Last week, in You Can’t Tax Scrooge McDuck, I made three claims and granted one courtesy. The claims: a fortune is claims on goods, not the goods themselves — you can no more tax Scrooge’s idle hoard into a bridge than you can eat a photograph of lunch. A flat statutory rate on capital hides an effective bite that grows with the horizon, under every architecture: 23.8% on annually taxed gains becomes 37% of a working life’s consumption and two-thirds of a century’s, while the realization system we already have quietly converges, through inflation alone, toward taking that same 23.8% of an asset whose owner never got a dollar richer. And a billionaire who burns his fortune makes everyone else fractionally richer — which tells you exactly what the fortune was, and was not, taking from you. The courtesy: I granted, and still grant, that the concern for the poor behind the wealth-tax proposals is entirely sincere.

This week the economists get their turn: the famous zero, the corrections that bent it, the strongest wealth-tax model in the literature, the one argument I cannot refute, what Sweden learned at the price of an estate — and, at the end, two Berkeley economists whose errors all point in the same direction. The compounding wedge, stated in plain English, is the famous result that economists argue about. So let’s argue about it.

What Chamley and Judd actually proved (and what they didn’t)

In the mid-1980s, Christophe Chamley and Kenneth Judd independently proved a result the rest of public finance would lean on for a generation: in a standard growth model, the optimal tax on capital income, in the long run, is zero. The intuition is exactly the compounding wedge I walked you through last week. A constant capital tax is an ever-growing tax on consumption the further out you look, so taxing capital is a spectacularly inefficient way to raise revenue compared with taxing the thing you’re trying to fund more directly.

For years this got flattened, by people who liked the conclusion, into a bumper sticker: the optimal capital tax is zero, science says so. It deserved puncturing, and the serious opposition punctured it. But watch what actually burst: the science says so, not the zero. Zero might still be the answer — whether it is, or whether the honest answer is merely low, is the question this whole post is chasing. Peter Diamond and Emmanuel Saez (about as credentialed as optimal-tax theorists get) argued in their “Case for a Progressive Tax” that capital income should be taxed, and that the Chamley-Judd zero-tax result is simply not policy-relevant. They are the strongest people on the other side, and it is them you have to answer, not the cartoon version.

So meet them. Notice what they actually deliver: a direction, not a magnitude. “Capital should be taxed positively” — but at what rate? The number, in their framework, hangs on the long-run elasticity of capital with respect to the net-of-tax return, and that elasticity is, in the literature’s own word, elusive — nobody has pinned it down. So the rate is left conveniently unspecified — hold that thought, because we will meet the genre it belongs to before the end. Diamond and Saez, to their credit, show their work and accept the constraints.

Then in 2020 Ludwig Straub and Iván Werning came at the question from pure theory, and “Positive Long-Run Capital Taxation: Chamley-Judd Revisited” did the field the larger service: they showed the zero-tax conclusion doesn’t survive inside Chamley and Judd’s own models. Under some preference specifications the optimal long-run capital tax is positive and significant; under others it converges to zero only after centuries. And they explicitly reject the argument that the ever-growing consumption-tax equivalence (my asymptote, in its formal dress) settles the question against capital taxation by itself. Their result does not repeal the wedge; it establishes that the wedge is one term in a welfare calculation, not the whole calculation. That sharpens the question rather than weakening it. The real fight is quantitative: how large are the countervailing forces in the economy we actually inhabit? That is a fight about magnitudes, which is where this essay has been trying to drag the debate all along.

The benchmark holds one more result, and this one argues for taxing wealth — which is exactly why it belongs here. Inside the model, the existing stock of wealth is the one thing that can be taxed without distortion — the past cannot respond retroactively — so a truly one-time levy on wealth already accumulated is the closest thing public finance has to a free lunch, and Diamond and Saez say so. The catch, which they also say, is the phrase one-time. A government cannot credibly promise the levy won’t recur. The first once-in-a-lifetime wealth tax is politically indistinguishable from the first installment of a series, and the moment savers expect a second helping, the lump-sum magic evaporates and you are back to taxing future saving with all the compounding machinery of last week’s essay. The free lunch exists only if no one believes you will ever eat it again — and no one ever believes that. Ask the citizens of any country that has ever levied a “temporary” income tax.

What the people who cheer all this miss — and it is the hinge of the entire post — is that positive is not the same as high.Diamond and Saez want capital taxed; Straub and Werning proved zero isn’t forced; incomplete-markets models find a positive optimum too. Grant every one of them. None of it repeals the wedge — it locates the exceptions, the places where the clean benchmark is too clean: finite lives, borrowing constraints, earnings uncertainty, the arbitrage problem we’ll get to. And with one exception, which I will meet head-on in a moment, none of those corrections pushes the optimal rate to where we tax labor, let alone to where the wealth-tax enthusiasts want to go. The disagreement among people doing the actual accounting is over how far above zero a low number sits. It is rarely between “low” and “tax it like a paycheck.” The economists are fighting over the first ten yards of the field. The political proposals are standing in the parking lot.

Now the exception, because it exists and it deserves better than a footnote. In 2023, Fatih Guvenen and coauthors published “Use It or Lose It” in the QJE, built on a fact the standard framework assumes away: returns differ persistently across people. Some entrepreneurs compound at twenty percent for decades; some heirs earn the T-bill rate on a fortune they didn’t build. Under a capital income tax, the productive entrepreneur pays more precisely because she produces more. Under a wealth tax, both pay the same on the same fortune, which shifts the burden toward the unproductive holder, nudges idle fortunes toward whoever can use them, and in their calibration produces a positive optimal wealth tax, at rates in the neighborhood of the real-world proposals, with real efficiency gains. Use it or lose it. With persistent return heterogeneity, a wealth tax can be a selection device. It is the strongest sophisticated case for a wealth tax. The theorem stands. Implementation is another matter. The tax authority observes asset values, not entrepreneurial productivity, and valuation discounts, leverage, private-company appraisal games, and plain mobility scramble the intended selection badly — a tax designed to dislodge unproductive owners tends in practice to dislodge the most liquid and transparent ones, which is roughly the opposite of the intent. Sweden, as we’ll see shortly, ran that experiment. And one twist from inside their own model: the reason wealth taxation looks attractive there is that capital-*income* taxation is significantly more distorting — the best pro-wealth-tax paper in the literature agrees that taxing the return is the thing you must not do. The wedge survives even in the model built to escape it.

The newest entrants — a cluster of automation-era optimal-tax preprints, Nakatani’s among them — mostly reinforce the pattern while demonstrating its fragility: depending on the version, the instrument mix, and the substitution elasticities you feed them, the same modeling family produces optimal capital rates anywhere from near zero to the low thirties, alongside a robot tax that keeps coming out at or near zero. I cite none of it as proof of anything. The instability is itself the instructive part: the optimal number is wildly model-sensitive even where the intertemporal wedge is not — which is the entire thesis of this essay.

That’s the thing about a conservation law. You can come at it from any direction you like. It’s still there when you arrive.

So why isn’t it zero?

Because of the single best argument for taxing capital — an argument I want to hand you at full strength, because it is the reason the true number is low rather than literally nothing. I owe the sharpest version of it to Tyler Cowen:

I used to favor a zero tax rate of capital, but I no longer hold that view. There are too many options for reclassifying labor income into capital income and thwarting the purposes of the tax system altogether. Nonetheless, subject to this constraint, I think taxes on capital should be as low as possible.

It goes like this. Suppose you set the capital rate to zero while taxing labor at forty percent. You have just handed every high earner who can plausibly incorporate a standing invitation: stop calling your income wages and start calling it a return on your equity. Form the LLC, pay yourself a distribution instead of a salary, and watch your tax rate fall by half. The line between labor income and capital income is not a wall of nature; it is a lawyer’s construction, and a zero capital rate next to a high labor rate puts an enormous bounty on erasing it. The labor tax base would hemorrhage from the top down. So you need a positive tax on capital, not because the optimal-tax mathematics wants one (it wants roughly zero), but as a fence around the labor tax: high enough that relabeling wages as returns isn’t worth the lawyer, the LLC, and the audit risk.

This is a serious argument. It is also, note, an argument I made myself, years ago, as the one caveat to the two-extremes case — that absent some rule or some small tax, workers will incorporate and rebrand their paychecks. Cowen states it better than I did. But the fence argument delivers less than it seems, so let me be exact about what it does deliver. It sets the rate from below — it does not, by itself, set it low. How high the floor sits depends on how cleanly the law can separate wages from returns: tight rules on owner-manager compensation, payroll-tax integration, and organizational form keep the floor low; sloppy rules, next to a forty-percent labor rate, can push it uncomfortably high, because the fence must be tall enough to beat the lawyer. What the argument does establish, unconditionally, is the purpose of the positive rate: it exists to protect the labor tax base, not to soak the pile. A fence, not a confiscation — and the fence’s height is a question of legal drafting, nothing grander.

Now look where that lands. The optimal-tax literature sets the ceiling: it starts from the benchmark’s zero, and the corrections (finite lives, incomplete markets, Straub-Werning) raise the defensible rate by amounts everyone agrees are contested and most agree are modest — so the ceiling sits low. The arbitrage argument sets the floor: the rate cannot sit at zero without the labor tax base leaking out through the LLC door — so the floor sits above zero, at a height that depends on legal drafting. Neither bound is a single number. But look at what both directions agree on, because it’s the thing the proposals deny: the rate is set by specific, nameable mechanisms (the wedge, the fence, the exceptions) and not by the size of the pile. Nothing in either direction says “there sits a great fortune; take a bite because it is there.” The bite-because-it-is-there theory is the one theory with no model behind it. That is the convergence, and it is enough.

The inversion

Now put the pieces together, and watch the wealth tax turn inside out in your hands.

The intuition that animates the tax is that the rich are sitting on a pile withheld from everyone else, and the tax claws some of it back for the common good. Every step of that is wrong. The pile, to the extent it’s a genuine hoard, is already a gift — withheld from no one, because unexercised claims feed everyone else’s claims exactly as a bonfire would. And to the extent it isn’t a hoard, it isn’t a pile at all — it’s deployed capital, the productive machinery whose returns set tomorrow’s wages, and the tax is a compounding wedge driven straight into it.

So this is what the tax actually does, granting its champions every ounce of the sincerity I promised them. It reaches past the harmless, self-donating hoard — which it can’t really touch and wouldn’t help anyone by touching — and it lands on the deployed capital instead. It shrinks the capital stock over time (how much depends on saving elasticities, capital mobility, and where the revenue goes, and the models disagree about the size), and a thinner capital stock pays lower real wages, and lower real wages fall hardest on exactly the people who own no capital and have nothing but their wages: the future poor. The size of the shift is model-dependent; its existence is not, and it is not zero merely because the tax bill bears a billionaire’s name. The policy aimed at the pile misses the pile and hits the future low-wage worker, through the one channel, the wage-setting power of the capital stock, that the champions of the tax never put on the slide.

And now the other end of the pipe, which never makes the slide either. Capital is deferred consumption; that is the premise this whole pair of essays runs on. So look at what the revenue actually buys. Federal R&D, the closest thing the budget has to a line that builds the future, runs around three percent of outlays, and roughly half of that is defense. The overwhelming mass of federal spending is Social Security, Medicare, Medicaid, other transfers, and interest on the debt. Present consumption, nearly all of it. That is no accusation — consumption is the entire point of an economy; the only reason to defer any of it is to do more of it later. I am not calling Social Security waste. I am calling the accounting dishonest. The machine takes deferred consumption in at one end and pays present consumption out at the other, at a compounding loss in between, and the loss lands on whoever will be doing the deferring later. That is not a metaphor. It is a description of the two ends of the pipe, available from the outlay tables. Pretending the trade is free — that you can move consumption from the future to the present and pay nothing in foregone growth — is the shared vice of the whole debate, left and right alike, and the most popular defense of the capital tax, that the money buys something that grows, is an argument about three cents on the dollar.

The exceptions are real, and they get their due. Some transfers really are investment: the work on the marginal value of public funds finds certain child-directed programs with returns high enough to pay for themselves several times over. And most public capital that does grow the future (roads, schools, universities) is built at the state and local level, not by the federal government a wealth tax would fund. And yes, revenue is fungible: the marginal dollar wears no name tag, and a capital tax funding a high-return childhood program is a different bargain from one funding undifferentiated present consumption. But composition is destiny in expectation: when ninety-seven cents of the existing dollar buys present consumption, the sensible prior on the marginal dollar is not “this one’s different.” The federal base is old-age transfers and interest, and those are present consumption however worthy.

There is a joke buried in the ledger, too. The fastest-growing line in the federal budget is net interest — compensation for lending, yes, but follow the composition: it flows to bondholders, which is to say to the holders of capital. The state is already writing checks to the very people the wealth tax is aimed at, and it is writing bigger ones every year.

The billionaire with a match makes everyone fractionally richer. The capital tax makes the future poor poorer, in the name of helping them. The bonfire is a better philanthropist than the tax bill, and it isn’t close.

One real complication remains, and it deserves full strength. Some top-end wealth really is low-velocity: art on a freeport wall, raw land held purely for appreciation, the empty trophy apartment. That looks more like the bin than the factory, and there the harmlessness argument is weaker. But two things. First, it’s a small slice of the relevant base; the overwhelming mass of great fortunes is equity in operating companies, not Picassos in a vault. Second, even on that slice, a wealth tax mostly distorts the composition of what people hold rather than raising clean revenue. And the one genuine exception, the asset whose supply truly can’t flee the tax, is land, which is the entire Georgist insight and has been sitting in plain sight since 1879. If you want to tax something inert and unproductive without the compounding catastrophe, tax land. That is a serious argument. “Mark Elon’s shares to market every April” is not that argument, and the people making the second one rarely have the courage to make the first.

What Sweden learned the hard way

One more objection needs answering, because it is the reflexive one, and a certain kind of American politician reaches for it constantly: but Scandinavia. The Nordic countries are held up as proof that you can soak capital and wealth and thrive. So look at what they actually did, as opposed to what they are imagined to have done.

They are not socialist. They are open, capitalist, free-trading economies that happen to run high taxes on income and consumptionto fund generous welfare states — which is a real choice with real trade-offs, but it is not a tax on capital. And on the specific question of taxing accumulated wealth, the region ran both arms of the experiment, which is more useful than either alone. Sweden walked away entirely: it abolished its inheritance and gift tax in 2004, by unanimous vote of the Riksdag (when the Boxing Day tsunami killed hundreds of Swedes days after the vote, the repeal was even backdated so that the victims’ estates would escape it) and abolished its wealth tax three years later, in 2007, reforms begun under a Social Democratic government and finished by the center-right one that followed. Norway abolished inheritance tax in 2014 but kept its net wealth tax, roughly one percent a year at the top, and has spent recent years watching a well-publicized procession of its wealthiest founders decamp for Switzerland, taking their tax bases with them. Denmark alone of the Scandinavian three still levies an inheritance tax; Norway alone still levies a wealth tax, and Norway’s experience with it is the cautionary half of the tale. The supposed socialist heartland ran both arms of the trial: the country that repealed is fine, and the country that kept is the case study in exit.

And they repealed it not on theory but on a corpse. Sweden’s inheritance tax kept doing the precise thing the deployed-capital argument predicts: because great fortunes are equity in operating companies, not coins in a vault, the tax repeatedly forced heirs to strip liquidity out of working businesses, or dump shares, just to pay it. The canonical case is almost too on-the-nose. When Sally Kistner (widow of Erik Kistner, the largest individual shareholder in Astra, the pharmaceutical company that would one day become half of AstraZeneca) died in 1984, her roughly 300-million-krona estate was mostly Astra stock. The top inheritance rate of 65% applied, and because the heirs would have to sell shares to pay it, a capital gains tax of almost 26% stacked on top: a combined bill near 90% before a single share had moved. Then the shares moved. The market understood the heirs would have to dump a large block, anticipated the flood, and the price fell — and the falling price dragged the total tax bill above the entire value of the estate. The estate was declared insolvent. The tax was larger than the thing it taxed. And note what did not break: Astra sailed on — it’s the Astra in AstraZeneca — because the company was never the problem. What broke was the estate, the family’s holding, and the incentive of anyone watching to keep concentrated ownership of a productive Swedish company inside Sweden. “Taxing the pile” never touched the factory. It destroyed the holder, and it taught every other holder the lesson.

And it didn’t just eat that estate; it exported the next one. When Fredrik Lundberg, today one of Sweden’s wealthiest industrialists, emigrated to Switzerland in 1985, he pointed to the Kistner case by name as his reason for leaving. He was not the only one; Ingvar Kamprad of IKEA and the Rausings of Tetra Pak had already gone, taking their companies’ futures with them. The tax didn’t redistribute the pile. It chased the pile out of the country and bankrupted the pile that stayed. The Swedes watched it happen often enough that the left decided it wasn’t worth it.

That is the mechanism above, running live. One ruined estate and a procession of emigrants do not estimate an aggregate counterfactual — case evidence never does — but the aggregate estimate now exists, and it is worth quoting both halves. Jakobsen, Kleven, Kolsrud, Landais, and Muñoz, using Scandinavian administrative data, find that a one-point rise in the top wealth-tax rate shrinks the stock of wealthy taxpayers by about 2%, and that the businesses of the owners who leave are measurably damaged. The exit is real, and it takes firms with it. The authors call the aggregate losses modest: two-hundredths of a percent of employment, a tenth of a percent of value-added, per percentage point of tax. Put a ruler on that adjective before accepting it. A tenth of a percent of value-added is, at American scale, roughly thirty billion dollars, and it recurs every year the tax stands, growing to the extent the missing investment compounds the capital stock downward. Set that against what the tax would actually collect (Summers and Sarin’s twelve to forty cents on the advertised dollar) and the recurring loss is the same order as the realistic revenue. Transplanting Scandinavian elasticities to America is heroic, so hold the number loosely; but hold it, because it is the honest one. The direction is theirs and it is certain. The adjective is theirs too. The compounding is ours.

Why they sweep it away

Which brings us, finally, to the rug.

None of the findings above is fringe. The wedge is freshman arithmetic; Chamley-Judd is in every graduate textbook; the corrections are published in the field’s best journals; and what they jointly establish is not a number but a burden of proof: each capital tax must justify itself through a specific, nameable mechanism, because the one theory none of them supports is take-a-bite-because-the-pile-is-there. Is the optimal rate exactly zero? Almost certainly not. But it is far closer to zero than anyone who loves capital taxation will say out loud, and the one model that reaches proposal-scale rates depends on the tax authority observing what it cannot observe. So why does the political conversation price capital taxation as if the optimum were high and the mechanism question settled, when the literature says neither? Ask the proposers which model backs their rate. Go ahead. I’ll wait.

Crickets.

And when I say the policy comes first, I am not guessing at anyone’s soul. I am reading the residuals. Any hard empirical problem forces dozens of judgment calls — which income concept, whose incidence assumptions, what counts as a transfer — and an honest researcher’s calls scatter: some flatter the thesis, some wound it. Random errors have no favorite direction. So look at the published record of Emmanuel Saez and Gabriel Zucman, the Berkeley economists whose numbers carry the wealth-tax movement. Their top-income series has been reconstructed from the same tax data by Gerald Auten of the Treasury and David Splinter of the Joint Committee on Taxation and comes out far flatter once taxes and transfers are counted. Their book-length headline — billionaires pay lower rates than secretaries — required excluding the refundable credits that do the most work at the bottom of the distribution, and reversing the corporate-incidence convention their own QJE paper had used the year before. Two switches, both moving the headline the same way. Their revenue score for the Warren wealth tax — a proposal they themselves developed, advised, and, in Lawrence Summers’s own words, “played a major role in developing and validating” (author, advocate, and referee in a single chair) came in, per Summers and Sarin, at somewhere between a tenth and four-tenths of the advertised figure once valuation, avoidance, and enforcement were priced in. Each call, taken alone, is defensible, and their fights are live — Piketty, Saez, and Zucman have rebutted Auten and Splinter, and the rebuttal is now itself being defended — so I am not declaring a winner on the measurement. I am pointing at something that requires no winner: the sign. Every disputed judgment call leans toward the conclusion. In my trade that is called a systematic error. Random mistakes cancel; these compound, and always in the same direction. None of this requires bad faith. I granted sincerity at the top of this pair and I am not taking it back, because motivated reasoning does not feel like lying from the inside; it feels like being right. But the direction of a bias term is not a feeling. Twenty judgment calls, twenty heads. Draw your own conclusion about the coin.

Because the result is inconvenient for the stated goal itself, and that is the one thing harder to look at than an unfair tax. An honest redistributionist could say “tax capital heavily, accept the efficiency cost, call it fairness” — and reasonable people could weigh that trade. But the arithmetic doesn’t offer that trade. It says the heavy capital tax makes the future poorer, and the future poor poorest of all. There is no billionaire-versus-everyone-else ledger to balance here; there is only a present-versus-future one, and the tax is on the wrong side of it. So the result goes under the rug, and the rug is woven entirely out of good intentions.

I’ll grant the good intentions. I said I would, and I meant it. But good intentions are not a conservation law. They don’t suspend the arithmetic. A wedge driven into a compounding quantity grows without bound no matter how kind the hand on the hammer, and the future worker whose wages it lowers will not be consoled to learn it was done out of concern for him.

You cannot tax Scrooge McDuck. If his bin is full of idle gold, you cannot take anything from him at all: as Part 1 showed, his consumption was zero before the seizure and zero after, so beyond the psychic sting of a smaller pool to swim in, he is untouched. What changes is that the government now consumes more — and since the goods are the goods, everyone else consumes less. The tax on the idle man is paid, to the last dollar, by everyone who is not the idle man. If you carry one sentence out of this pair of essays, carry that one. If his bin — like every real fortune — is deployed in the world building the things the rest of us use, then taxing it is a slow tax on the future, falling hardest on the people with no bin of their own. Either way you miss the duck. The only question is who you hit instead.

It is almost always someone who couldn’t afford to be hit.

You Can’t Tax Scrooge McDuck is Part 1 of this pair — the duck, the arithmetic, and the bonfire. If you found the terminal-accounting move useful, it’s the same one I ran on the wealth-management industry in The Conservation Law of Tax Alpha — different pile, same conserved quantity, same trick for finding where it went.

This piece is one argument out of a book’s worth I’ve already written up — a whole manuscript of economics done this way, from first principles, following the arithmetic wherever it goes. If you happen to be a literary agent and any of this is your kind of thing, get in touch. I have no appetite for firing proposals into the void; the void, as far as I can tell, is managing perfectly well without them.

My book, The Science of Free Will, asks an equally uncomfortable question about an equally cherished story. For the India series, start with The Paradox of India and the full collection; for Britain’s long institutional decline, first seen in 1979, there’s Albion.

The pitch is always the same, and it is always seductive.

Somewhere out there is a pile. A great, glittering, obscene pile of money, and a small number of people are sitting on top of it doing nothing useful. The pile just sits there, the way piles do, while teachers buy their own classroom supplies and the bridge falls into the river. So we will reach into the pile — just once, just a little, just the tip of it — and we will use what we take to do some good. A wealth tax. A mark-to-market tax on unrealized gains. Capital gains taxed as ordinary income. The mechanism varies; the picture underneath it never does. There is a pile, the pile is static, and reaching into it costs no one anything except the people who have more than they could ever need.

I want to tell you why this picture is wrong. Not wrong in the sense that reasonable people disagree about the rate. Wrong in the sense that the central object in the picture — the static pile you can dip into without consequence — does not exist, and the thing that does exist behaves nothing like it. And I want to do it without the usual libertarian throat-clearing, because the usual libertarian throat-clearing is exactly what lets the people who love these taxes wave the whole argument away as ideology. So let me grant them everything. Let me grant that the inequality is real, that the need is real, and grant, charitably, that the concern for the poor is entirely sincere. None of that is the problem. The problem is that the instrument does not do what they think it does, and the cleanest way to see it is to try to tax a cartoon duck.

The duck

Scrooge McDuck has a money bin. You remember it: a building-sized vault filled to the rafters with gold coins, into which the richest duck in the world dives each morning for a swim. It is the platonic ideal of the pile. If you have ever felt the intuition that we should tax great fortunes, you have felt it about that money bin. There it sits. Tax it.

So tax it. Walk into the bin, take a fifth of the coins, walk out. What just happened?

Nothing happened. Or rather: nothing happened to anyone but Scrooge. The coins in the bin were not doing anything. They were not financing a factory, not paying a wage, not buying a machine, not building a house. They were a hoard — inert, idle, swum-in. Whatever harm you imagined that hoard was doing to the rest of us, it was not doing, because a pile of metal in a vault does not reach into anyone else’s life. You took a fifth of it and you accomplished precisely one thing: the pile is now smaller. The bridge is no closer to being fixed than it was, except to whatever extent the gold you carried out can be spent — and the moment you spend it, you’ve simply moved the pile, not unlocked any productivity that the pile was strangling. There was no strangling. It was a swimming pool.

Here is the part that should stop you, and it is an argument I owe to Steven Landsburg, who made it years ago in a piece called “The Man Who Can’t Be Taxed.” To tax someone is to lower their consumption — that is what bearing the burden of a tax means, in the currency this argument is denominated in, which is real goods and nothing else. (And to keep the point clean, treat the coins as what modern fortunes actually are — entries in a ledger — rather than gold, which is itself a real commodity you can fill teeth with.) If your consumption is unchanged, you have not been taxed in that currency, no matter what the ledger says. So watch what happens when the government empties the bin. Scrooge’s consumption before the seizure was zero; he was swimming in coins, not spending them. His consumption after the seizure is also zero; he’s just swimming in a smaller pool. His consumption did not change. So he bore no burden in consumption terms (other than swimming in a little less gold, which for the purposes of this argument we can consider negligible). But the government now has the coins and will spend them — which means the government consumes more, which means, in an economy that is using its resources, someone else must consume less. And it cannot be Scrooge, because his consumption was zero and stayed zero. So the burden of “taxing the idle rich man” falls not on the idle rich man but on the rest of us, whose share of real goods the government’s new spending bids away. You did not tax Scrooge McDuck. You taxed everyone who isn’t Scrooge McDuck, and sent him the bill addressed to someone else.

That argument runs in a stripped-down model where the only thing wealth is for is consumption — and I want to keep the toy model honest, because the real world is richer. A smaller bin might cost Scrooge something the toy model can’t see: the option to spend later, status, security, political heft, the bequest he meant to leave, the sheer pleasure of the swim. Those are real, and a wealth tax can reach them even when it can’t reach his dinner. But notice what the toy model isolates, which is the whole point: it separates real consumption from accounting claims, and it shows that to the extent a fortune is the latter and not the former, taxing it doesn’t free up goods for anyone — it just moves the claims, and the burden lands elsewhere.

Which means that to whatever extent a real Scrooge McDuck exists — someone sitting on idle wealth, consuming nothing — he is not a problem to be solved. He is the single most harmless rich person imaginable. He produced something, or holds claims on something others produced, and in exchange he has taken nothing out of the common pool. He doesn’t eat your lunch. He doesn’t bid up your house. He doesn’t consume the goods you wanted. His fortune is a number on a ledger and a pile of metal he likes to look at. In pure consumption terms, the miser on his hoard has donated his entire economic output to the rest of us and asked for nothing back. He is, whether he means to be or not, a philanthropist — and you cannot tax a philanthropist by confiscating the gift he has already given.

We will come back to that, because it turns out to be the whole argument. But first the obvious objection, which is correct: real billionaires are not Scrooge McDuck. There is no bin. Elon Musk does not have a swimming pool of dollars. And that is exactly the point, because it means the thing you are actually proposing to tax is not the bin.

There is no bin

Real wealth at the top is not a hoard of coins. It is claims on productive enterprise — shares in companies, stakes in firms. Those claims are not identical to the machines, buildings, software, and organizational knowledge beneath them; a trillion dollars of market capitalization is not a trillion-dollar warehouse of lathes, and equity value bundles factories together with rents, patents, growth expectations, and control. But the claims and the machines are coupled, and the coupling is where all the action is: the valuation of the claims sets the cost of raising capital, which sets what gets financed, which sets the capital each worker has to work with, which sets — more than almost anything else — what those workers get paid. A recurring tax on the claims does not scoop coins from a vault. It changes the terms on which productive assets are financed, owned, and expanded, and the wage effects arrive through that chain. (Someone always objects here that buying a share on the secondary market funds nothing — the “it’s only paper” move. A recent working paper says otherwise, through a channel almost nobody expected: Sammon and Shim find that when index funds buy, the sellers on the other side are, roughly one-for-one, the firms themselves — issuing new equity, largely through stock-based compensation — so two decades of passive buying has financed corporate equity issuance. The paper claims and the real machines are joined more tightly than the objection assumes; the full story deserves its own post.)

This matters because it splits the wealth tax in two, and the tax misses on both halves.

If the wealth is in fact hoarded — the actual bin, idle and consuming nothing — then taxing it releases nothing. You are taking from the one person who, by construction, took nothing from anyone, and the seizure conjures no bridge, no nurse, no machine — it relabels who holds the tickets.

And if the wealth is deployed — which is the real case, the overwhelming case — then it isn’t a bin at all. It’s the capital stock. And now your tax is not a one-time dip into a static pile. It is a recurring wedge driven into the exact machinery that builds future wages. Every year the capital exists, the tax bites again. And that recurrence — not the unfairness, not the politics, the plain arithmetic of a wedge that compounds — is the thing nobody prices in.

Let me show you the number, because this is where assertion turns into arithmetic.

The terminal accounting

One clarification before the arithmetic, because it governs everything after it. The question is never whether to tax. The government is going to spend what it spends, and the revenue has to come from somewhere. What every serious model in this literature actually asks is narrower: which instrument raises a given sum at the lowest cost — capital, labor, or consumption? So nothing below is a claim that taxing capital is bad and taxing nothing would be better. It is a claim that of the three doors available for raising the same dollar, one of them opens onto a wedge that compounds and the other two do not. That is a narrower argument than the one usually made in my direction, and it is the one that survives contact with someone who does this for a living.

One more piece of scope, so nobody gets ambushed later. My target is recurrent taxation of the normal return to deployed capital — annual mark-to-market taxation, recurring wealth levies, and their accrual cousins. Other instruments are different animals with different arithmetic, and I will keep them separate — fold them together and “capital taxation” becomes a suitcase into which every policy dispute has been stuffed. A one-time levy on existing wealth gets its reckoning next week, land and rents get theirs there too, and the realization-based tax with deferral — which is what America actually has — stays labeled as such throughout. The compounding objection applies most directly to the normal return required to make deferring consumption worthwhile. It does not carry over unchanged to land rents, monopoly profits, or windfalls.

Capital is deferred consumption. This is not an ideological statement; it is what capital is. A dollar of capital is a claim on future goods — goods you chose not to consume now so that you (or your kids, or whoever holds the claim later) could consume more of them later. When you tax capital, you are taxing future consumption. You are reaching forward in time and taking a slice of the goods someone was saving for.

Now run the numbers on what a flat, modest, annual rate does to that future slice — and one piece of precision first, because the precision is the point. The arithmetic that follows describes an annual tax on capital as it accrues: a mark-to-market tax on unrealized gains, a wealth tax, an annual levy on capital income. It is not the ordinary U.S. capital gains tax, which is charged once, at sale, and lets the gain compound untaxed until then. But the annual, accruing form is the architecture of the marquee proposals — Senator Wyden’s plan would mark tradable assets to market and tax the gain annually; the Treasury’s billionaire-minimum design is a gentler cousin that treats the annual payment as a prepayment credited at eventual sale — and it is the annual form, not the realization form, where the compounding catastrophe lives. So I’ll use today’s top capital-gains-plus-NIIT rate as a familiar number, not as a claim that gains are taxed annually today. They are not. The proposals want to change that.

Take that number: twenty percent, plus the 3.8% Net Investment Income Tax, for 23.8% all in. Nobody calls it confiscatory; the people who want to mark gains to market annually consider it gentle. Now levy it every year on a real return of five percent — which is what “tax unrealized gains” means — and ask the only question that matters: of the future consumption you were actually saving for, how much does this “gentle” annual tax remove?

At a one-year horizon, almost nothing — about one percent. Fine. But you don’t save capital for one year; you save it for a life, for a retirement, for a generation. At ten years the tax has eaten about 11% of your future goods. At twenty years, 20%. At thirty years, 29%. At forty years — a working life — the headline rate of 23.8% has become an effective tax of 36.6% on the consumption you were deferring. At fifty years, 43%. And over a hundred years — the horizon a foundation or a family office plans across — a “gentle” 23.8% annual rate becomes a 68% tax on the goods at the far end. The saver was told the government takes 23.8% of his gains. The terminal accounting says the government takes 37% of a working life’s goods, and two-thirds of a century’s. Two words about what that 68% is, because this is where a critic will aim: the state does not collect 68% of the year-hundred account. The statutory rate stays 23.8% of each year’s return, forever; what the accumulated wedge removes is 68% of the consumption available in the untaxed counterfactual. Your wealth still grows. It just grows into a third of what it would have been. (These are also partial-equilibrium numbers. In general equilibrium a smaller capital stock pushes the pre-tax return up, so the individual saver’s realized burden is somewhat lighter than this — in part because some of it has been shifted onto labor, through the lower wages a thinner capital stock pays. How much shifts is model-dependent, but the direction of the correction makes the incidence worse for workers, not better — which is next week’s essay arriving early.)

A recurring wealth tax runs the same machinery with a different dial, and the translation is worth doing because the dial is deceptively marked. A 1% annual tax on wealth is not “only one percent”: against a 5% real return, it absorbs a fifth of the return, every year, whether or not the return shows up. Run it out and a portfolio compounding at 4% instead of 5% lands about 32% below the untaxed path at forty years and 62% below it at a century. The proposals on the table start at 2%, which against the same return is a 40% annual tax on the return before compounding — and the wealth tax collects it in loss years too, which the income tax at least has the manners not to do.

The headline rate never moved. The rate that actually lands climbed the entire way up. Why? Because each year’s tax is itself a brick pulled out of the wall that compounds. You don’t merely lose this year’s skim — you lose every future dollar that this year’s skim would have earned, and every dollar those would have earned, forever. The flat rate is conserved. The thing it taxes grows exponentially in the horizon. So the effective bite has to grow to match. (Readers of The Conservation Law of Tax Alpha will recognize the move: find the quantity that’s conserved, then do the accounting all the way to the end. The fund managers hide the conservation law in the fee structure. The state hides it in the compounding.)

Now the line that should reframe the entire debate. That climb never stops. Run even the gentlest rate anyone seriously proposes — fifteen percent — out far enough and it asymptotes to a one hundred percent tax: about 51% at a century, 97% at five hundred years. For any positive rate, the effective tax on consumption N years out climbs toward total confiscation as N grows. No flat capital tax is gentle enough to spare the far future; give it enough time and it takes essentially all of it. That is not rhetoric. It is the geometry of compound interest, and it is the same for a Marxist and a monk.

One limit on what the asymptote proves, before anyone proves it for me: it establishes persistence, not optimality. A one-percent management fee also compounds toward eating everything, given enough centuries — readers of the Tax Alpha piece watched me run exactly that accounting on the wealth-management industry — and the conclusion there was not that fees must be zero. It was that you must price the compounding before you sign. Same here. The asymptote tells you the wedge never stops mattering; it does not, by itself, pick the rate. Picking the rate still requires weighing everything else — redistribution, risk, what the revenue buys, what the alternative taxes cost. The asymptote is not the verdict. It is the invoice.

The same compounding that makes a flat rate confiscatory over time also dismantles the most popular argument for capital taxes — the Warren Buffett line, the one about how it’s a scandal that the billionaire pays a lower rate than his secretary. The sharpest demolition is again Landsburg’s, in a post called “Getting It Right.” Take Alice and Bob (physicists always reach for Alice and Bob). Both earn a dollar of wages and both pay a 50% wage tax, leaving each with fifty cents. Alice spends her fifty cents immediately. Bob invests his, it doubles to a dollar, and then he spends it. With no capital gains tax, has Bob gotten away with something? No. Alice consumed fifty cents out of a potential dollar — a 50% bite. Bob, absent the wage tax, would have invested the full dollar and ended with two; his one dollar of actual consumption is also exactly 50% of his potential. Identical. The two are taxed at the same rate, because the only difference between them was when they chose to consume. Now add a 10% capital gains tax — levied, as it actually is, on the gain. Alice is untouched, still at 50%. Bob’s fifty cents doubles to a dollar, he pays ten percent on his fifty-cent gain — a nickel — and is left with ninety-five cents against a potential two dollars: an effective bite of 52.5%. The capital gains tax didn’t equalize Alice and Bob; it taxed Bob more, purely for the crime of waiting. The wage tax already shrank the principal both of them started from; the capital gains tax adds a second wedge that binds only on the one who postponed. The secretary-versus-billionaire comparison counts the second wedge and forgets the first — it mistakes the patient saver for a tax dodger. (None of this proves capital income must never be taxed; Bob’s return might contain luck, rent, or relabeled wages, which is exactly where next week’s essay goes.)

The same timing problem haunts the estate tax, and here, for once, the story is messy, so let me name the mess. Part of any estate is income that was already taxed in the hands of the person who earned it; taxing it again on transfer is a straightforward double-count. But part of an estate is unrealized gains that were never taxed at all — and under current law the heir’s basis is stepped up to market value at death, so those gains escape income tax entirely. That cuts both ways against the simple story: the estate tax is partly a double-count and partly the only tax those unrealized gains will ever face. The right fix isn’t to add an estate tax on top — it’s to kill the step-up, so gains are taxed exactly once, instead of either twice or never. Taxing gains once and taxing inheritance as such are two different aims, and stuffing them into one instrument hides both. (That, too, is the buy-borrow-die machinery from the Tax Alpha piece, viewed from the far end of a life.) Either way, the lesson holds: a tax on accumulated saving is usually a tax on consumption deferred across a life, or across generations, not a painless dip into a pile.

One more wedge before we leave the arithmetic, because this one belongs to the system we actually have, and I don’t want anyone walking out of this section feeling smug about realization. The current tax is levied on nominal gains. Nobody indexes your basis for inflation. So consider the most boring asset imaginable: one that merely keeps pace with the CPI, forever — zero real return, the owner never one dollar richer in purchasing power. Under a nominal-gains tax there is no such thing as “no gain.” At 3% inflation, sell after thirty years and the tax takes 14% of the asset’s entire value; after fifty years, 18%; hold long enough and the take converges to the full 23.8% — of the asset, not the gain, because at long horizons the basis shrinks toward irrelevance and the “gain” converges to the whole thing. A capital gains tax on a zero-real-gain asset is a slow confiscation of principal, at a rate that climbs with the holding period toward the headline rate itself. And for assets that do earn something real, the same mechanism silently marks up the rate: at 2% real and 3% inflation over thirty years, the 23.8% headline is an effective 41% on the gain you actually made. Same signature as before — a flat statutory rate hiding a bite that grows with the horizon — different channel: the proposals’ wedge compounds through the return; the current system’s compounds through the price level. Giesecke and Nassios just ran this arithmetic for Australia, whose effective top rate of 23.5% sits within rounding of ours by pure coincidence, and they added a design warning worth pocketing: half-indexation — recognizing real gains but real losses only when they are also nominal losses — can be worse at thirty-year horizons than not indexing at all. The moral generalizes, and it completes the fix from a moment ago: kill the step-up and index the basis — real gains, taxed exactly once. Anything else taxes either twice, never, or the inflation. Tax the gain, not the price level.

The bonfire

So the bin doesn’t exist, and to the extent it does, its contents are harmless — you cannot tax the hoard, and its mere existence is already a gift. But the intuition is bin-shaped, and it will not die until you follow it all the way down. Landsburg takes the argument to the vault. Let me take it to the match.

Suppose a billionaire wants — really wants — to give his entire fortune to the rest of the country. Not to a foundation, not to a cause, not to a board of trustees who will summer in Aspen deciding how to disburse it. To everyone, directly, with zero overhead and zero leakage. How?

Simple. Liquidate the entire net worth, convert all of it to dollars, and incinerate the cash. Strike a match. There you are. (Grant me the liquidation as a premise — yes, selling that much moves prices, and someone is on the other side of every trade. The toy model is about what happens to the claims once they’re cash, so stipulate the conversion and let’s move on.)

Stay with me, because this is rigorous, not a joke — or rather, it’s both. Money is not wealth. Money is a claim on wealth: a ration ticket entitling the bearer to some slice of the real goods and services the economy produces. The actual wealth is the goods themselves — the food, the housing, the machines, the labor. Dollars just determine who gets which slice. So when our billionaire burns his fortune in cash, he destroys his claims while leaving every real asset on earth exactly where it stood. The factories still hum. The farmland still grows wheat. Not one good has been destroyed. He has simply torn up his own ration tickets and walked away from the table.

And because the same pile of real goods is now divided among a slightly smaller stock of dollars, everyone else’s dollars are worth fractionally more. He has made a gift to every holder of the currency, in proportion to what they hold, via a whisper of deflation — no Treasury in the middle, no administrator skimming, no application filed. (This is a fixed-money thought experiment; a central bank can offset the deflation, and we’ll deal with the Fed in a moment. The toy model is here to isolate one fact, not to model the plumbing.) Every dollar-holder on earth gets a sliver, the instant the ash cools.

That is the cleanest charitable act available to a very rich man, and it requires no Giving Pledge, no press release, no naming rights on a hospital wing. Just a match.

Now hold that next to the hoarder, and watch the two nearly collapse into one. The hoarder’s unexercised claims do the identicalfavor to everyone else’s claims that burning them would — today, and every day the hoard stays a hoard. In neither case is he pulling real goods out of the common pool. The one difference between them is tomorrow: the hoarder keeps the option to march back to the table, and the burner does not. A concert ticket sitting unused in your pocket and a concert ticket torn up differ in exactly one respect — until the doors close, you might still go. So the exact statement is this: the vault and the bonfire are the same act in the present tense, and they stay the same act for every year the vault stays shut. Which still leaves the question the wealth-tax intuition can never answer: if a billionaire torching his entire fortune is a pure gift to the public, what exactly was that fortune takingfrom anyone while it sat unexercised? Holding claims you don’t cash is, for every year you don’t cash them, identical in real terms to having handed them out. The thing the tax imagines it is recovering was never being taken in the first place — only, at most, reserved.

There’s even a wrinkle that shows how specific this transfer is, in a way a blunt tax is not. The gift goes to whoever holds the currency you burn. Incinerate dollars and you’ve made a gift to dollar-holders — disproportionately Americans, but also every foreign central bank and saver sitting on Treasuries and greenbacks. Burn a globally-held currency instead of a domestically-held one and you change who receives your gift. The incidence isn’t a mystery; it follows the denomination of the ash. Subject to the central bank’s response, you can in principle steer who benefits just by choosing what you set alight — which is more than you can say for a tax, where the incidence is whatever the equilibrium decides, not whatever the legislator intended.

I’ll be honest about the size, because overclaiming here would be the same sin I’m accusing the other side of. The gift from any single bonfire is tiny and diffuse — even a vast fortune is a rounding error against the money supply, so the deflation is microscopic, smeared across hundreds of millions of holders. (You might object that a central bank targeting inflation would simply reprint to offset the deflation. But the Fed isn’t targeting the man with the match — it’s steering an aggregate, and whatever it prints flows to whoever it transacts with, in proportions having nothing to do with who burned what. That’s a separate transfer with its own incidence; it does not reach into the ash and hand the gift back. In the limiting case of a complete offset, the recapture accrues as seigniorage to the consolidated government — our billionaire would, in effect, have mailed his fortune to the Treasury by the scenic route, which is a worse joke but still a transfer. In practice the question never arises, because the burn sits so far beneath the Fed’s measurement noise that the objection is hypothetical: not a drop in the bucket, a drop in the reservoir.) So the claim is not that burning money is a large transfer. It is that the transfer is clean, leakage-free, and real — and that its smallness is of a piece with the rest of this story. The magnitudes are modest. The direction is the entire argument.

One more thing before the coda, because I can hear the objection forming: the serious literature does not say the optimal capital tax is zero, and you haven’t dealt with it. Correct, and next week I deal with all of it, at full strength — Chamley and Judd’s famous result and the corrections that bent it; the strongest wealth-tax model in the literature, which deserves far better than a footnote; the one argument for taxing capital that I cannot refute and won’t pretend to; what Sweden learned in 1984, at the price of an entire estate; and two Berkeley economists whose errors all point, mysteriously, in the same direction.

Next week: Scrooge McDuck Goes to Stockholm. No, he doesn’t get a Nobel.

If you found the terminal-accounting move useful, it’s the same one I ran on the wealth-management industry in The Conservation Law of Tax Alpha — different pile, same conserved quantity, same trick for finding where it went.

This piece is one argument out of a book’s worth I’ve already written up — a whole manuscript of economics done this way, from first principles, following the arithmetic wherever it goes. If you happen to be a literary agent and any of this is your kind of thing, get in touch. I have no appetite for firing proposals into the void; the void, as far as I can tell, is managing perfectly well without them.

My book, The Science of Free Will, asks an equally uncomfortable question about an equally cherished story. For the India series, start with The Paradox of India and the full collection; for Britain’s long institutional decline, first seen in 1979, there’s Albion.

Sometime this spring I did something I had been putting off for two decades, and I mean the title literally. Since 2005 I have kept three repositories of research ideas. Papers, links, half-formed signals, a note from 2 a.m. that says only “iceberg detector — trade size decay?” and an attachment I can no longer open. That was the year I walked away from a famous pod shop’s IP contract and started building execution systems that belonged to me, which is why the earliest items in the archive are about detecting hidden orders rather than about anything so grand as a theory of markets. I never intended any of it as a dataset. It was a place to put things so I would stop losing them.

It turns out that when you put things somewhere for twenty-one years, and each thing arrives with a date, you have built a dataset whether you meant to or not. There are 1,163 dated items in there. And when I finally coded them by year and theme and stared at the resulting heat map, I found that what I had accidentally recorded was not a list of good ideas. It was a record of my own attention moving, across several market regimes, under adversarial conditions, with real money downstream of every shift.

That record is one evidence base for a chapter I have just published in an IntechOpen volume on investor behavior and investment strategy. This post is about the one thing the archive says that I did not expect it to say.

What the archive is, and what it emphatically is not

Let me kill the obvious objection before it gets comfortable.

This is an N of one. It is a record of what a single practitioner chose to save, filtered through his interests, his blind spots, his employment history, and whatever he happened to be reading that week. It is not a random sample of anything. It is not evidence that any of these ideas made money. Selection bias is not a bug in this dataset; selection bias is this dataset.

Good. That is the point.

A published survey tells you what a field decided was worth writing down. An archive like this one tells you something different and, for a practitioner, more useful: what somebody with capital at risk kept coming back to, across market environments that had no obligation to be kind to him. An idea that shows up in 2007, again in 2013, and again in 2022 has survived three different worlds. An idea that shows up once and never again has told you something too.

So: not a performance ranking. A persistence ranking of research attention. A weaker claim than “these things work,” and a stronger one than “these things were fashionable.”

The arc nobody plans

The first surprise was the shape. The heat map itself is in the chapter, which is open access and free — I would rather you went and looked at it than took my word for the pattern. The caveats are with it: items can carry more than one label, cells are normalized within year, and 2026 is partial.

The early years — 2005 through roughly 2011 — are almost entirely plumbing. Iceberg detectors. Trade-size models for inferring informed flow. Intraday signals. Spread estimation from OHLC data. Pairs logic and scalper logic. Nothing in there asks why markets do anything. It is all measurement and execution, because that is what was missing.

The middle years, call it 2012 through 2018, are the factor years. Momentum. Value. Quality. Accruals. Piotroski. Factor timing. The factor zoo, and then the papers about how the zoo had gotten out of hand. This is the period when I still thought the question was what predicts returns.

And then from 2019 onward the archive changes character entirely. Composite alpha construction. Risk parity arguments. Volatility state. Regime conditioning. Downside correlation structure. Drawdown metrics.

Early years: build what was missing. Middle years: chase what predicts. Later years: conclude that the chase was the mistake.

I did not design that trajectory. I did not notice it happening. I saw it only when the years were laid out side by side, at which point it was so obvious that I felt slightly foolish for having needed a heat map.

The obvious objection is that this is occupational rather than intellectual — that the reading simply follows the job title. But that gets the causality backwards. Nobody assigned me to plumbing in 2005. I needed plumbing, because you cannot run a portfolio without it, and it did not exist. That is what a bottleneck is: the thing that has to be built before the next question can even be asked. Once execution worked, finding something to trade became the constraint. Once I had spent years looking, what became the constraint was the discovery that the looking had been aimed at the wrong thing.

So the archive was never tracking my job description. It was tracking the order in which the constraints bound. Which is precisely what you would want a record of practitioner attention to do, and considerably more interesting than a reading list.

I want to be careful about what the heat map can and cannot carry. It is a seismograph of attention, not a performance ledger. It cannot show that stacks beat signals. What it can show is that as responsibility widened, the recurring problem moved from finding an edge to keeping one alive. The stronger claim comes from the failures, and I will get to one.

Momentum, and the word “housed”

The most persistent theme is momentum. It appears in 17 of the 22 annual columns, in every form — time-series, cross-sectional, factor, cross-asset, option momentum, acceleration, reversal. If you want a one-word answer to “what recurs,” it is that word.

But here is the more interesting pattern, and the reason the archive is worth more than a citation count. The momentum papers appear throughout. The papers on momentum crashes, and on blending momentum with valuation, start piling up after about 2015 and never stop.

The lesson two decades of my own attention encodes is not “buy momentum.” It is: momentum works, but it must be housed correctly. And how it fails depends on which momentum you mean, which is why the archive’s habit of hoarding all of them turns out to matter. Cross-sectional equity momentum dies on the violent rebound after a stressed market, because it is short the beaten-down names at precisely the moment they scream off the bottom. Time-series trend following frequently does its best work during those same persistent crises, and dies instead on whipsaw — on the reversal that arrives the week after it commits. They fail on different days. What they share is that both get worse when liquidity thins, and worse again when crowding makes the rebalance expensive enough that everyone tries to leave through the same door on the same afternoon.

Which reframes everything else. Value and quality can obviously stand alone as return sources; what the archive records is that they increasingly show up as stabilizers, the thing that stops momentum from degenerating into pure extrapolation. Volatility, breadth, sentiment, and macro are not forecasts; they are state variables that tell you whether your other signals are operating in a friendly environment or a hostile one. Microstructure is not a technical afterthought; it is the boundary between paper alpha and tradable alpha.

The durable unit is not the signal. It is the stack: signal, sizing, execution, risk control, monitoring, and the governance that decides when to override the other five.

Let me be exact here, because there is a lazy version of this claim and I would rather not be caught holding it. Architecture is not a substitute for having something to trade. A beautifully governed portfolio with nothing in it is an expensive way to earn nothing, and I have met several.

But let me be equally exact about what I am not claiming, because the obvious reading is wrong. I am not saying I got better at forecasting and then decided governance mattered more. I stopped forecasting returns more than a decade ago. I have argued at length elsewhere that the hunt for a durable alpha source is largely a hunt for something that is not there, and nothing in twenty-one years of saved ideas has talked me out of it. The middle stretch of this archive is not a record of learning what predicts. It is a record of finding out that the question was wrong.

What survives that discovery is not a forecast. It is a rule for taking exposure conditional on a state you can actually measure, sized so that being wrong is survivable.

The sad part is that pouring money and brain hours into predicting the market is completely unnecessary to achieve success. Trading is about probability and risk management, not about predictions. You do not need to know what happens next in order to make money. You need to know how to size a position, under what conditions to hold one at all, and how to be wrong cheaply.

The conservation law nobody writes down

Here is where I want to plant a flag, because I think this is the gap that behavioral finance keeps walking up to and then declining to cross.

Behavioral finance is very good at the first step. Attention is scarce, so information gets incorporated in stages. Investors overweight private signals, disagree, and trade too much. They realize gains early and defer losses. They extrapolate. They herd, especially when career risk is involved. All of this is well documented and, as far as I can tell, true.

But every one of those mechanisms produces a pattern in prices. And a pattern in prices is not a return.

Between the pattern and your P&L sits a toll booth, and the toll is collected in several currencies. Spread and market impact. Borrow cost. Capacity — how much money the effect absorbs before it stops being an effect. Crowding, which is capacity’s angry cousin. Regime dependence, because the pattern is conditional and the conditions move. And path risk, which is the possibility that the sequence of returns bankrupts you before the mean arrives.

Now ask the conservation-law question. If a behavioral bias creates a measurable pattern, and the pattern does not show up as compounded wealth, where did the value go?

There are two answers and only two. Either the edge was never there — a false positive, a specification artifact, noise wearing a t-statistic — or the toll booth collected it. The anomaly literature is largely occupied with the first question. Practitioners find out about the second with money.

That is what a conservation law actually buys you. It does not tell you which channel the energy went down. It tells you the books balance, and that no amount of admiring the pattern will make the arithmetic come out differently.

There is also a seventh currency, and it rarely turns up in the anomaly papers. It has a literature of its own — implementation shortfall, operational risk — but that literature lives in a different building from the alpha research, and the two do not visit often.

In 2005 I was supplying signals to a market-neutral long-short book at a prop shop, trading the firm’s capital. A separate team, over which I had precisely no authority, was responsible for turning those signals into orders. The people involved shall remain nameless. On the day Hurricane Katrina made landfall, that team pushed a change to the trade placement algorithm. The change was simple enough, they judged, that it did not need to be tested.

It was not simple enough.

The algorithm began entering positions at roughly one hundred times their intended size. And unbalanced. Which is to say that a book whose entire risk architecture rested on being market neutral was, for the duration, not market neutral in any respect whatsoever. It had become a large, arbitrary, directional bet, placed by software, during Hurricane Katrina, while natural gas was coming apart in the Gulf.

It took about an hour to lose the book’s profits for the year.

Note carefully what did not happen. The signals were not wrong. The mechanism was real, the pattern was real, the edge was real, and it had been converting into money all year. Not one of the six line items above laid a finger on it. The alpha died in the gap between my research and the exchange, inside a system I did not control, because of a decision I was not consulted on.

And here is the part that bothers me more, twenty years on, than the money ever did.

I never ran a post-mortem. I was far too angry at the time to conduct one, and the precise sequence is now lost to whatever passes for posterity in this business. Every material loss I have taken since 2016 has produced a post-mortem, because the post-mortem is the machine that converts a loss into a system improvement. It is the only mechanism by which losing money is ever worth anything.

You cannot run a post-mortem on a system you do not control. You cannot diagnose it. You cannot fix it. You cannot stop it recurring. You can be told what happened, by the people who did it, and then go back to feeding signals into the same machine.

Capacity, crowding, and cost are the tolls the anomaly literature prices. The operational chain between your signal and the exchange is the one it doesn’t — and that toll takes your ability to learn along with your money.

Operational failure is the violent case, and it is mercifully rare. The ordinary case is slower: a real signal dies by degrees, which is also why single-signal investing usually fails. Efficacy varies by regime. Expected return compresses as capital arrives, and unwind risk rises with it. Turnover and slippage grow faster than linearly exactly when liquidity thins, which is exactly when the signal is shouting at you to trade. And relationships drift, because the microstructure of 2005 is not the microstructure of 2025, and pretending otherwise is a decision rather than an oversight.

One special case deserves its own paragraph, because it has eaten more good funds than the rest combined: market neutrality is incomplete neutrality. Dollar neutral is not liquidity neutral. Beta neutral is not crowding neutral, funding neutral, or crisis neutral. Quantitative market-neutral books tend to hold similar positions because they use similar data, similar models, and similar optimizers; when one is forced to deleverage, the selling moves prices against everyone else’s identical positions, which forces more deleveraging. August 2007 and March 2020 are the canonical exhibits. In both, dollar-neutral strategies took correlated losses at precisely the moment the hedging architecture was supposed to be doing its job.

The hedge was against the wrong thing. It was hedged against the market. It was not hedged against the other people running the same hedge.

It is not a controlled experiment — nothing in this business ever is — but consider what happened inside a single firm in 2020. Renaissance’s Medallion fund, open only to current and former partners, returned 76% net. Renaissance’s institutional equities fund, RIEF, lost about 19% on the year. RIEF is not a market-neutral book, incidentally — it runs long-biased with a low target beta — and that is rather the point, because what separated the two funds was not the hedge. Same firm. Same software. Same senior management team. The difference, per an investor who was in both, came down to holding period and risk architecture: RIEF held positions for six months to a year and hedged with factor-based risk models, while Medallion turned over fast enough to adapt as the regime changed. Medallion also ran more leverage, and the same investor was clear that almost nothing else about the two funds was correlated. So: not a clean instrument. Nearly a hundred points of spread inside one firm is still a difficult thing to pin on the signals.

And here is the line from that investor that I would put on a wall: there was nothing wrong with RIEF’s models.

I believe him. But in trading, the world is the test set. The return model was never the whole system.

Risk management is not a tax

The standard framing treats risk management as a drag on return — the thing you do to sleep at night, paid for in basis points.

In a live book the opposite is true. Risk management preserves the horizon over which alpha is permitted to operate. A valid signal with a bad drawdown profile does not underperform. It gets liquidated, or its manager capitulates, and then it produces nothing at all, forever. There is no partial credit.

Which is not a license for threshold rules. Mechanical drawdown cutoffs mostly fail, and the standard metric used to evaluate them is arithmetically broken in a way that ranks catastrophes as mild; I went through all of that in January and won’t repeat it. The line worth carrying forward is that a generic risk rule is not risk management. A rule that knows nothing about your entries, your instruments, or your holding periods cannot possibly know when to override them.

What a risk framework can rest on is a single empirical asymmetry that I think remains underappreciated even by people who know it. Short-horizon volatility is dramatically more predictable than short-horizon returns.

An illustrative check, SPY from February 1993 through April 2026. Regress this month’s realized volatility on last month’s: R-squared of roughly 0.41. Now run the analogous regression for returns, this month’s on last month’s: R-squared of roughly 0.00003.

Four orders of magnitude. That is not a difference of degree. It is a difference of kind. The exact coefficients depend on sample, universe, frequency, and estimator, and I would not defend the third decimal place of either number. I will defend the gap between them until the sun burns out.

This is the volatility clustering Engle formalized and took a share of the 2003 Nobel for, and its design implication is direct: model risk, not return. But note the limits. GARCH and its descendants fail most reliably at regime transitions — at the volatility spikes, at the collapses that start recoveries — which is to say, exactly when you need them. So the right operation is classification, not point forecasting. Do not stake the book on a number the model produces most confidently right before it is most wrong.

My own 2020 is a process story rather than a forecasting story, and I want to be precise about that, because I had no idea what was coming. Nobody did. What I had by then — having started trading my own capital in 2016 and built every piece of the stack myself, for reasons the reader can probably now guess — was a risk classification that did not require me to know. It was built to identify which state we were in, not to predict where we were going. Exposure came down partway into the drawdown and went back up a few weeks later, when the state changed back. No judgment call was involved at either end, which was the entire design intent.

Whether that was the right thing to do in any particular week is a separate question from whether the machinery did what it was built to do. The machinery did what it was built to do.

And it is the reason I own my plumbing now.

The protocol is a list of things that have already gone wrong

The chapter proposes a research-to-portfolio protocol. I want to state it in the only form I find useful, which is to attach each step to the specific disaster it exists to prevent.

1. Mechanism. Start from a behavioral or structural reason, not a mined predictor. Otherwise blind search gets mistaken for explanation. The exemption here is narrow, and you do not qualify for it: unless you have ninety PhDs on the payroll, something like a Markov chain running underneath the whole apparatus (if you have to ask, don’t even try), and a fund so uninterested in your money that it stopped accepting outside capital in 2005, you do not get to skip the mechanism. Renaissance can brute-force its way to an answer nobody on the premises can explain in words. You cannot.

2. Proxy. Translate the mechanism into transparent variables with documented measurement error. Otherwise vague psychology becomes an unverifiable signal.

3. Chronology. Test out of sample, strictly in date order. Otherwise the future leaks into the past.

4. Costs. Stress turnover, market impact, and execution delay. Otherwise paper alpha survives purely on the strength of costs you declined to model.

5. Integration. Combine signals with shrinkage and regime-conditioned weights. Otherwise you have bet the firm on one noisy input.

6. Governance. Set the leverage limits and the drawdown response ex ante, because otherwise the discretionary override gets invented at the moment of maximum stress by the person least qualified to invent it — you, at 3 a.m. Set the kill switches separately and mechanically, and note that these are not the drawdown thresholds I spent an earlier section arguing against: a kill switch fires on operational failure, corrupted data, a breached exposure limit, a model invalidated rather than merely disappointing. Had one been watching position size against intent on a certain morning in August 2005, this essay would be shorter.

7. Audit. After deployment, track signal decay, crowding, and slippage against backtest. Otherwise slow deterioration gets filed as temporary noise until it isn’t.

Governance matters more for behavioral strategies than for any other kind, for a reason that is almost too neat: the biases you are exploiting in the market are running in your own head at the same time, on the same inputs.

Coda: the humility premium

Any chapter claiming to take academic finance seriously has to reckon with Chen, Lopez-Lira, and Zimmermann, who mined 29,000 accounting ratios by brute force and found that the resulting predictors survive out of sample about as well as a couple of hundred peer-reviewed ones from the top journals. I went through that result in detail here and will not relitigate it.

What matters for this argument is narrower. It looks superficially like a refutation of “mechanism first,” and it isn’t. Peer review improves measurement, interpretation, transparency, error detection. It is simply not a test of tradability, and was never built to be. And “mechanism first” never claimed that theory guarantees the result. The mechanism supplies the search direction and the interpretive frame. It does not supply the blessing. The data still has to speak, and it is under no obligation to say what you hoped.

What theory cannot do under any circumstances is get you around cost, capacity, regime dependence, and path risk. Those bind regardless of how elegant the story is. Some patterns do carry real content; that has never been the dispute. What they do not carry is a durable edge — which is exactly why the game is exposure and survival rather than collection.

And then there is the term no quantitative framework contains.

A strategy that does not fit your temperament will not be executed. Not “will be executed poorly” — will not be executed. A risk-averse person will not hold momentum through a crash. A patient value investor will not trade intraday reversals. They will deviate at precisely the wrong moment, and they will deviate for the most convincing reason available, which is that every instinct they have will be screaming that something is wrong. I have said this to aspiring quants often enough that it has become a refrain, and I have watched it happen to good people with valid strategies more times than I can count, and I have never once seen the strategy win the argument.

Which returns us to where the chapter starts. The same forces that create mispricing out there — loss aversion, overconfidence, scarce attention — are operating in here, on you, in real time, while you decide whether to override the model. A strategy design that ignores the internal dimension is incomplete no matter how good its Sharpe looks in the deck.

One last thing the heat map shows, which I have deliberately left alone until now. The newest and fastest-growing cluster in the archive is machine learning and language models applied to return prediction. The evidence there points in two directions at once: these models can extract real information from news, and they also over-extrapolate recent trends and exhibit miscalibration that looks uncomfortably human. Which stands to reason. They were trained on us. AI may not abolish behavioral finance; it may industrialize it. There is a post in that, and it is coming.

But whatever that cluster turns into, it arrives at the same toll booth as everything before it. Cost. Capacity. Decay. Crowding. Governance. A language model does not get to skip the queue because it is new.

Which is the archive’s actual verdict, and the reason that twenty-one years and 1,163 saved items produced a conclusion about architecture rather than a conclusion about signals: the unit of survival is not the signal. It is the governed architecture that carries it.

The chapter is open access and free: “From Investor Psychology to Strategy Design — Evidence, Mechanisms and Implementation,” IntechOpen, DOI 10.5772/intechopen.1016617. The drawdown work behind it is “The False Promise of Drawdown Rules: New Evidence and a Better Framework,” Journal of Portfolio Management, 2025, 52(1), 145–161, DOI 10.3905/jpm.2025.1.765.

Related in this series: The Stop-Loss That Stops Gains on why drawdown rules fail and what to use instead. The Simplicity Tax on what happens when the binding constraint turns out to be computational rather than preferential. The Emperor Has No Alphaon what peer review does and does not certify.

Disclosure: I am the managing member of VS Asset Management, LLC, which uses some of the methods discussed here. Nothing above is a recommendation or a solicitation. Past performance is not indicative of future results, and I would gently suggest you not need to be told that twice.

If you enjoyed the argument that the governing architecture matters more than the component it carries, you may enjoy an equally uncomfortable question about an equally cherished story: my book,The Science of Free Will.

Also here: The Paradox of India and the full India series, on why the world’s largest democracy keeps producing outcomes nobody chose. And Albion, on Britain’s institutional decline — which I first noticed in 1979 and which has, if anything, accelerated.

Ask anyone to explain why quantum computers are fast and you will get the same answer. A classical bit is 0 or 1, but a qubit can be both at once, so a quantum computer “tries all the answers in parallel.” Three hundred qubits, 2^300 basis-state amplitudes, more numbers than atoms in the universe, checkmate.

This explanation appears in TED talks, in investor decks, in the mouths of CEOs announcing quantum initiatives, and — I regret to report — in the writing of people with physics PhDs who should know better. It is wrong. Not simplified-but-basically-right wrong. Wrong wrong. And its wrongness is not a pedantic footnote. It is the door to everything actually interesting about quantum computing, including the one application that I think is genuinely, permanently fascinating — which, in a twist nobody in the press releases mentions, is a form of analog computing. The technology we spent seventy years triumphantly abandoning.

One more promise before we start swinging. The mathematics beneath nearly everything in this post — the light that steers these machines, the noise that kills them, the channels their information survives through — leads back to one man, my thesis advisor, whom you have almost certainly never heard of. Hold that thought. We will get to him.

Getting there takes some demolition first. Bring a hard hat.

The Toll Booth

Here is the problem with “tries all the answers at once”: at the end of the computation, you don’t get all the answers. You get one. Measurement returns a single classical outcome, drawn according to probabilities you may have done nothing to shape. If your algorithm genuinely just “tried everything in parallel,” measurement would hand you a random wrong answer with overwhelming probability, and your 2^300 amplitudes would be worth exactly one lottery ticket.

The entire art of quantum algorithm design — the actual content of Shor’s algorithm, the reason it took a genius to find it — is choreographing interference. Amplitudes, unlike probabilities, carry phase: they can be negative, they can be complex, and above all they can cancel. In the canonical algorithms, the computation is arranged so that paths leading to wrong answers interfere destructively and annihilate, while the structure you care about survives into the final probability distribution. When it works, the measurement at the end is no longer a lottery; the right answer has been made overwhelmingly likely before you look. The parallelism is real, but it is useless without the choreography, and the choreography only exists for problems with a very particular kind of structure. Hold that thought too; it is the hinge of this entire post.

So even in the completely orthodox account, measurement is not an afterthought. It is the toll booth every quantum computation must pass through, and the toll is steep: of your exponentially many amplitudes, you get to keep one classical outcome per run.

General Purpose, Pointlessly

Now the second demolition. You will often hear — I have said loose versions of it myself — that a quantum computer “isn’t a general-purpose computer.” Stated that baldly, it is false, and the way in which it is false is more damning than the falsehood.

A universal quantum computer can simulate ordinary classical computation — the Toffoli gate hands you reversible classical logic, and in complexity language, P sits comfortably inside BQP. Anything your laptop can compute, a quantum computer can compute. A sufficiently large one could, in principle, run Excel.

It would do so at spectacular overhead, at millikelvin temperatures, inside a chandelier of dilution-refrigerator plumbing that costs more than your house, with zero speedup of any kind. That is the real point. The quantum computer is not forbidden from general-purpose computation. It is merely pointless at it. It is not a better computer. It is a coprocessor — an accelerator card for a narrow class of problems — and the honest public case is dominated by three motifs:

One: problems with hidden periodic structure. This is Shor’s algorithm and its relatives — factoring, discrete logarithms, the hidden-subgroup family. Here the speedup is genuinely exponential (over the best known classical algorithms; nobody has proven factoring is classically hard, a caveat the press releases also omit). This is the motif that gets essentially all the funding attention, so let’s look at the economics. A prominent 2025 surface-code estimate put the resources for factoring a 2048-bit RSA key below one million noisy physical qubits running for under a week; newer proposals using different error-correcting codes claim lower counts still. The numbers move quickly. The invariant does not: cryptographically relevant factoring demands sustained fault-tolerant computation far beyond anything yet demonstrated. And — here is the punchline — the world is not waiting. NIST has standardized post-quantum algorithms and the migration is underway right now, precisely because everyone can read the same papers. It will be uneven and slow — legacy systems have the survival instincts of cockroaches — and “store now, decrypt later” is a genuine problem not just for intelligence agencies but for anyone holding data whose confidentiality must outlive the decade: medical records, legal files, industrial secrets. But the headline stands: Shor’s algorithm is a billion-dollar key being cut for a lock the world is in the process of replacing.

Two: unstructured search. Grover’s algorithm searches N possibilities in roughly √N steps. Quadratic, not exponential. And a quadratic advantage is exactly the kind that gets eaten alive by constant factors: once you include fault-tolerance overhead and the cost of actually implementing the thing you’re searching, careful analyses find the quadratic advantage can disappear entirely on plausible early machines. The machine wins the asymptotic race and loses the actual one.

Three: simulating quantum systems. The subject of the rest of this post.

(Specialists will note that other algorithmic families exist — quantum walks, amplitude estimation, linear-system solvers, sampling schemes. Noted. None of them supplies a general-purpose accelerator, which is the claim on trial.)

Before moving on, notice the shape of what we’ve just walked through, because there is a conservation law hiding in it. My readers know the reflex by now: whenever someone offers you a free lunch, ask where the mass goes. The quantum speedup is not conjured from the vacuum. It is relocated. You get it exactly when the problem’s structure already matches the structure of quantum interference — when, in a sense, nature has pre-computed the hard part by being the kind of thing it is, and the algorithm merely arranges for you to read the answer off. Periodic structure matches the Fourier transform that quantum mechanics performs natively; so factoring accelerates. Generic problems have no such match; so generic problems don’t. The “power” of quantum computing was never a general-purpose power. It was always a resonance phenomenon. Which raises the obvious question: what problem resonates perfectly?

The Founding Memo

In May 1981, at a conference at MIT, Richard Feynman gave a keynote asking whether a classical computer could efficiently simulate physics. His answer was no — and everyone in the room already half-knew why. Write down a quantum system of n spins — n two-level systems — and the state you must track has 2^n amplitudes. Every additional spin doubles the bill. A dense state vector for 40 spins occupies roughly 16 terabytes; at 50, roughly 16 petabytes — and that is before you do anything to it. Not because engineers are lazy, but because 2^n is 2^n. Nature, Feynman observed, “isn’t classical, dammit” — and if you want to simulate nature, you had better build your simulator out of quantum mechanics itself. The talk was published in 1982 as Simulating Physics with Computers. Feynman was not entirely alone — Yuri Manin and Paul Benioff had already approached quantum computation from adjacent directions — but that lecture became the founding manifesto of quantum simulation, which is precisely the part of the field this post is about.

Read it today and the striking thing is what it does not contain. No factoring. No cryptography. No search, no optimization, no finance, no “quantum machine learning.” The founding memo proposes one central application: use a quantum system to simulate quantum systems. Everything else — four decades of it — has been a detour from the original purpose. A profitable detour, measured in venture funding. But a detour.

The one thing a quantum computer is unambiguously good at is being a quantum system. That sentence sounds like a tautology. It is actually a business plan, and the only honest one in the industry.

The Revenge of Analog

Here is where it gets genuinely beautiful, and where the story bends back on itself in a way I find delicious.

Before digital computers conquered the world, computing was analog. Lord Kelvin’s tide predictor computed tides with brass gears and pulleys because the gears’ motion obeyed the same equations as the tides. The Norden bombsight solved ballistics with cams and gyroscopes. Op-amp machines integrated differential equations as voltages, because a capacitor charging is an integral. The analog principle was always the same: don’t calculate the physics — build the physics, in a medium you can control, and read off the answer.

Digital killed analog for two reasons, and it is worth being precise about them. First, generality: a digital computer is one machine that runs every program, while an analog machine is married to its equations. Second — and this is the deeper one — error correction. Digital signals are discrete, so noise below threshold can be wiped clean at every step. Analog errors accumulate, drift, compound. Digital computing is the technology of keeping errors from breeding. It won so completely that “computer” now simply means “digital computer,” and analog machines sit in museums next to the slide rules.

Now watch what the most interesting corner of quantum computing is actually doing. (I mean analog here in the standard quantum-simulation sense: the computation proceeds through direct Hamiltonian evolution, not through a long sequence of fault-tolerant logical gates.)

Take an array of qubits — real physical systems: atoms, ions, superconducting circuits. Do not compile a program. Do not run gates. Instead, tune the couplings between them — the interaction strengths, the fields, the geometry — until the Hamiltonian of your array matches, term by term, the Hamiltonian of a system you care about. Then let it evolve. The array is not computing a description of the target system. Within its couplings, it is the target system — or, said with the lawyers present, it instantiates the same effective Hamiltonian, the same governing pattern of interactions, in a medium you can actually see and steer. Nature runs the program, at nature’s clock speed, in nature’s native instruction set. You have built Kelvin’s tide predictor out of atoms — with the one upgrade that matters: the gears are quantum mechanical. In the generic, highly entangled regimes where every known exact classical description pays the exponential state-space bill, the quantum device pays nothing extra merely to remain quantum. It was going to be quantum anyway.

This is not a thought experiment. It is a working research program with results:

At Harvard, arrays of hundreds of rubidium atoms held in optical tweezers, with couplings set by Rydberg interactions, have been programmed — in the analog sense, by arranging 219 atoms on the links of a kagome lattice — into a state showing the signatures of a quantum spin liquid, an exotic topological phase of matter that Philip Anderson conjectured in 1973 and that decades of classical numerics could characterize only partially. Cold atoms in optical lattices have realized the Fermi-Hubbard model — the model many condensed-matter physicists suspect holds the secret of cuprate superconductivity — as a literal physical object: an artificial crystal whose “electrons” are atoms you can photograph one by one, whose hopping and interaction parameters you can turn with a dial. Trapped-ion crystals have simulated quantum magnets with more than fifty spins, watching dynamical phase transitions no classical computer could track exactly.

Notice how these machines are being used. Not as universal computers running fault-tolerant gate sequences — even where the underlying platform could in principle support that — but as special-purpose simulators, each one married to a family of Hamiltonians, exactly as the tide predictor was married to tides. They have given up generality — the first thing digital computing won on — to get fidelity to a specific piece of physics. In this mode of operation they are analog computers in the full, honest, brass-gears sense of the term. The frontier of computing in 2026 is a technology from 1873 with the classical gears swapped out for entangled atoms, and almost nobody in the industry will say this sentence out loud, because “we have reinvented the analog computer” raises less capital than “we will break all encryption.”

I will say it out loud. It is the best thing in the entire field.

The Man Under the Floorboards

And now I get to do something I have wanted to do in these pages for a long time, which is talk about my thesis advisor.

Nearly every load-bearing formalism in this post — the description of the light that steers these machines, the noise that kills them, the channels their information survives through, even some of the tricks for keeping qubits alive — runs on mathematics written by a man most readers have never heard of: E. C. G. Sudarshan, of the University of Texas at Austin, where I did my PhD under him. George, to those who knew him. He died in 2018, and it is as if the field agreed to use his work forever while never mentioning his name. Consider the ledger:

In 1961, with Mathews and Rau, Sudarshan introduced the theory of dynamical maps — the mathematics of how quantum states transform when you don’t track everything they’re entangled with. Today these are called quantum channels, they are foundational objects of quantum information theory, and modern analyses of noise and information flow in quantum processors are routinely conducted in this language.

In 1963, he proved the optical equivalence theorem and introduced the diagonal coherent-state representation of light — the representation in which laser light, squeezed light, and the rest of the quantum states of the electromagnetic field are analyzed to this day. The 2005 Nobel Prize for the quantum theory of optical coherence went to Roy Glauber. The representation is called, when people are being careful, the Glauber-Sudarshan P-representation. People are not usually careful.

In 1976, with Gorini and Kossakowski, he derived the general equation of motion for open quantum systems — the master equation governing decoherence, which is to say: the equation that describes why qubits die. Lindblad published the infinite-dimensional version the same year, and the result is almost universally shortened to “the Lindblad equation”; the careful minority writes GKSL. Every quantum computing company on Earth is fighting, daily, at enormous expense, against a differential equation with Sudarshan’s name correctly on it and colloquially sanded off it.

In 1977, with Misra, he predicted the quantum Zeno effect — a watched quantum pot never boils — since demonstrated experimentally and turned into a working technique for suppressing errors in quantum processors.

Add the part physicists half-know: in 1957, Sudarshan — then a graduate student — and Marshak worked out the V-A structure of the weak interaction. The preprints appeared the same day, September 16, 1957: Feynman and Gell-Mann’s went to Physical Review and ran on New Year’s Day 1958; Sudarshan and Marshak’s was read at a conference in Padua and buried alive in the proceedings. And here is the part that makes the story genuinely strange rather than merely unjust. Feynman — a man constitutionally incapable of claiming credit that wasn’t his — said it explicitly, on the record, in 1963: the theory was “invented by Marshak and Sudarshan, published by Feynman and Gell-Mann” and completed by Cabibbo. Read that again. The most famous physicist alive stood up and said the other guys got there first. The field’s response? Crickets. Not a correction, not a reappraisal, not a prize. The textbooks kept the attribution they had, and the man Feynman pointed at went on being a footnote in the story of a force of nature he got to first — by Feynman’s account, not mine. (And we haven’t even discussed tachyons: the relativistically covariant framework for hypothetical faster-than-light particles is Bilaniuk, Deshpande, and Sudarshan, 1962 — worked out five years before the word “tachyon” existed. Response from the field: also crickets.)

One man. The weak force, the quantum theory of light, quantum channels, the equation of decoherence, the Zeno effect, and, for good measure, tachyons. The persistent rumor is that he holds the dubious distinction of having been nominated for the Nobel Prize more times than anyone who never won it. The nomination archives are sealed for fifty years, so I cannot prove the count. What I can tell you is firsthand. Steven Weinberg — my other advisor, whose office sat perhaps fifty feet down the hall from George’s, and who by universal agreement did not get along with him — confirmed the essence of it to me in person, in not so many words, when he told me that he backed George every single time. Fifty feet of hallway, decades of frost, and a vote cast for the man at the other end of it, every time. That is what it looks like when the work is undeniable and only the crediting mechanism is broken.

I am not telling you this only because he was my advisor and I loved the man, though he was and I did. I am telling you because Sudarshan’s ledger is this field in miniature: the durable substance is unglamorous, sits in the foundations, and the credit flows elsewhere. The press releases go to factoring demos and “supremacy” announcements. The load-bearing structure — the master equations, the channel formalism, the coherent states, the tuned Hamiltonians quietly impersonating unsolvable materials — is where the actual future is, and it is precisely the part nobody’s PR department mentions.

The Case Against My Own Case

Fair is fair; I’ve audited everyone else’s claims, so here is the hostile audit of mine.

Today’s analog simulators are not fault-tolerant. The thing that killed classical analog computing has not been repealed. Analog quantum simulators have no fault tolerance; their errors drift and accumulate exactly as op-amp errors did. Their defenders argue, plausibly, that for many-body physics you don’t need every amplitude exact — you need phases, transitions, qualitative structure, which are robust to modest noise. Plausible. Not proven. And here is the real epistemic problem: in the classically hard regime — the only regime that justifies building the machine — no general, efficient classical procedure exists to certify every detail of the answer. There are partial courts of appeal: limiting cases that classical methods can check, conserved quantities, agreement between independent platforms. None is universal. It is entirely possible that noise quietly corrupts the fine structure of exactly the phase diagrams we most want, precisely where no referee can look. That should bother you. It bothers me.

The classical frontier keeps moving. Many headline claims of quantum advantage have acquired a shorter shelf life than their publicity suggested. The canonical recent example: in June 2023, IBM reported a 127-qubit experiment in Nature as evidence that noisy quantum processors might already be useful in regimes where leading classical methods struggled — a circumspect claim, whatever the coverage implied. Even so, classical physicists using tensor-network methods rapidly reproduced or surpassed the results — some of them on a laptop. This keeps happening because classical simulation is not a fixed target: tensor networks exploit the fact that low-entanglement states compress beautifully, and clever people keep extending how much entanglement “low” can mean. Nor is the analog side exempt: even the Harvard spin-liquid result drew classical follow-ups — tensor-network calculations accurately tracked the state-preparation dynamics, and theorists continue to debate how much genuine long-range topological order the experiment established versus local patches of it. The result stands as a landmark. It does not stand as untouchable. The honest form of the quantum claim is asymptotic: classical cost grows exponentially in entanglement, so quantum simulators must win eventually, somewhere. Every finite, dated, specific claim of victory should be written in pencil.

In one representative episode, a team at the Technical University of Denmark, using a printer-sized photonic processor from the startup Orca Computing, reported that a quantum-enhanced generative model designed better protein-binding peptides than its classical counterpart — with the biggest gains exactly where training data was thin. The Orca device is a boson sampler: single photons threaded through nested fiber loops, interfering in time rather than across a wall of beam splitters, with room-temperature detectors reading off which time-bins the photons exit — a sample from a distribution that is genuinely hard to compute classically. (No dilution-refrigerator chandelier, which is precisely why it fits on a benchtop.) Drop that hard-to-mimic distribution into a data-starved generative model and you get a richer spread of candidate peptides. Interesting, and worth watching. And to their considerable credit, the researchers said the quiet part aloud: the machines are still so small that a larger classical model could likely do better today. Which means the contest was run against a size-matched classical control, not against the classical state of the art — a promissory note, not an advantage. A sampler injecting exotic diversity into a small model is precisely the kind of edge a cleverer classical prior may erase within a year, and boson-sampling advantage claims have been spoofed classically before. Write it in pencil. And note what it is not: it is not quantum simulation of a physical system. It is the sampling motif — Orca’s own literature calls these devices co-processors — pressed into service accelerating an AI model. A detour from the detour.

And there is a wildcard under the whole table. Stephen Wolfram — cited by name, as usual; the man is right about too many things for the establishment’s comfort — has proposed that quantum mechanics is itself the emergent description of a deeper multiway computation, and in that picture a definite observation is not a free primitive. The observer must do computational work to knit branches into an outcome — in effect, to outcompute the system itself. When his group “compiled” quantum factoring into their models, the measurement stage demanded a pile of additional multiway events — an explicit cost of measurement — and how that cost scales is a question they have not answered. Precision here is friendlier to Wolfram than hype: this is not a hidden defect in Shor’s theorem, which includes measurement and stands firmly within standard quantum mechanics. It is a test that Wolfram’s deeper framework must pass — can it recover the standard speedup without smuggling an exponential bill into the observer? — and the test is unresolved. (It is also a different objection from Gil Kalai’s better-known noise skepticism: Kalai says the machine can’t be built; Wolfram asks whether, even built, the accounting closes.) But note who would be left standing if the answer ever came back negative. Not the grand interference machines, which lean their entire weight on the contested step — but the analog impersonators, which ask their measurements only for ordinary, robust physical quantities: correlations, densities, order parameters, sampled over many runs rather than decoded from one choreographed global answer. Make of that alignment what you will.

What would change my mind? A fault-tolerant digital quantum computer solving a classically-verified-intractable problem end to end, measurement included, with the full error-correction overhead on the books. I will cheerfully eat this post. I am not planning the menu.

What to Watch

So: ignore the factoring countdowns. Ignore “quantum advantage” announcements that get reproduced on a laptop by August. Watch, instead, for three things. Watch for an analog simulator mapping the phase diagram of the two-dimensional Fermi-Hubbard model in the regime classical numerics cannot reach — because somewhere in that phase diagram, many condensed-matter physicists suspect, lies the mechanism of cuprate high-temperature superconductivity, and a civilization that finally understands the mechanism gets to go shopping for better ones. Watch for spin-liquid and topological-matter results that hold up when the classical spoilers come hunting. And watch for whether the field’s money and language quietly rotate from “quantum computer” to “programmable quantum matter” — because that rotation, when it comes, will be the industry admitting what Feynman told everyone in 1981, what the brass gears knew in 1873, and what this post has been arguing all along.

The quantum computer was never a better computer. It is a piece of nature you can tune — the universe’s own instruction set, exposed through a control panel. That is not a lesser thing than a computer. It is arguably a greater one. It just isn’t the thing on the pitch deck.

If you enjoy watching cherished stories audited down to the studs, my book, The Science of Free Will, asks an equally uncomfortable question about an equally cherished story.

More physics? This post’s cousins live in the Physics series. For markets, see Trading. New here? Start with The Paradox of India and the full India series — and for related country-specific work, there’s Albion, Britain’s institutional decline, first spotted in 1979.

There is an Indian journalist named Bibhuti Bhushan Nayak who holds a world record for being kicked in the groin.

I want to be precise about this, because the precision is the joke. In 1998, Nayak — a journalist and martial artist associated with the Times of India — invited four of his karate students to kick him, unprotected, in the groin. They managed forty-three kicks in ninety seconds. He was not content with this. He went on to set a record for the most concrete slabs broken over his groin with a sledgehammer (three), the most one-handed cartwheels in a minute (thirty-four), and 1,448 sit-ups in an hour. Often he set three records in a single day, having given himself no time to practice.

Hold that man in your mind. We will come back to him, because he is not a curiosity. He is the entire argument.

The thing about a billion people

I was watching the World Cup recently — Egypt, Cape Verde, Curaçao, Morocco. Curaçao has a population of about 150,000 people. That is not a typo. It is roughly the population of a mid-sized Indian suburb that nobody outside India has heard of. And Curaçao was at the World Cup — they lost 7–1 to Germany, a result they earned the hard way, against a Germany that was actually trying. India would never make Germany try. The gap would be so total that the Germans would stop of their own accord somewhere around the eighth goal, not out of mercy but out of sheer pointlessness, the way you stop pressing a button that no longer does anything — unless, of course, they decided to keep going purely to see how high the number could climb. (Number go up.) India, 1.4 billion people, one in every six humans alive, has never played in a World Cup at all. Not once. (India technically qualified for the 1950 tournament, by the elegant mechanism of every other Asian team withdrawing, and then withdrew. This is the most Indian sentence I will write today.)

The men’s national football team hovers somewhere around 140th in the FIFA rankings, in the company of nations you would struggle to find on a map. And before you reach for the comforting explanation — cricket-mad country, football just isn’t the thing — look at the Olympics, which span everything humans have invented to do with their bodies. At the 2024 Paris Games, India won six medals. Zero gold. Six medals for 1.4 billion people works out to roughly one medal per 233 million human beings, which places India, reliably, Games after Games, among the very worst large nations on Earth by any per-capita measure. Grenada did better. Dominica — population seventy thousand — did spectacularly better, winning the first Olympic medal in its history, a gold. Saint Lucia did better. Pretty much every speck of a country that medaled at all did better, per person, than the nation of 1.4 billion.

Now hold that in your mind, next to the man being kicked in the groin, and notice that something does not add up.

Because here is the same country. India produced Viswanathan Anand, who held the world chess championship for years, and then a whole generation of grandmasters behind him — Gukesh, who in 2024 became the youngest undisputed world chess champion in history, Praggnanandhaa, Erigaisi, an entire conveyor belt of teenagers dismantling the world’s best. India was the dominant force in field hockey for an era, with eight Olympic golds. India is, in cricket, simply the center of gravity of the entire sport — its best players, its biggest league, its money, its future. And in the long, strange tail of human achievement — spelling bees, memory competitions, the Guinness book itself — India shows up again and again and again.

So which is it? Is India terrible at sport, or world-class at it?

The answer is: both, and the gap between the two is the most interesting thing about the country. India is either historically, catastrophically, almost-comically bad at something, or it is at or near the best on the planet. What India is almost never is mediocre. It does not do the vast, boring middle of the distribution. It lives in the tails.

Why the middle goes missing

Here is the strongest version of the objection, and it deserves a fair hearing. India is poor. For most of its modern history it was very poor, and decades of childhood undernutrition do not build Olympic sprinters. There are no playing fields. There is no money. A talented kid in Lagos or São Paulo has more grass to run on than a talented kid in most of urban India. Poverty is real, and it is not nothing.

But poverty cannot be the whole story, for the simple reason that poverty is a property of the whole country — and the whole country is the thing that simultaneously tops the chess world and fields one of the worst football teams on Earth. If deprivation were the explanation, it would suppress everything uniformly. It would not pick chess to win and football to lose. Something is selecting. And once you ask what, the answer falls out, and it is not about money. It is about what kind of machinery a given pursuit requires.

Consider what it takes to be merely good — not great, just solidly, professionally competent — at football. You need youth academies. You need coaches who themselves were coached. You need leagues with a hundred clubs and a thousand journeymen, a scouting system, age-group competitions, physios, pitches, a whole institutional pyramid whose entire function is to manufacture the fifty-thousandth-best player in the country and make him good. Excellence in football sits at the top of a very tall pyramid of competence. No pyramid, no peak. You cannot skip the middle, because the middle is the thing that produces the top.

Now consider what it takes to be the best chess player in the world. You need a board. You need an opponent, or these days a laptop. And you need one obsessive human being, drawn from a very large population, who will not stop.

That is the whole difference. The tails are demographic. The middle is institutional. Excellence in a board-and-a-fanatic pursuit can emerge from raw population times raw obsession: take one human in ten million who is wired for it and refuses to quit, and out of 1.4 billion people you have produced a hundred and forty potential world-beaters with no system required. But competence — the dull, dependable, fiftieth-percentile-of-a-deep-field competence that you can only get by building it — requires exactly the thick institutional middle that India, for reasons this entire series has been about, never built.

India has the demographics. India lacks the institutions. So India gets the tails and misses the middle. Football needs a system, and India is bad at systems. Chess needs a genius, and India has, by the pitiless arithmetic of scale, an enormous number of geniuses. The empty World Cup qualifying table and the world chess crown are not two facts. They are one fact, seen from two angles.

Let me give you the missing middle as I watched it happen with my own eyes, when I was nine years old. Delhi, 1975, the DCM Trophy — one of the big domestic football tournaments of the day. South Korea, for reasons I have never fully understood, sent a university team: Hanyang University. Students. They were playing against the best professional clubs India had. And in their opening match — I do not remember the opponent — these Korean undergraduates were such obvious gentlemen that they visibly took it easy, coasted, and won by a single goal, plainly operating at about five percent of their actual capacity. One of the Delhi papers ran a solemn column the next day about how the South Koreans had disappointed in their opener. As Dave Barry would say, I am not making this up.

Here is the part I think about. I was nine, and even I could see what the journalist could not: the Koreans were not disappointing. They were being kind. They were holding back to spare their hosts’ feelings — the way you let a small child win at cards — and a grown professional sportswriter looked at that mercy and mistook it for the limit of their ability. Hanyang, having been thus chided, proceeded to wake up and dismantle every subsequent opponent, and in the final they beat East Bengal — then as now one of the two greatest clubs in the country — 2–0. A college side. Touring undergraduates, knocking over India’s finest professional team to win the trophy, while the local press worried they weren’t quite living up to expectations.

And here is the detail that elevates the whole thing from anecdote to indictment, the part I did not appreciate until I went back and looked at the records decades later: 1975 was not a fluke. It was the beginning of a habit. Hanyang came back the very next year and shared the title, having drawn its finals against the Border Security Force. Then Myongji University won it in 1981. Then Incheon University won the 1982–83 tournament. Then Kyung Hee University won it in 1990. Then Incheon University won again in 1992–93, beating East Bengal in the final for the second time in the tournament’s history. Four different South Korean universities, six titles between them, across seventeen years — visiting students treating one of India’s flagship football tournaments as a kind of working holiday, with India’s best professional clubs lining up obligingly to be the runners-up. That is the missing middle, rendered not as a single embarrassing afternoon but as a two-decade structural fact that nobody in Indian football apparently found alarming enough to do anything about: India’s best could not reliably beat other nations’ undergraduates, and the gap was sometimes so wide that the undergraduates had to pretend, out of politeness, that it was closer than it was.

The billiards room at the Gymkhana

Let me tell you about Michael Ferreira.

Ferreira — “the Bombay Tiger” — is a lawyer and a three-time World Amateur Billiards Champion. In 1978 he became the first amateur in history to break the thousand-point barrier, scoring 1,149. He was, for a stretch of the 1970s and 80s, plausibly the best player of English billiards alive. He was also a man who understood the argument of this post in his bones, and was furious about it: he spent his playing career campaigning for India to give some facilities, any facilities, to sports other than cricket and hockey. He felt it so keenly that in 1981 he refused the Padma Shri — a national honor — in protest, on the grounds that the cricketer Sunil Gavaskar had been given the more prestigious Padma Bhushan and that his own world titles were worth no less. (The Padma Bhushan did eventually follow, after his third world title; the official award rolls list it in 1984. He had made his point.) Here was a world champion, produced by a country that had given him essentially nothing, who spent his career angry about the missing middle. He is the thesis with a civilian award.

Ferreira was a friend and colleague of my mother’s. And one day my father and I took him to the billiards room at the Gymkhana Club in Delhi.

You need to understand the Gymkhana. It is one of those colonial-era institutions where the waiting list for membership is measured in decades and the disapproval is measured in micrometers. My father was a member. Michael Ferreira — three-time champion of the world — was not, and that is the whole point. Ferreira is a Bombay man, a Bombay lawyer; in Bombay he had his clubs and his tables. But he was in Delhi, away from home, and the best billiards player alive wanted somewhere to practice — and in Delhi the way that happened was for a member of a club like the Gymkhana to sign him in as a guest and lend him a table. Read that sentence again. A reigning world champion, visiting the national capital, could not simply walk in somewhere and find a table to keep his eye in; he had to be somebody’s guest. This is the missing middle you can stand next to. It is a man with three world titles standing at the door of a Delhi club he did not belong to, waiting to be signed in, so that he could go inside and be the best on Earth at the thing the club happened to own a table for.

So my father signed him in, and got, as his reward, the privilege of being Ferreira’s practice partner. I put “partner” in quotes too, and “play,” because playing billiards with Michael Ferreira consisted of the following procedure. You broke. Then you sat down. Then you watched, in respectful silence, as he ran the entire table — every ball, methodically, without apparent effort, the way you or I might tidy a desk. Then, when he was finished, you got up, you broke again, and you sat back down. That was the game. You were not an opponent. You were the man who occasionally reset the balls so that the demonstration could continue.

Now here is what happened. Other members of the club, seeing a world champion at the table, began to drift over. And then they began to grumble. They wanted a turn. They felt that my father was hogging Ferreira — monopolizing him, refusing to let the rest of the room have a crack — as though a man’s own invited guest were a shared club facility, like the sauna or the good snooker cue, that it was somehow selfish of him to keep to himself.

My father started laughing, and could not stop.

Because think about what those men were asking for. They were lining up to “play” a three-time world champion as though he were a club amenity — the good cue, the corner table, a thing you sign up for in fifteen-minute slots. And the contest they were demanding a share of did not exist. There was no game to be had. The gap between Michael Ferreira and a competent, enthusiastic Gymkhana club member was not a gap you could play across. It was the gap between a world-class tail and a keen amateur — total, unbridgeable, and faintly absurd the moment you saw it laid out on green baize. You broke, and you sat down. That was all anyone in that room was ever going to do, member or not.

That billiards room is India. One man at the very top of the world, drawn from a vast population by sheer individual obsession. A queue of pleasant, capable amateurs who imagine the distance is smaller than it is. And no structure of any kind connecting the two — no rungs on the ladder, no middle to climb through, just the champion clearing the table and everyone else waiting for a turn that was never going to be a contest.

The tails of every distribution

Once you see it in sport, you see it everywhere, because it is not really a fact about sport. It is a fact about how India distributes excellence and competence across everything.

The man with the longest fingernails ever recorded on a single hand is Shridhar Chillal of Pune — an aggregate of 909.6 centimeters, about the length of a London bus, grown on his left hand over sixty-six years. He started in 1952, after a schoolteacher beat him for accidentally snapping the teacher’s own long nail and told him he would never understand the care it took not to break one. So he spent the next two-thirds of a century proving the teacher right. The man with the longest mustache is Ram Singh Chauhan of Rajasthan — roughly fourteen feet, untrimmed since 1970, oiled daily with coconut and mustard oil and once featured in a James Bond film. In Varanasi there are ascetics, the Khareshwari, who have taken a vow never to sit or lie down again, not even to sleep, for years on end. And there is Amar Bharati, a former Delhi bank clerk who, by every account, raised his right arm above his head in devotion to Shiva in 1973 and has not lowered it since — the muscles long withered, the fingers fused and feelingless, the arm now a thing he could not bring down if he wanted to. Half a century, pointing at the sky.

None of these is a record in the institutional sense — there is no federation of arm-raising, no national mustache development squad, no junior-varsity circuit feeding talent up to the Standing Babas. That is exactly the point. These are not the products of systems. They are the products of one human being, drawn from an ocean of human beings, who decided to do an extreme thing and would not stop. Population times obsession, with the institutions subtracted entirely. They are Gukesh and the empty football table wearing a different costume.

But they built the largest city on Earth

At this point a reasonable reader objects: if India is so hopeless at organization, explain the Kumbh Mela.

It is a fair challenge, and it nearly sinks everything, so let me take it seriously. Every few years, on the banks of the Ganges, the same Indian state that cannot reliably deliver tap water conjures the largest temporary city on the face of the Earth. Tens of millions of pilgrims — at the largest gatherings, well over a hundred million people across the full span — arrive at a stretch of riverbank that was empty mud a few weeks earlier and is now laid out with roads, pontoon bridges, electricity, water points, sanitation, hospitals, police, and a grid of camps stretching to the horizon. Then, when it is over, the whole thing is struck and the river takes the ground back. A metropolis the size of a major world capital, summoned and dissolved on a schedule, decade after decade. (Not flawlessly — there have been terrible crowd crushes, including recent ones, and I will not pretend otherwise.) But as a feat of mass logistics it is genuinely staggering, and it is unmistakably the work of the same government that elsewhere cannot keep a municipal drain clear.

So which is it — can India organize, or can’t it?

And the answer, again, is that the question is wrong, because it lumps together two completely different things. The Kumbh Mela is a project. It is a single, episodic, overwhelming mobilization toward one legible goal, with a hard deadline and the full weight of the civilization pointing at it. India is magnificent at projects. It put a spacecraft in orbit around Mars on a shoestring, on the first attempt. It runs the largest election in human history — nearly 970 million eligible voters, 642 million ballots actually cast in 2024 — every few years. It builds a city for a hundred million people and then takes it down. When India decides, all at once, to do one enormous thing, it is as good as anyone alive.

What India historically could not do is the other kind of organization — not the surge but the sustain. Not the project but the institution: the boring, permanent, unglamorous machinery that does the same modest task competently every single day for forty years with nobody watching and nobody’s salvation on the line. The Kumbh is the arm raised for fifty days. The functioning water utility is the pyramid maintained for fifty years. India has always been able to summon the first. It is the second that goes missing — and the second is precisely what produces the fiftieth-percentile midfielder, the mid-sized firm, the city that works on a Tuesday in March when nothing in particular is happening.

You can see the whole distinction inside a single Indian household. Indian homes are, very often, spotless — swept, scrubbed, polished to a gleam, maintained with fierce daily attention. Step outside the front door and the city is filthy as all hell. The same people, the same hands, the same standards of cleanliness — immaculate within the boundary of what an individual or a family can personally maintain, and a catastrophe the instant it requires a collective institution to do the maintaining. The house is the surge you own. The street is the sustain you have to build together, and the machinery for building it together is the thing that was never there.

And it is worth saying plainly why it was never there, because this is not some eternal Indian incapacity written into the soul. A great deal of that missing machinery was deliberately destroyed. When the British arrived, India was not a backwater waiting to be modernized — it was, in several of the most demanding industries on Earth, the world leader. In 1750 India produced something like a quarter of the entire planet’s manufacturing output; by 1900, after a century and a half of the Raj, that share had collapsed to about two percent. That is not a statistic about poverty. That is a statistic about demolition.

Take the two industries India most conspicuously dominated. The first was textiles. For centuries the handloom weavers of Bengal had made the most coveted fabrics in the world — muslins so fine they were described as woven air — exported west to Egypt and Persia and east to Java and Japan, the cloth of choice for European aristocrats who had nothing at home that could touch it. The British did not merely out-compete this; they also legislated against it — taxing Indian cloth, protecting Lancashire’s mills, and, once Bengal was theirs to administer, running the supply chain in reverse: extract the raw Indian cotton, mill it into fabric in England, and sell that fabric back to the people who used to make the best in the world. (Honest historians note the decline had other causes too — Mughal political fragmentation, drought, transport costs, and above all the brute productivity of the power loom — and I will not pretend Britain did all of it single-handed. But the policy of suppressing a rival by statute rather than beating it on quality alone is a matter of record.)

The second was shipbuilding, and here the detail is so good I can hardly believe it is true. Indian shipwrights built their hulls from Malabar teak, which is naturally resistant to the marine worms that bored through European oak; an Indian teak ship could stay at sea for fifty to eighty years, while a British oak ship was often scrapped after twelve. The Parsi Wadia family of Bombay built vessels said to be “vastly superior to anything built anywhere else in the world” — so superior that the Royal Navy itself commissioned warships from them. One of those Bombay-built ships was HMS Minden, and there is a persistent legend — disputed by the kind of historian who ruins parties, and most likely false, since the better-attested account has the man aboard an American truce ship — that it was on the Minden‘s deck that an American lawyer, watching a battle in 1814, wrote the poem that became the United States’ national anthem. The detail may be apocryphal; the dependency is not. The British Empire, at the height of its naval power, was having warships built for it by Indian craftsmen because it could not build better ones at home. And then, over the following century, that industry too was regulated and starved out of existence, until the masters of the Indian Ocean were spectators on their own coastline. As one historian put it, the decline was never due to a lack of skill. It was due to a lack of sovereignty.

This is the deepest layer of the missing middle. Chandni Chowk, the great commercial artery of Mughal Delhi, was once among the richest market streets on the planet, the beating heart of an institutional ecosystem — guilds, financiers, manufacturers, the whole connective tissue of a working economy — that the Raj hollowed out over generations until it became a monument to what had been demolished. India did not simply forget how to build the middle. In the industries where it was the best on Earth, the middle was taken apart on purpose.

And then — this being India, and this being the through-line of everything I have written in this series — the country took the wound the British had inflicted and, for its first forty-four years of independence, made it dramatically worse. From 1947 to 1991 India answered colonial deindustrialization with homegrown socialism so creatively destructive that it managed to suppress the middle the colonizers had missed. This is not hyperbole. In 1979 the Finance Minister, Charan Singh, stood up to deliver his budget and, in the name of “a ruthless curbing of luxury expenditure,” raised the excise duty on toothpaste from 10.5 percent to 25 percent — because in the socialist paradise, clean teeth were a bourgeois indulgence. (I am not paraphrasing; the man said it, and I quoted the speech in From Goddess Lakshmi to Ration Cards.) You do not rebuild a wounded economy by taxing toothpaste as a luxury and rationing the right to manufacture; you finish the demolition the empire started, only this time with your own hands and a five-year plan. The roots did not begin to grow back in 1947. They began to grow back in 1991, when India finally stopped strangling itself.

But here is the detail that proves the whole argument — that the missing middle was always a policy, never a people — and it comes, as the best evidence in my life usually does, from my own family. My grandfather was the accountant to the original Mr. Oberoi, the man who built what became some of the finest hotels on Earth. And those hotels were magnificent during socialism, in the depths of the License Raj, at the very moment the rest of the economy was queuing for a scooter it would receive a decade later. How? Because hotels brought in foreign exchange — and foreign exchange was the one thing the regime wanted badly enough to permit excellence in pursuit of it. (The same postdescribes the full horror of what that regime did to ordinary people who needed dollars — the pittance you were allowed to take abroad, the impossibility of paying for a college application.) So in a country that forbade competence almost everywhere, there was a narrow corridor where competence was allowed, and into that corridor India poured world-class hotels, instantly, as good as anyone’s anywhere. The Oberoi was the missing middle made visible — a glowing island of institutional excellence in an ocean of deliberate scarcity. And its existence proves the scarcity was a choice. Indians did not lack the capacity to build the dependable, world-beating middle. They were forbidden from building it, except where the state had a reason to look the other way.

I know this because I spent many happy hours in those hotels as a boy in the early 1970s. They had, among other marvels, air-conditioning — which in the India of that era was not a convenience but a frank luxury, a thing that marked a different and cooler universe. My own family could not afford an air-conditioner until I was much older, and when one finally arrived it was a gift from my mother’s brother — given, let the record show, to a decorated naval fighter pilot and a barrister trained at Lincoln’s Inn. Those were my parents. Between them, a man who had landed burning aircraft for his country and one of perhaps a few dozen Lincoln’s Inn barristers in all of India earned a combined sixteen dollars a month. Sixteen.

And now I can hear the objection forming, because I have heard it from every earnest defender of the era: ah, but that’s the dollar figure — purchasing-power parity, the rupee went much further at home. No. That objection is the very thing this essay is about, and it is wrong twice over. First, the sixteen dollars is the official exchange rate — and under FERA, the foreign-exchange regime, the official rate was a polite fiction miles from the black-market rate, for the simple reason that you were forbidden from actually obtaining dollars. The gap between the two rates was not a technicality. It was the strangulation, expressed as a number. And second, the whole premise of purchasing-power parity is that there is something to purchase — and there was, to put it precisely, almost nothing. You could not buy the car (ten-year waiting list), or the telephone (eight-year waiting list), or, past a certain excise bracket, the toothpaste. The rupee “went further” the way the wrong key goes further in a locked room — impressive in the hand, useless in the lock. That is what a socialist exchange rate and a rationed economy did to a household stuffed with exactly the kind of talent a functioning country builds its competent middle out of — it left them unable to buy a window unit, waiting on the generosity of a relative. The cool air of the Oberoi and the heat of our flat were the two Indias of the License Raj: the tiny permitted enclave where India was allowed to be the best in the world, and the vast rationed scarcity where everyone actually lived.

That is why this is a story of hope and not lament. The capacity was never missing. The permission was. And the astonishing thing — the thing you can walk into and feel for yourself — is how fast the middle has grown in the three decades since India finally granted itself that permission. Go and stay in an Indian luxury hotel now, the descendants of the ones my grandfather kept the books for. Not a tail event, not one freakish genius — a whole category, a deep bench of trained professionals maintaining a flawless standard every single day, which is to say the very competent middle this entire post has been mourning. The first time you experience it you may quietly conclude you do not especially want to stay anywhere else ever again. The difference between then and now is not that Indians got better. It is that India got out of their way.

The exception that proves the mechanism

Here is the test of whether any of this is real, rather than a clever story I have told myself — and this series has spent its whole run catching satisfying stories that did not survive scrutiny, so I am obliged to apply the knife to my own.

If the missing middle is genuinely institutional and not just a matter of national temperament or some mystical Indian incapacity for team effort, then the mechanism makes a falsifiable prediction: build the institution, and the middle should appear. Where India has actually invested — money, academies, coaching, a real pyramid — it should stop living only in the tails and start populating the middle that produces reliable, repeatable, plural success.

And that is precisely what you see. In shooting, where money, coaching, academies, and real international competition began to accumulate, India went from Abhinav Bindra’s lone, miraculous gold in 2008 to Manu Bhaker winning two medals at a single Olympic Games in 2024 — not one freak champion, but depth. In badminton, where an actual training ecosystem grew up around academies and serious institutional backing, India produced not just a single star but a cohort — Olympic medals, world championship medals, and in 2022 the Thomas Cup itself, won for the first time, which is a team trophy and therefore exactly the kind of thing the missing-middle thesis says India should fail at. Unless, of course, you build the middle. In which case it shows up, right on schedule.

This is not a defeat for the argument. It is the argument’s strongest confirmation. The mechanism predicts its own exceptions. India is not constitutionally incapable of the boring competent middle. It simply does not get the middle for free, the way it gets the tails for free, because the middle has to be built and the tails only have to be born. Give India the institution and the middle materializes. Withhold it, and you are left with a billion people, a handful of world-beaters thrown up by the law of large numbers, and an empty space where the journeymen should be.

So the next time you wonder how a country of 1.4 billion cannot find eleven men to qualify for a World Cup, remember that it is the same country, governed by the same arithmetic, that produced the youngest chess champion in history and a man who let his students kick him in the groin forty-three times in ninety seconds. The midfielder India does not have and the sledgehammer-groin journalist it does are the same phenomenon. One requires a system India never built. The other requires only a single, magnificent, irreducible human being — and of those, India has always had more than its share.

You break. You sit down. And somewhere, someone clears the table.

If you found this essay worth your time, you may enjoy my book The Science of Free Will, which asks an equally uncomfortable question about an equally cherished story.

If you want to know more about India, start with The Paradox of India, the first essay — or browse the whole series.

Related country-specific work: Albion — Britain’s institutional decline, by someone who first saw the place in 1979.

There are old traders, and there are bold traders. But there are no old bold traders. — trading-floor proverb (stolen from the pilots, as traders steal everything that works)

The Wall Street Journal reconstructed the week almost cinematically, and I could not improve on the staging if I tried. As the week began, Leopold Aschenbrenner, 24 years old, was preparing for his wedding: a multiday celebration in Carmel, ceremony at a Tuscan-style villa, send-off at a spa in the forest, and — because this is San Francisco money in the year 2026 — a pre-wedding colloquium, with breakout sessions, so the guests could discuss ideas. The couple’s only request: no gifts.

By the time the guests began to arrive, his $45 billion hedge fund was reportedly being carved up on the phone at midnight. Lenders were issuing margin calls. Rival desks were reportedly trading against his anticipated forced sales. And Ken Griffin’s Citadel was negotiating to buy the bulk of his public portfolio — roughly $16 billion of stock, per the Financial Times — at a discount of more than 10% to market. The deal was struck just before Thursday’s open, while the caterers were presumably confirming headcount.

The couple had asked for no gifts. The market brought one anyway.

The Nostradamus of AI

Let me compress two years for you.

In 2024, Aschenbrenner — a former OpenAI researcher barely out of college, who had departed the lab after a disputed dismissal — published a 165-page essay about the coming of superintelligence. It went viral. Michael Dell shared it. Ivanka Trump shared it. And the essay became the intellectual foundation of a hedge fund: he named the firm Situational Awareness, after the manifesto, and described it as a “brain trust on AI.” Within a year it managed $1.5 billion. The pitch was simple: nobody understood AI like he did, and therefore — therefore — nobody could invest in it like he did. “We’re going to have way more situational awareness than any of the people who manage money in New York,” he told a podcast. Note the logical structure of that sentence. We will return to it, with a crowbar.

And for a while, the returns were spectacular. Up 47% after fees in the first half of 2025. A stumble when DeepSeek hit in January of that year, quickly recovered; up roughly 200% for 2025. Then the real fireworks: up about 439% for 2026 through June, assets swelling past $20 billion and, per the Journal, toward $45 billion — a size that took Ackman and Loeb decades to reach, achieved in under two. Jane Street, a firm that almost never allocates to outside managers, wrote a check. The quarterly 13F filings were treated as scripture; a retail app rolled out a feature to copy his disclosed trades; Tim Ferriss called him the Nostradamus of AI, “as close to clairvoyant as you could possibly be.” He wore dark turtlenecks and rarely appeared in public, which in this business is worth at least 200 basis points of mystique.

Then July 2026 arrived, the AI complex wobbled, and Situational Awareness lost 67% in a month.

Not 67% of profits. Sixty-seven percent of the fund. In one month. And here is a fact I will not hide from you, because we will need it later: even after that, the fund reportedly remained up roughly 80% for the year. Both things are true at once, and the arithmetic checks.¹ The letter to investors — written, per the Journal, as the wedding guests were arriving — has since been published in full, and I would encourage you to read it, because it is better than the coverage of it suggests. It opens: “We let you down this month,” a sentence that does a great deal of load-bearing work with the word “month.” It also contains the line “I take full responsibility for these events,” and it does not hide behind the market. It notes adverse trading in names publicly associated with the firm and likens the dynamic to a bank run — vulnerability begetting more vulnerability — but the framing throughout is that a fund must be structured to take a loss and survive. That is the correct lesson, stated plainly, by the man it cost the most. Hold that thought too; we will need it at the end.

So here is the question this post exists to answer: how does a man with no track record raise billions from the most sophisticated names in American finance and then lose two-thirds of it in thirty days? The short version involves two clocks running at different speeds, and we will get to them. But first, the arithmetic — because I promise you it is short.

IDIOT!

Let me teach you everything you need to know about risk management. It will take one paragraph. There are no prerequisites. There is no math beyond addition and multiplication. This is not Trading 101. This is Trading 000.

Start with the financing. One dollar of investor capital plus three borrowed dollars buys four dollars of assets. Lose 25% on those assets and the investor’s equity is gone — all of it. Borrow four dollars instead and the wipeout threshold falls to 20%. Options can push it lower still. Now the other half of the index card: any tech stock can drop 30% at any instant. Not “might, in a bad year” — at any instant. This is true of real, profitable, moat-owning giants; Google, Amazon, Apple and Meta have all done it, some of them inside a few weeks. It is certainly true of a Korean memory-chip maker, a fuel-cell company, and a cloud-computing provider whose common thread is that they all appear in the same manifesto. Now put the halves together, and notice the horror: the move that kills you is smaller than the move that happens routinely. That’s the whole course. It fits on an index card. And according to the Journal’s reporting, Situational Awareness was borrowing $3 to $4 for every $1 of investor capital — “or sometimes more” — with options layered on top for extra spice.

I should note, because it is too perfect to leave out, that Aschenbrenner had made a habit of deriding “wordcels” — his term for people who traffic in words rather than mathematics, and who therefore could not see the AI revolution coming in the numbers. He was not wrong about the numbers he was looking at. He was undone by a different set of numbers, requiring one operation, taught in primary school.

I can hear the objection forming: but he was hedged! He shorted the AI losers! Think, for one moment, about what that short book actually was. It did not hedge the AI thesis. It expressed the thesis a second time: long the companies expected to build the future, short the companies expected to be eaten by it. That construction may reduce ordinary market beta. It offers no protection against the one unwind that matters — AI longs falling while crowded AI-victim shorts are squeezed upward by everyone else covering to meet the same margin calls. Both legs lose. Everyone attempts the same exit through the same door.

You do not have to take my word for this. Pull up the tape for July, from the June 30 close to the July 29 close — the last session before the block trade cleared the book. His disclosed longs: Sandisk down 55%, Nebius down 46%, Bloom Energy down 46%, CoreWeave down 39%, Micron and IREN down 36%. Over the identical twenty sessions the enterprise-software complex went the other way: Workday up 37%, Adobe up 28%, Intuit up 28%, Salesforce up 20%, Veeva up 17%.⁴ He had been perfectly explicit about this trade — on Dwarkesh Patel’s podcast in June 2024 he said he was bearish on the software “wrapper companies” because they were “betting on stagnation.” Both legs. Simultaneously. Precisely as constructed.

Matt Levine put the general form of this well last week: a thematic long-short book has not reduced risk, it has rotatedit. You remain insulated against the market and completely naked to your one idea. Which is exactly what buyers found when the book came up for sale. Situational’s representatives reportedly told rival firms they were selling hedges — modest protective positions, now in distress. The rivals opened the file and found monster directional bets against companies like Adobe, forming a huge share of the firm’s total investments. They were hedges only in the narrow sense that they were short something while the fund was long something else. Against the risk that actually materialized, they were an accelerant.

You need not take my characterization of this either. Here is the man himself, in the letter: many AI names drew down by half or more, he wrote, while the fund’s positive long/short spread reversed violently. That is the whole mechanism, conceded by the person who built it. The longs fell, and the thing that was supposed to cushion the fall did the opposite, at the same time, on purpose, by construction.

And there is a name for what happened to the market that week, which turns out to matter more than “AI selloff.” Per BTIG, Morgan Stanley’s sector-neutral Momentum Index fell 17.4% in four trading days — the worst such move on record, surpassing the dot-com unwind, the pandemic shock and the 2022 bear market. BTIG’s chief market technician called it the largest and fastest momentum crash in modern history, and not particularly close. The momentum ETF had posted its best month ever in April and its worst on record in July. Situational Awareness was not really long AI and short software. It was long momentum and short anti-momentum, at four or five times, and momentum picked that fortnight to do the thing momentum does. Readers of the Emperor series will know why I find this less than shocking: momentum is not a stable factor you can lean against. It is a tradable measurement of non-stationarity, which is a fancy way of saying it works until it violently doesn’t, and nobody rings a bell.

And now the number that should be tattooed somewhere visible. Over that same stretch, the S&P 500 fell 2.32%.

Two and a third percent. Nvidia — supposedly the beating heart of the AI trade — fell 5%, and finished July up slightly. There was no crash. There was a rotation, out of the capital-hungry end of the AI complex and into the cash-generating end, which is to say out of everything he owned and into everything he was short. The broad market barely flinched, and a $45 billion fund lost two-thirds of its value. That is not a market event happening to a portfolio. That is a financing structure converting an ordinary Tuesday into an extinction.

And notice what kind of risk this really is, because the vocabulary matters. This was not ordinary stock-picking risk — not the kind a diversified book washes out. It was concentrated thematic risk wired directly into funding risk: the same market move that hurt the portfolio also raised collateral demands, made lenders nervous, and announced to every rival desk that a very large fund might soon be forced to sell. You can diversify away a company-specific accident. You cannot diversify away a financing structure that forces you to sell precisely when everyone knows you must. And here is the part that should genuinely worry you: I have watched experienced professionals fail this exam in person, with the answer key open on the table. Hold that thought.

You Levered a Thesis and Called It Alpha

Imagine a fund that does exactly one thing. Every morning the manager comes in, pours his coffee, sells put options on the S&P a few percent below wherever the market opened, and goes to lunch. That is the entire strategy. It will be up almost every single month. Its Sharpe ratio will be magnificent, its investor letters serene, its manager interviewed about his process and his edge. And then one day the market gaps down through his strikes and he loses, in one session, on the order of everything he has made since inception — quite possibly more.² Given enough repetitions the tail stops being theoretical, and it need arrive only once before your financing horizon ends. This is called picking up pennies in front of a steamroller: the pennies are not small in dollars — they can be 439% in six months — they are small relative to the steamroller.

Now, a levered long book is not literally a short put; a quant can draw the two payoff diagrams and have me arrested before lunch. But under a margin constraint it sets the same trap for the allocator. Calm markets produce repeated gains and the financing tail is simply absent from the sample, so the returns look smooth, the Sharpe looks superb, and the allocator concludes he is watching skill. Then a drawdown raises collateral demands, forced sales worsen prices, rivals smell the seller, and a linear mark-to-market loss becomes a path-dependent fight for survival. Static leverage magnifies losses. The margin clock makes them nonlinear. Every month the market failed to gap down, Situational was collecting the financing premium and booking it as genius — until it was clawed back in a single July, at the lows, at a 10%-plus discount, by a buyer who had been standing beside the steamroller for decades. We will get to him.

And there is a mechanical detail here that deserves more attention than it gets, which Matt Levine has made in print twice now, first about Bill Hwang and again this week. When you borrow to buy stocks and the stocks go up, you deleverage automatically: the same debt now sits under a larger pile of assets. So a manager committed to a target leverage level must, every morning after a good day, borrow more and buy more. That is a machine for buying high. Run it for two years in a bull market and you arrive at the top of the move holding the largest position you have ever held, financed by the most debt you have ever owed. Situational returned something like 1,000% since inception and was reportedly still running at four to five times. The natural question — after a 1,000% run, why would you still need the leverage? — apparently did not come up.

The lenders, to their credit, did eventually ask it. Bloomberg reported that as its positions were making all-time highs, Situational went looking for more prime-brokerage capacity, and Barclays turned the firm away out of concern about its exposure to a single sector. Sit with that. Before the margin calls, before the rout, before any of this was obvious — a bank looked at the book and said no. The risk was visible from the outside, in advance, to a party whose only job is to ask what happens when the collateral falls.

Why can’t the thesis save you? Because — put this on the index card too — your thesis cannot post margin.Aschenbrenner’s thesis may well be right. AI may eat the world roughly on his schedule; I suspect much of it will. Irrelevant. The thesis runs on a decade. The margin clock runs on T+1. Correctness on the slow clock buys you nothing — nothing — on the fast one. Many of LTCM’s positions eventually recovered enough for the rescue consortium to unwind the book in an orderly way. That did not restore the partners’ capital, and it did not make the financing structure sound. The market does not pay you for being right about the future. It pays you for remaining solvent while being right, and those are different products with very different prices.

Now the best objection to everything I have just written, which I have seen made forcefully in the past few days: leverage is only a magnifying glass. It explains how fast he lost, not why. He lost because his thesis was wrong — wrong about the software companies, which turn out to sell the rails rather than the labor, and wrong about whether the capital that builds a transformative network ever earns its cost. All true, and worth its own essay. But it is answering a different question than the one a risk system exists to answer. Being wrong is not an anomaly in this business; it is the base case. Everyone is wrong, regularly, and the good ones are wrong perhaps four times in ten. Plenty of managers were long that same second-derivative complex in July and are sitting at their desks this morning, bruised and employed. What distinguished Situational Awareness was not the error. It was that the error was fatal. At one times capital, being that wrong costs you a quarter of the fund and a humiliating letter. At four or five times, with lenders holding a same-day option on your survival, the identical wrongness ends the firm. The thesis determines whether you lose money. The financing structure determines whether you survive losing money. Only one of those is under your control on the morning it matters.

And this is not merely a saying. It is a theorem, and Cliff Asness reached for it within days of the collapse: you never go full Kelly — a phrase he suggested filing under things of which one should be situationally aware.

The Kelly criterion answers the question “given a genuine edge, how much should I bet?” The answer is a specific fraction of capital, and the fraction is finite. Bet less than Kelly and you grow more slowly than you might have. Bet more than Kelly and something remarkable happens: your growth rate starts falling, even though your edge has not changed at all. Bet exactly twice the Kelly fraction and your long-run growth rate is precisely zero. Bet more than twice and it goes negative — you now lose money, with certainty, over time, while continuing to be right.

Read that again, because it is the entire post compressed into a sentence of mathematics. Beyond a threshold, position size alone converts a winning strategy into a losing one. No change in the thesis is required, no error of analysis: the edge can be real and durable and exactly as advertised, and the overbet destroys you anyway. I ran the numbers to be sure — with a healthy edge, sizing at three times Kelly leaves the median outcome below where you started. The strategy works. The bettor is bankrupt. Which is why professionals bet a half or a quarter of Kelly: you never know your edge exactly, and overestimating it pushes you past the ruin threshold without your noticing.

So when the Journal reports borrowings of three to four dollars per dollar of capital, with options on top, in a concentrated single-theme book — the question is not whether he was right about AI. Kelly does not care whether you are right. Kelly cares how much you bet on being right, and it has a number, and the number was exceeded by a wide margin.

The truly damning part is that he had already taken the practice exam, and had been handed the answer besides. January 2025: DeepSeek drops, Nvidia and the whole AI complex convulse, Situational takes real losses. The fire alarm went off. He stood in the parking lot and concluded that the lesson was dips recover. And well before that, in June 2024, on his friend Dwarkesh Patel’s podcast, Patel had raised the shelf of business-history books about investors who got the thesis right and the timing wrong, and asked whether he had thought about how these things blow up. His answer: “not blowing up is task No. 1 and 2.” He knew the words. He did not know the thing the words refer to. On July 24, one week before the fire sale, he reportedly wrote to investors that the sell-off was creating some of the most attractive opportunities since early 2025 — which, in a certain light, it was. Just not for him.

So Who’s the Idiot?

Here is where I am supposed to pile on the 24-year-old, and believe me, the material is there. But let’s be honest about what happened. A 24-year-old with no investing experience, handed billions in a raging bull market for his exact theme, levered it to the moon and blew up. Is that surprising? That is the base case. That is what the model predicts. He performed his role in this drama flawlessly. He at least has the excuse of being 24.

The more interesting question is the checks.

The Collison brothers. Daniel Gross and Nat Friedman. Later, Jane Street — with minimum check sizes reported at $25 million for some investors, and multi-year lockups. What were they actually buying? Strip away the mystique and the honest answer is: nobody outside could tell, and that is precisely the problem. The 1,000%-plus since inception did not, by itself, identify skill. It was compatible with skill, with genuinely privileged private access (the early Anthropic stake was, to be fair, a real coup), with levered thematic exposure in a roaring bull market, with luck — or with some combination the headline number could not disentangle. Readers of the Emperor series will recognize this instantly: it is an identification problem, the same one I’ve spent three posts (so far — the emperor has many more clothes to not wear) documenting in the academic literature. The professors run factor regressions and call the residual skill. The allocators looked at a return that was compatible with SHORT FINANCING PREMIUM in forty-foot letters and called it foresight. A high Sharpe measured before the tail is not a tail-risk model. Steady gains plus a rare catastrophic loss produce a beautiful Sharpe right up until they produce a crater — and the beauty of the Sharpe is precisely the evidence that you have not yet seen the tail.

Now, in fairness: I cannot see Jane Street’s position size, side letters, or portfolio construction. Perhaps they knowingly bought a small, carefully sized slice of exactly this tail; perhaps the venture access dominated the economics; a $25 million check can be socially enormous and economically trivial to a firm of that scale. It is also reported that when the book came up for sale last week, Jane Street looked at it and passed — which tells you something about how they price risk when they can actually see it. And I would extend them a courtesy I would not extend to an allocator: Jane Street is a proprietary trading firm. It was risking its own capital, not somebody’s pension. A prop shop taking a sized bet on a volatile young manager and eating the loss is not a scandal; it is Tuesday. The people who owe someone an explanation are the ones who were investing other people’s money. But that is exactly why the rest of the market could not rationally treat the Jane Street name as a substitute for its own arithmetic. Their check revealed almost nothing about how much risk they had accepted. In public, however, the name functioned as permission.

My diagnosis — and I offer it as diagnosis, not as a transcript of anyone’s mental state — is that in the Valley, diligence is a social graph. Once the Collisons are in, Gross and Friedman follow; once Jane Street allocates, the questions stop. The problem was not necessarily that no one had done diligence. It was that each successive investor treated someone else’s participation as evidence whose underlying calculation remained invisible. It was social proof all the way down.

And now the detail that turns that diagnosis from a hunch into something closer to a controlled experiment. When Aschenbrenner took the fundraising tour east, New York was distinctly unimpressed. Rob Copeland reports in the New York Times that three people he pitched declined, describing him as a lightweight and a one-hit wonder; Blackstone, the largest hedge fund investor on earth, passed. One investor who took the meeting asked him the single question the entire enterprise turned on — what is the plan if the AI revolution doesn’t arrive as hoped? — and reports getting no detailed answer. Aschenbrenner simply believed it would work out.

That is the whole post in one anecdote. The same manager, the same pitch, the same numbers, presented to two populations. One asked what happens if you’re wrong, got no answer, and declined. The other had built its fortunes on being right about the future and heard, in his certainty, a familiar and admirable sound. The money that got destroyed was not the money that failed to understand AI. It was the money that understood AI and had never been asked to price a margin call.

The easy version of this is that New York is smarter than San Francisco, and since I have spent thirty years running money an hour outside New York, you should discount me accordingly. New York is not smarter. New York bought Amaranth, funded Archegos through five separate prime brokers none of whom knew about the other four, and invented Long-Term Capital Management. What New York has is not judgment. It is scar tissue. Those allocators have personally been carried out before, which is why “what happens if you’re wrong” is not a clever question for them but a reflex, asked in the same tone as “where are the fire exits.”

Which is to say: the split isn’t geographic at all. It is variolation, applied to allocators rather than to traders. The New York money got its inoculation in 1998 and again in 2008 and has the antibodies to show for it. The pool of capital that has flooded into the AI trade has, structurally, never seen the down. It is not stupid. It is immunologically naive, which is a different condition with the same symptoms and a much better prognosis, because this is exactly how you acquire immunity. Last week was the first dose.

And the same applies, with more force, to the manager. The Financial Times made the point with admirable dryness: in his defense, Aschenbrenner was in his infancy when the financial crisis hit. He was six or seven years old in 2008. Every market he has ever traded has gone up. This is not an excuse — the arithmetic was available to him, in books, at any point — but it is a mechanism, and mechanisms predict. If you want to know which manager blows up next, do not look for the one with the worst thesis. Look for the one who has never personally watched a thesis fail to matter.

And here is what the discipline looks like when it has been earned. Some years ago, a Nobel laureate in economics came around pitching a new fund to a large institution I know well. The institution asked the ordinary questions — what is the strategy, what is the thesis, what is the risk management. His answer, in substance: I have a Nobel Prize. Invest with me.

They passed.

That is the entire discipline in one meeting. Not cleverness, not superior insight into whether the man was right — he may well have been right, he had a Nobel Prize — but the refusal to accept a credential as a substitute for a process. And note the exquisite detail that the pitch ignored: the most spectacular leverage-driven collapse in the history of finance, the one everybody on Wall Street reached for the moment Situational Awareness started sliding, was run by a firm with two Nobel laureates in economics on its masthead. In this particular industry, the prize is not merely insufficient evidence. It has a track record.

“We’re going to have way more situational awareness than any of the people who manage money in New York” is the same sentence. It is a credential offered where a process was requested. The people who manage money in New York had, in fact, considered his credential and passed — not because they knew more about AI than he did (they certainly did not), but because he could not tell them what would happen if he was wrong, and they had all met someone before who could not answer that question.

None of which is an argument against ambition. There is nothing whatever wrong with wanting to change the world, and nothing wrong with taking outsized risk to do it — the entire venture model is built on outsized risk, correctly sized and honestly disclosed, and it has produced more value than any other allocation of capital in modern history. There is, however, something wrong with being an idiot about the financing. Those are separable, and the separation is the whole ballgame.

But here is the detail that should end the argument about whether anyone was misled. The Times reviewed an investor document in which Situational Awareness stated that it would set no limits on the types of investments it might make, nor on the concentration of its investments, nor on the amount of leverage it might use. No limits. In writing. In the documents. Before a dollar went in.

So let us dispense with the idea that this was hidden. It was disclosed, in the plainest possible English, in the one document every allocator claims to read. Nobody was deceived about the leverage. They were told there would be no limit on it, and they wired the money anyway, because the returns were 439% and the manager was clairvoyant and the Collisons were already in. That is not a failure of disclosure. It is a failure to believe the disclosure — which is a much older and much more human problem, and one no regulation has ever fixed.

One more distinction the headline numbers erase: return since inception is not investor experience. The LP who wired money at launch is still, even now, sitting on a multiple. The LP who wired money in May, at peak mystique, has been vaporized. Same fund, same manager, same thesis — entirely different outcomes, sorted purely by entry date. When the marketing quotes “1,000% since inception,” ask which investor actually lived it.

You want to know how deep the underlying confusion runs? It is not a Silicon Valley failure. It is not an amateur failure. Let me tell you about the professionals.

The Exam, Administered in Person

Years ago I consulted for a well-known hedge fund. Experienced traders. Real careers, real track records, people who say “risk-adjusted” without air quotes. They had a position they were very pleased with: long Google, short eBay. Dollar neutral. The market could do whatever it wanted. Low risk, they told me.

I gently pointed out that this trade was riskier than simply buying the S&P.

They did not believe me. Fair enough — extraordinary claims and all that. So I ran a full-scale bootstrap analysis and put the distributions side by side. In the sample we tested, the loss distribution of the Google–eBay spread was worsethan the S&P’s. Because dollar neutrality is not beta neutrality, and beta neutrality is not risk neutrality. They had cancelled a broad, diversified source of variation — the cheap risk, the kind the market pays you to hold — and retained a concentrated residual spread containing two distinct ways to be unpleasantly surprised on the same morning.³ Google could miss earnings while eBay caught a takeover bid, and the trade would lose on both legs simultaneously, on a day the S&P closed green.

They looked at the bootstrap. They looked at me. And they left the room muttering: but it’s dollar neutral... how can it be risky... I don’t get it...

They were right about exactly one thing.

These were professionals with decades of combined experience, and they could not price the difference between “dollar neutral” and “risk neutral” with the answer key on the table. Now take the same misunderstanding, hold it with the same sincerity, run it at four-to-five times assets over equity, scale it toward $45 billion, and make the short book express the same theme as the longs. The Google/eBay boys at least had the decency to run it small.

The Variolated Shark

One more thing, because the universe has a sense of humor and I would hate for you to miss it.

Who bought the forced sale? Ken Griffin. And Griffin did not invent this playbook last week. In September 2006 he and JPMorgan took Amaranth’s energy portfolio off its hands after $6 billion of natural-gas losses. In July 2007 he bought Sowood’s distressed book — a deal reportedly agreed at half past three in the morning, because Citadel’s team was still working when everyone else had gone home. Note the hour. Some habits are load-bearing.

Then, in 2008, Citadel itself nearly died. The flagship Kensington and Wellington funds lost roughly 55%, the firm was running something like seven-to-one leverage, the correlations that were supposed to be independent went to one, and Griffin gated redemptions for the better part of a year to avoid liquidating into a frozen market. Counterparties wondered aloud whether the firm would survive. It did — in part because Citadel had studied LTCM closely and pre-negotiated far stronger financing terms with its prime brokers, which is the single most underrated sentence in this entire post. The firm returned 62% the following year.

And what has Griffin said about 2008 since? He told Institutional Investor that the firm’s large leveraged balance sheet had been its Achilles heel, and that the episode was incredibly humiliating. He has repeatedly accepted the blame, called it his low point, and admitted he had not understood how precarious the banking system was. He did not, so far as the record shows, blame the short sellers. That is what separates a variolation from a grievance.

He is not alone. Stanley Druckenmiller — as good as anyone has ever been at this — piled into a handful of tech names at the top of the dot-com mania and got destroyed. His own verdict, delivered years later: the first thing he was ever taught in the business was that bulls make money, bears make money, and pigs get slaughtered, and he was here to tell us he had been a pig. That is a man who has been inoculated. You can hear it in the tense.

Because that is what 2008 was: a sublethal dose, survived, conferring immunity. And it is precisely because he has been variolated that Griffin was ready this July when a 24-year-old’s phone needed answering at midnight — the standing playbook, the fortress liquidity, the patience to wait beside the steamroller for however many years it takes. The apex predator in this business is not the one who never came close to blowing up. It is the one who came close once, at survivable scale, and never forgot.

The staging was almost too neat. Griffin is 57. Izzy Englander, circling on behalf of Millennium the same night, is 77. Both were phoning in from Europe well past midnight. On the other end was a 24-year-old whose fund was named for his ability to see what others could not. eFinancialCareers ran the story under the observation that the wunderkind had been saved by the old guys, which is the proverb at the top of this post reduced to a headline. And the trade was, for Citadel, a fat pitch of exactly the kind Griffin has been swinging at since Amaranth: Bloomberg reported that Citadel’s Wellington fund had been up less than fifty basis points for the month through July 24 — a nothing month — before it bought a portfolio at a double-digit discount and watched those names rebound better than 20% the following day. Patience beside the steamroller does not merely keep you alive. Periodically it pays the rent.

You see now what the proverb at the top of this post actually is. It is not a warning. It is a census. There are no old bold traders because the market runs the selection experiment continuously — the fragile are removed, and the survivors are, by construction, the ones who learned. And note the fine print, because Citadel is hardly a monastery devoted to timidity: Griffin got old not by ceasing to be bold, but by separating boldness from fragility. The census does not count the bold. It counts the bold who could not survive being wrong on the wrong day.

Class Dismissed

Back to Carmel, briefly. I hope the wedding was lovely — by every account the couple asked only that nobody bring gifts, and I find that genuinely endearing. But if the breakout sessions needed a discussion topic, the groom’s own investor letter supplied one. “Vulnerability begetting more vulnerability,” he wrote of the market’s dynamics — apparently without noticing he was describing his financing structure, not the market. The market supplied the shock, and what a modest shock it was: the S&P fell 2.32%. The market does that on a schedule it never conceals. The leverage is what converted an ordinary month into a survival event.

So: was he an idiot? A Wall Street source gave the New York Post the verdict that will probably stick — a leveraged idiot who was right until he was wrong. It is a good line, and it is very nearly correct; I would only move the emphasis. The leverage was idiotic — that is simply what the arithmetic says, and the arithmetic is not open to interpretation. But he was bold, in the precise actuarial sense of the proverb: bold traders are the ones the census has not counted yet. And the census, this time, has not quite counted him. The fund survived — because Citadel provided an exit for the public book, at a price his investors will remember. It reportedly remains up roughly 80% for the year, has kept its Anthropic stake, and has removed its leverage. Early investors may still be substantially ahead. None of this vindicates the risk system. It separates two questions that spectacular P&L had temporarily fused: did the trade make money, and was the amount of capital placed at risk rational? A bridge can survive an overload. That does not validate the engineering.

He is 24. The tuition has been paid — in full, in public, largely with other people’s money. The expensive question is whether survival gets read as vindication or as instruction, and the letter is genuinely better than I expected on this. He takes full responsibility, in those words. He closed every short and removed all leverage. He now runs the public book fully paid for — long stock and long fully-paid-for options, no margin, no liquidation risk. He writes that the fund must always be structured so that it can take a loss and fight another day, which is precisely the right sentence.

And then there is one clause, and it is the whole ballgame. The public book will be run on a fully-paid-for basis, he writes, while we draw the lessons from these developments. While. The Times noted the same thing: he stopped short of pledging not to borrow again. So the leverage is not gone; it is suspended, pending the completion of learning, on a schedule he sets himself. Somewhere in the next bull market there is a morning when the lessons will feel sufficiently drawn, the opportunity set will look extraordinary — the letter already says the current opportunity set looks extraordinary — and the prime brokers will be delighted to hear from him. That morning is the exam. Everything before it is the review session. Victor Niederhoffer, having sold naked puts into the crash of October 1997, sued the exchange alleging that floor traders had colluded to drive the market down and force him out; ten years later he blew up again. The scars were expensive. They are only invaluable if they are permanent.

And I am afraid the surrounding ecosystem is not helping. Within seventy-two hours the comeback chorus had assembled — don’t count him out, the man has a track record, he still has billions and a strong view of the future. One prominent voice put it almost perfectly: if he is right, and keeps playing the risk-taking game, there is a high probability he generates tens of billions for his partners.

If he is right.

Read that sentence and then read this post again. He was right. He was right about AI for two solid years, right enough to return more than 1,000%, right enough that the Journal wrote him up as an oracle and a retail app let strangers clone his trades. Being right is what got him to $45 billion. Being right is not what took it away, and being right is not what will bring it back. The chorus, seventy-two hours after watching a fund lose two-thirds of itself in a month, has already reconstructed the exact syllogism that produced the wreckage: he sees the future, therefore he will make money. That is the error. Not the leverage — the leverage is downstream. The error is the therefore.

The index card, one last time. One dollar of equity plus three borrowed dollars buys four dollars of assets; lose 25% and the equity is gone. The thesis may be brilliant. The future may arrive exactly on schedule. Your thesis cannot post margin. There are old traders, and there are bold traders.

Class dismissed.

¹ Reuters reports the fund gained about 439% through June, then lost 67% in July: 5.39 × 0.33 ≈ 1.78, i.e. up roughly 78–80% on the year. The reported numbers cohere, which in stories sourced to “people familiar with the matter” is a small miracle worth savoring.

² How much more depends on strike distance and sizing, so I claim the direction and not the magnitude; the universal statement is that the one bad day is of the same order as the accumulated winnings, and it need only arrive once. The graveyard offers confirming instances rather than a theorem. Victor Niederhoffer, having sold naked S&P puts alongside puts on Thai bank stocks, was wiped out on October 27, 1997, when the Dow fell 554 points — 7.2% — and the resulting margin call exceeded roughly $130 million he did not have; his Matador Fund then blew up again in September 2007, down more than 75%, on rising margin requirements against overexposed options positions. Or consider XIV, the inverse-volatility note: on February 5, 2018, it fell from $108.37 to $4.22 — about 96% — on a day the S&P 500 was down all of 4.1%. Credit Suisse announced the acceleration the next morning. Read that pair of numbers again: 4.1% and 96%. That is what negative convexity does to you, and it is the same arithmetic as the leverage in this post, wearing a different hat.

³ For the quants: the spread variance is σ_G² + σ_E² − 2ρσ_Gσ_E. Cancellation requires the right covariance structure, not equal dollar signs on opposite sides of a blotter. Remove the common factor and what remains is the residual risk of both names — the risk the market does not pay you to hold.

⁴ Price moves are my own, computed from daily closes, June 30 to July 29, 2026. The quoted remarks about the wrapper companies are from the June 4, 2024 Dwarkesh Podcast episode, as is the exchange about blowing up. One necessary caveat on the positions: 13F filings disclose long positions in U.S. stocks, not short positions, so the long basket above is a matter of public record while the short basket is not. Adobe has been reported as a short; the remaining software names are the complex he was on record as bearish about, not confirmed positions. The direction of the argument does not depend on the exact roster — what matters is that the theme he was long and the theme he was short moved violently in opposite directions, against him, at the same time.

A note on sourcing, and on a genuine discrepancy. Peak assets are reported at roughly $45 billion by the Journal and Business Insider, but at around $30 billion by the New York Times; what remains is put at about $10 billion by the FT and about $8 billion by the Times; the stock sold over those thirty hours is $16 billion per the FT and roughly $20 billion per the Times. I use the larger figures where the Journal supports them and flag that the range is real — which is itself instructive, because a firm whose size cannot be pinned within 50% a week after the fact is a firm whose risk nobody outside was in a position to measure. The 67% July loss, the $3–4 of borrowing per $1 of capital, the 10%-plus Citadel discount, the ~439% through June and ~80% year-to-date come from reporting by The Wall Street Journal, Reuters, the Financial Times, Bloomberg, CNBC and the New York Times between July 30 and August 3, 2026, much of it attributed to people familiar with the matter: the WSJ’s wedding-week reconstruction, Citadel purchase, July losses and June profile; Reuters on the July letter and year-to-date figures and the Citadel transaction. The investor letter is quoted from the full text published by Business Insider; I have relied on the letter itself rather than on characterizations of it, and recommend the same to anyone writing about this. The Barclays refusal and the leverage-seeking at all-time highs are Bloomberg’s reporting (Burton, Natarajan, Gillespie and Kumar); the New York fundraising reception, the Blackstone pass, the Goldman margin call, the mislabeled “hedges,” the “no limits” investor document and the Pfeffer quote are Rob Copeland’s in the New York Times; the momentum-crash figures are BTIG’s, via CNBC; the Citadel Wellington month-to-date figure, the ages of the principals and the “saved by the old guys” framing are Sarah Butcher’s in eFinancialCareers, citing Bloomberg and the FT; the comeback-chorus remarks were collected by Business Insider; the Asness remark on full Kelly, Griffin’s Institutional Investor comments, the Druckenmiller quote, the FT’s line about Aschenbrenner’s infancy and the New York Post verdict all come via Joe Nocera’s account in the Free Press; the deleveraging-ratchet observation and the rotated-risk framing are Matt Levine’s in Money Stuff, where he has been making the first of them since Archegos in 2021. I claim directions confidently and hold magnitudes loosely — though note the verdict is insensitive to the error bars: the arithmetic condemns the financing at 3x or at 5x, at $30 billion or at $45 billion, which is rather the point.

If you enjoyed watching a cherished story about foresight collide with arithmetic, my book, The Science of Free Will, asks an equally uncomfortable question about an equally cherished story: what if seeing the future — even correctly — was never the same thing as controlling it?

New here? Start with The Paradox of India or browse the full India series. And if institutional decline is your genre, Albion is about Britain’s version — which I first saw coming in 1979.

The Emperor series resumes shortly. The news declined to wait.

But set the insult aside, because underneath it is a real question, and it turns out to be one of the deepest questions in biology. For those of us who are smarter than a chimpanzee — what, precisely, is the difference? Not “we have bigger brains,” which is true and useless. I mean mechanically. You and a chimpanzee both start life knowing almost nothing and end it knowing a great deal. You both got from the first state to the second by learning. So the difference between you and the chimp is not that you learn and it doesn’t. It’s how. And the how is stranger, and more specific, than you think.

It begins with a paradox you can verify at any hospital.

The paradox in the maternity ward

A human being, the smartest animal that has ever lived — as far as we can tell, which is a hedge I’ll come back to — does not know what to do with its own baby. It has to be told. There are classes. There are books. There is, in the more affluent precincts of the species, a woman you pay by the hour to come to your home and explain why the baby is crying, whether the crying means anything, and whether your soul will ever fully re-enter your body. First-time parents stand over the crib at 3 a.m. googling whether that breathing pattern is normal. On this one task — the single most important thing any organism ever does, the thing evolution cares about with the cold intensity of a tax auditor — we are spectacularly, visibly incompetent.

Now consider a solitary wasp. It will die before its offspring hatch. It has a nervous system you could sketch on an index card. It never met its mother and never will. And it digs the burrow, hunts the caterpillar, stings it into paralysis rather than death so the meat stays fresh, seals the egg in with its living larder, and flies off to die — the entire program of parenthood executed once, flawlessly, with nobody to copy and nothing to learn. Zero training. Zero doubt. No 3 a.m.

So the conclusion writes itself, and it is the kind of conclusion that feels wonderful because it flatters us: every other animal just knows what to do with its young, and only humans have to be taught, and that is the clean bright line between us and the rest of creation. Instinct traded away for intelligence. The price of the big brain.

I believed a version of this for years. It is a lovely theory. It is also wrong — and, as usual, the way it is wrong is more interesting than the theory was.

The animal that should have made the case, and didn’t

Start with the counterexample a biologist reaches for first, because it looks like it should settle the matter. The theory says instinct is the crude, simple option and learning is the sophisticated one — so the most instinct-bound animals should have the simplest parental behavior, and anything elaborate should require teaching.

That is just false, and one bug is enough to show it.

The white-margined burrower bug, Sehirus cinctus, guards her eggs and then forages to bring seeds back to her newly hatched young — real provisioning, adjusted to how hungry the brood is (documented maternal behavior). It is genuinely elaborate. It is also not taught. She is not handing down a recipe — not listen carefully, Brenda, the mint seeds are best in late May. The behavior is species-typical, built into the bug’s inherited developmental program, run more or less the same way by every female of her kind whether or not she ever watched another do it. Which makes the useful point, the one to carry forward: elaborate behavior is not the same as learned behavior. Complexity does not imply a teacher. A great deal of impressive parenting, across the whole animal kingdom, is simply run off the genome — no watching, no instruction, no tradition. So the question is not whether behavior is complex. It is where the information lives.

Now swing to the animal nearest us: the great ape. If the human helplessness at the crib were really the price of our particular intelligence, it should be ours alone — a gorilla ought to have kept the instinct we supposedly traded away, and mother competently the way the wasp does.

It doesn’t. A great ape is as helpless with its first infant as we are with ours, and for the same reason: it never learned. A captive gorilla raised without watching other gorilla mothers frequently cannot mother at all — it won’t hold the baby, won’t nurse, has to have the whole thing taught to it. This is common enough that zoos run programs for it. The Smithsonian’s National Zoo has coached an expectant mother through holding and nursing using photos of gorilla mothers and a plush infant that plays newborn calls (their account); the Cincinnati Zoo has gone further, putting human surrogates in faux-fur vests to carry rejected infants, vocalize like gorillas, and teach the baby to cling as a gorilla mother would (documented by the zoo). We are, and I want you to sit with this, running classes for gorillas on how to be a mother — the same remedial course we run for ourselves.

So the gorilla is not a case below the human; it is the same case as the human. Both are big-brained, elaborately capable animals whose mothering is not shipped with the body but has to be picked up from other members of the species — and both fail, identically and helplessly, when it isn’t. (A hand-reared ape is damaged in many ways at once, not just maternally, so no single case settles exactly how much rides on learning. But the pattern is not subtle, and it repeats.) The bright line we were promised — humans who learn on one side, animals who just know on the other — has quietly stopped falling between the species and started falling inside them: between the individuals who got to learn and the ones who didn’t.

And the gradient runs down through the mammals. First-time mothers across a huge range of species are measurably worse at the job — less milk, more stillbirths from botched deliveries, higher infant death — and they improve with each birth. Not because the mother got smarter between pregnancies. Because she got experience. The competence that looked like a birthright is, across much of the mammal line, at least partly a skill.

Which means the wasp and the gorilla are not two marks on one ruler. They are running two different machines. And that is the question worth an essay — the same question we started with, sharpened to a point:

When a species could transmit knowledge by learning — when it lives long enough, and is smart enough, that teaching its young is genuinely on the menu — does evolution take the option? Or does it keep using the genome anyway?

Where the information lives

To answer that, stop thinking about babies for a moment and think about information — the stuff a parent has to get into a child: what to eat, what to fear, how to handle prey that fights back, how to raise the next one in turn.

Behavioral information reaches an animal by three broad routes. Some is built in by natural selection — written into the developmental program, arriving with the body, no experience required. Some is acquired by individual learning — the animal works it out itself, by trial and error. And some arrives by social learning — from other individuals, by watching them or being shown.

Those routes differ in cost and in reach, and the differences decide everything.

The built-in route is a magnificent medium with one crippling limitation. It is reliable, it is instant, it works even if the parent is dead before the child hatches — the wasp, the burrower bug, the monarch butterfly that navigates from Canada to a grove in Mexico it has never seen, on sun and magnetic cues wired in before birth. But it cannot be updated within a lifetime. You get the program your ancestors got. If the world has changed since they wrote it, you lose.

So the moment a species lives in a world that changes faster than the genome can track — shifting prey, new predators, new terrain, other minds — it has to move some of the information off the genome and into the living animal, through learning. Individual learning is one way, and it is cheap for everyone except the learner: a New Caledonian crow works out in a water-displacement task that dropping stones raises the water and floats a treat within reach, with nobody to show it — but trial and error is slow, and lethal when the first mistake is the last one.

Social learning fixes that by letting the naive animal skip the dangerous trials and crib off someone who already knows. Most of it is cheap for the knowledgeable animal too, because it isn’t doing anything extra — it’s just living its life in view. A wolf cub learns the hunt by tagging along while the pack hunts; the pack isn’t teaching, nobody slows the elk down for pedagogy, the hunt simply happens in front of the cub and the cub picks it up.

And then there is the expensive corner of social learning — teaching. This is the one case where the knowledgeable animal pays: it changes its own behavior, at a cost or for no immediate benefit, specifically so the naive one learns faster or more safely. That single fact — that teaching costs the teacher — is the hinge the whole question turns on. Because a cost is exactly the kind of thing natural selection refuses to pay unless it has to.

Nature is a tight-fisted CFO

So: a species that could teach — long-lived, iteroparous, young underfoot for years, a brain plainly up to the task. Does evolution reach for teaching?

Overwhelmingly, no. It leaves the capacity on the shelf and pays out of a cheaper channel.

Here is how we know, and it’s a genuinely startling fact. Biologists have a strict operational definition of real teaching — the Caro-Hauser criteria: the knowledgeable individual modifies its behavior only in the presence of a naive one, at a cost or with no benefit to itself, and the pupil learns faster as a result. Apply that definition honestly across the animal kingdom and the confirmed teachers are not the roll of brainy, long-lived species you’d expect. The classic experimentally-supported cases are meerkats, pied babblers, and tandem-running ants, with a thin and contested tail of proposed others; reviews still treat unequivocal evidence for teaching as scarce next to the far wider spread of ordinary social learning, and researchers keep arguing over whether the definition is too strict or exactly strict enough. Meanwhile elephants, wolves, most primates, the corvids — long lives, big brains, offspring around for years, every prerequisite satisfied — mostly do not show rigorous teaching at all. The capacity is everywhere. The use is rare.

And now the detail that flips the table. Look at who is on the list. An ant. A meerkat. These are not the intelligentsia. Teaching, it turns out, does not require a big brain at all — Alex Thornton, who ran the experiments proving meerkats teach, likes to say the whole point is that meerkats teach despite not being clever. So intelligence is not the switch that turns teaching on. Cognitive capacity opens the door; evolutionary economics decides how often anything walks through it.

Teaching evolves only when the young animal cannot get the skill any cheaper way — when the genome can’t carry it, watching can’t convey it, and trial-and-error would kill you before the lesson landed. Only when all the cheap channels are exhausted does evolution, grudgingly, pay for the expensive one.

Watch it happen in the one case that clears the bar. A meerkat pup has to eat scorpions. Scorpions have a stinger that can kill the pup. You can’t hardwire scorpion-handling — live scorpions vary too much to reduce to a fixed program. You can’t leave it to trial and error — the first error is the last. And watching an adult do it at full speed is no use to a pup that isn’t ready. So the adult does the one thing left: it bites the sting off, hands the disabled scorpion to the pup to practice on, and makes it progressively harder as the pup improves. That is teaching in the full technical sense, performed by an animal nobody would call bright, purely because for that one lethal skill there was no cheaper channel. The instant a cheaper channel exists, evolution uses it instead. It does not buy teaching because it can. It buys teaching when it’s cornered.

That’s the answer to our question. Capacity does not summon the behavior. Cost does. A creature can be a long-lived genius and be taught nothing by its parents, because everything it needed either fit in the genome or came free from watching. And a dim little meerkat teaches, because for one scorpion-shaped problem, nothing cheaper was available.

The child who copies nonsense

So the honest version of the human paradox is not “animals don’t teach and we do.” An ant teaches. It’s stranger than that — and here we finally meet the chimpanzee.

The classic nonhuman cases are strikingly domain-specific. The tandem-running ant teaches one route to one food source. The meerkat teaches one prey-handling trick — dead prey, then disabled, then live, as the pup matures. The pied babbler teaches its chicks to link one “purr” call with food. Each looks less like a school than a single-purpose appliance, installed because that one skill had no cheaper route. Human teaching is the opposite: general-purpose. We teach prey handling, mathematics, manners, music, metaphysics, and the correct way to load a dishwasher — sometimes all before lunch.

But teaching, on its own, still isn’t the deepest trick. Here is where the chimpanzee earns its place in the title. Put great apes and human children through the same broad battery of cognitive tests — and Herrmann and colleagues did, with chimps, orangutans, and two-and-a-half-year-old children — and the children were not simply better at everything. On the physical world, understanding objects and space and quantity, the toddlers and the apes came out about even. The children’s advantage was concentrated in social cognition — the skills for learning from and understanding other minds. (Note the boundary: this is toddlers versus apes on one battery. It does not show that adult humans barely out-think a chimp about the physical world. It shows where the human edge already lives at two and a half.)

Which is, at last, the hedge I promised at the top. “The smartest animal that ever lived” is doing quiet work, because our advantage is not what the phrase implies. It is not a bigger general-purpose processor grinding out every problem better; on raw physical-world cognition a two-year-old and an ape are startlingly close. Our edge is specialized. We come with in-built wiring an ape essentially lacks — above all a hunger to read other minds and a brain pre-built for language, neither of which is cultural, both of which are as much a part of our biology as an eye. And that specialized wiring is precisely what lets us do the thing the rest of this essay is about: it is the socket into which the external inheritance plugs. Not a better engine. A different one — one built, from the factory, to be finished by other people.

The sharpest demonstration is a puzzle box. Show a chimp how to open it, with some steps that are visibly pointless — tap here, wave there, then slide the latch. When the box is transparent and the causal structure is visible, the chimp drops the pointless steps and just gets the food. When the box is opaque and the causes are hidden, the chimp copies the pointless steps too. The human children copied the full sequence either way — transparent or opaque, causes visible or not.

In the transparent case, the chimp is the better engineer. It sees the pointless flourish and says, in effect: cute ritual — I’ll take the snack. The child looks dumber; it copies the flourish anyway. But that “dumbness” is the whole trick, and the difference is strategic, not a matter of ability. Chimps can imitate — in the opaque case they copy the useless steps too. What they are far less inclined to do, especially when the causes are right there in front of them, is preserve another individual’s apparently pointless action. The child does. And the child behaves as though the unexplained step might carry information it can’t yet see — perhaps it makes the box work, or perhaps it is simply how we open boxes. That second reading matters as much as the first. It is the same disposition that later absorbs a table manner, a ritual, a rule of grammar, a line of law: copy the form now, even the parts whose reason is invisible, on the bet that the reason exists — causal or social — and will surface later. High-fidelity copying of other people, faithful enough to carry the parts you don’t understand, is the thing humans do and chimpanzees mostly don’t.

If you think copying the steps you don’t understand is a childish failure we grow out of, I have some news about the most sophisticated manufacturing on the planet. Intel builds chips by a doctrine it named — capital letters, underline, exclamation point, they were not subtle — Copy EXACTLY! When Intel proves out a new process in its development fab and then builds a high-volume factory to run it, the mandate is to replicate the original down to the finest detail: the same equipment, suppliers, plumbing, clean-room conditions, and training methods. And the copying goes absurdly deep — engineers have duplicated variables as fine as lighting, air humidity, and even the maker and color of the paint on the walls, because paint can outgas contaminants that affect the silicon. Not because anyone is superstitious about wall color. Because modern chip fabrication is so complex that nobody can fully model which variables matter in advance, and the era when engineers “improved” the process on transfer was the era of mysterious yield collapses and ruined ramps. So Intel banned improvement. Copy everything, including the parts you’re sure are irrelevant, because you cannot reliably tell the load-bearing step from the ceremonial one — and the cost of guessing wrong is a dead fab. That is the puzzle-box child, in a clean room, wearing a bunny suit. It is over-imitation promoted to corporate law, by some of the most capable engineers alive, precisely because it works.

Banning improvement sounds like the opposite of engineering, but it is the correct response to a system too complex to predict. When you cannot compute in advance which variable is load-bearing, the safe move is to copy the entire known-good answer and let reality — not your intuition — be the editor. Fidelity is one of the ways a system stores answers it cannot yet derive. Which is roughly how DNA works too: replicate faithfully, don’t “improve” the working genes, let selection be the editor over generations. Faithful copying is how you keep an answer whose reasons you don’t have.

The memory outside the skull

Why does faithful copying matter so much? Because it is the only way to preserve a piece of knowledge whose reason nobody currently has — and knowledge whose reason no single person holds is exactly the kind that accumulates. This is the mechanism anthropologists call the ratchet. One generation does something; the next copies it faithfully, including the parts it doesn’t understand; occasionally someone adds an improvement; and the next generation copies the improved version, so it never slips back to zero. High-fidelity transmission is the pawl that stops the ratchet unwinding — the piece without which improvements leak away faster than they arrive. It is not the only ingredient — language, teaching, cooperation, sheer population size, and the division of cognitive labor all feed the ratchet too, and cultures can climb by more than rote copying. But fidelity is the load-bearing one, because everything else is wasted if each generation has to rediscover the base. Turn the ratchet enough times and you get a canoe, a cathedral, calculus, the common law, anesthesia, the iPhone, and the hard-won knowledge that one should never assemble IKEA furniture while hungry — none of it invented by any single mind, all of it the compounded work of thousands who never held the whole.

That is what humans did — not “we teach,” because an ant teaches, but that we combined high-fidelity imitation, open-ended teaching, and language into a channel that no longer stores knowledge inside any one head. We store it between us — in words, writing, tools, institutions, the culture itself — where it ratchets across generations and outlives everyone who fed it. Other animals have the beginnings of this; recent work shows chimpanzees socially acquiring skills no individual invented, so the line is not culture-versus-no-culture. It is that we did it at a scale, a fidelity, and an open-endedness nothing else on Earth has come near — and we spend the first two decades of every human life loading a child into the result.

That’s the difference between you and the chimpanzee. Not mainly the brain behind your eyes. The memory outside it — and the odd, faithful, copy-the-parts-you-don’t-yet-understand disposition by which you were plugged into it before you were old enough to ask why.

What the crib was telling you

Now the paradox reads correctly.

Humans did not trade instinct for intelligence. That is the part the flattering theory got backwards. The genome did not stop building the parenting system; you still arrive wired for the job in all the ways that can be wired in — pulled toward infant faces and cries, primed by hormones and touch, equipped to attend, attach, imitate, and care. What evolution did was subtler and stranger. It built a developmental program whose completed form requires a social world — one that expects a crucial part of the answer to arrive from other people, and is therefore unfinished until it does. A human infant is not born speaking Hindi or Hungarian, but nobody calls language “uninstinctive.” It is a structured program that runs on social input. Parenting is the same. What looks like missing instinct is really instinct built to be completed from outside.

And that is exactly why we need the grandmother, the midwife, the class, and — a thousand years ago, and still across most of the world — the whole village. Not as sentiment, and not as emergency tech support for a defective animal. The village is part of the animal’s expected environment — the place where a species that stores knowledge between its members keeps the half of the parenting program a wasp keeps entirely in its genes. Cut a human off from that world and mothering can fail, which is close to what we see in a great ape raised without other apes. The wasp never needed the connection. We are built to assume it.

So, no — you were not born knowing what to do with your baby. An animal like you couldn’t be, because so much of what would finish you was stored outside you, waiting. Your helplessness at the crib is not the opposite of your intelligence. It is intelligence’s peculiar developmental cost — the price of being the animal whose other half lives beyond the skull.

Are you smarter than a chimpanzee? Not because the brain behind your eyes is a bigger general-purpose engine — on the raw problems it barely is. Because that brain came wired to read other minds and to hold language, and those two gifts plug you into an inheritance no chimpanzee has at anything like your scale: dead people can teach you, strangers can correct you, discoveries can pile up for centuries, and you will faithfully preserve a step whose purpose nobody alive can yet explain.

The wasp had it easier. The wasp also cannot read this.

If you liked this, you’ll enjoy my book, The Science of Free Will — which asks, among other things, why we can’t trade with ants (they can detect cancer), what that has to do with the future of AI, and why a deterministic universe still needs a Supreme Court. Available on Amazon or anywhere else books are sold.

Somewhere in America right now, a chief financial officer is deciding whether to build a factory. Her analyst has computed the firm’s cost of equity — the return shareholders demand for tying up their money in something as undignified as a cheap, boring, profitable company. Using the standard model, the one taught in every MBA program on Earth, the answer comes back around five percentage points over the risk-free rate. Using a newer version of the same model — same data, same market, same firm — the answer comes back around seven.

Two percentage points of discount rate is not a rounding error. Compounded over the life of a factory, it is the difference between building it and not building it. Between the jobs existing and not existing. And the two numbers do not differ because anyone made a mistake. They differ because of a modeling choice — which definition of “risk” the analyst subtracted — made years earlier by academics the CFO will never meet, for reasons she will never read.

She thinks she is estimating a number. She is choosing one, and calling the choice an estimate.

Last time, I wrote about a trio of papers showing that the deep structural parameters of finance — demand elasticities, the machinery under the hood — are not measurements at all. I ended by promising a third essay about the strangest paper of the lot: one that set out to defend academic finance and ended up indicting it harder than the demolitions. That paper is Gormsen and Jensen’s “Conditional Risk,” and this is that essay.

The claim, stated up front so you can hold me to it: alpha — outperformance, skill, the thing this entire industry is organized around measuring — is a function of two arguments. Alpha of a strategy, against a benchmark. There is no benchmark-free alpha. Yet the field talks, publishes, and sells as if the function had one argument. (Strictly there are more suppressed arguments still — sample, estimator, information set, costs — but the benchmark is the biggest one hiding in plain sight.) Every fight about whether value works, whether momentum is real, whether your fund manager earns his fee, is partly a fight about what was subtracted before the residual got called skill. And the second argument is not supplied by nature. It has to be chosen, defended, and disclosed. The field is diligent about the first, sporadic about the second, and allergic to the third.

The Sixty-Second Edifice

To see why this matters, you need sixty seconds of history. I will be fair to the edifice — fairer than it deserves — because you cannot appreciate the demolition until you have admired the architecture.

In 1952 Harry Markowitz had the genuinely great idea: do not judge an investment by itself; judge it by what it does to your portfolio. Individual wiggles cancel; co-movement does not. Diversification is the closest thing markets offer to a free lunch, and Markowitz turned that observation into mathematics. In 1964 William Sharpe pushed the logic to its equilibrium conclusion: if everyone does the Markowitz math (and agrees on the inputs, and can borrow freely, and trades without friction), everyone holds the same portfolio of risky assets — the market itself — and exactly one quantity determines any asset’s expected return: beta, its co-movement with that market. Risk you could have diversified away pays you nothing, because why would it? This is the Capital Asset Pricing Model. Nobel Prizes followed — Markowitz and Sharpe shared the 1990 award with Merton Miller — and it deserved them, as theory.

In 1968 Michael Jensen did the operational thing: run the regression. Whatever return the benchmark explains is beta. Whatever is left over — the intercept — gets a Greek letter, and the Greek letter gets a priesthood. Alpha. Every alpha you have ever been quoted descends from that intercept.

Then in 1977 Richard Roll pointed out the crack in the foundation, and the field has been stepping over it for fifty years. The CAPM’s “market portfolio” means everything — every stock, every bond, every house, every farm, every piece of human capital, your uncle’s vintage guitars. It is unobservable. Not hard to observe. Unobservable. So every test of the CAPM ever run substitutes a proxy — in the academy a broad CRSP-style composite, in practice the S&P 500 — and is therefore, Roll proved, a joint test of the CAPM and of the proxy’s adequacy as the market portfolio. A rejection cannot tell you which one failed; an acceptance certifies nothing about the true market portfolio either. The profession read the paper, agreed it was devastating, and went right on running the regressions.

Hold Roll’s point in your hand. This essay is what happens when you refuse to put it down. The benchmark enters twice — once as a model (which risk adjustment?) and once as a proxy (which portfolio stands in for “the market”?). We will now watch both choices fail to be facts, in public, one at a time.

Oh — and whether anyone can actually estimate the inputs to Markowitz’s beautiful machine is a separate question, with an answer so funny it gets its own essay later in this series. Patience.

Act One: The Benchmark That Will Not Hold Still

Here is the most-studied number in empirical finance: the alpha of value — the return to buying cheap companies and shorting expensive ones, beyond what the CAPM says the risk deserves. Thousands of papers. Careers, Nobel arguments, trillion-dollar product categories. One number.

Except it is not one number, because betas move. A company’s market sensitivity is not a constant of nature; it drifts as the firm changes leverage and business mix, and as the market changes composition around it. The original CAPM ignores this — it is unconditional, one beta for all seasons. The conditional CAPM lets beta vary with time. Think of it as a camera: the unconditional model shoots a fifty-year film at a single exposure setting; the conditional model has auto-exposure. Same camera. Different photograph.

You would think the field could settle which photograph to use. It has been trying for thirty years. Watch.

In 2006, Lewellen and Nagel published a paper whose title is its thesis: “The Conditional CAPM Does Not Explain Asset-Pricing Anomalies.” Their argument is a plausibility bound, and it is elegant: for time-varying betas to rescue the CAPM, betas would have to swing with the market premium implausibly hard — several times harder than anything in the data. They measured betas directly, in short windows, no state variables required. Betas do move, meaningfully.[^1] But not in the right way, not remotely enough, and momentum’s conditional alpha stays enormous — on the order of one percent per month. Verdict: conditioning cannot save you. The anomalies are real. (Savor that word while it is still legal. An anomaly is a deviation from normal — and the “normal” here is an efficient market, which is to say the anomaly is itself measured against the very thing that does not exist. Even the field’s vocabulary is benchmarked against a fiction. Onward.)

In 2011, Boguth, Carlson, Fisher, and Simutin found the hole. The prior literature — Lewellen-Nagel’s bound included — had focused on betas co-moving with the market’s premium and largely waved off betas co-moving with the market’s volatility, a channel they showed is plausibly two to ten times larger. They also named a second sin, “overconditioning”: estimating conditional risk using information no investor could have had at the time, which can inflate measured alphas by a factor of two and a half. Run the estimation properly, with instruments, and momentum’s alpha drops twenty to forty percent. Not zero. Not one percent a month either. Verdict: the anomalies are smaller than you think, and the size of your alpha depends on your assumptions about what investors knew and when — which is to say, on your benchmark.

That same year Nagel himself, with Ken Singleton, built the statistically optimal version of conditional-model testing — different machinery, aimed at the era’s fashionable consumption-based models — and rejected those too, by yet another route. Better conditioning machinery did not produce an acquittal. It produced a fourth verdict. This is what careful people look like when the object they are chasing does not exist.

Then in 2024, Gormsen and Jensen arrived with the paper that was supposed to be the defense. Instead of estimating betas — rolling windows, state variables, all the machinery the previous combatants fought over — they built a tradable portfolio designed to harvest the return to conditional risk, both flavors, premium-timing and volatility-timing. To build it they still had to estimate the market’s conditional premium and its conditional variance. They did not eliminate estimation. They relocated it — into the benchmark itself, which is where this essay keeps telling you the bodies are buried. Add the factor to the benchmark and see what happens to the alphas.

What happens is a massacre. Value’s alpha in the full US sample, 1964 to 2022: 4.58 percent per year unconditionally, 3.33 percent conditionally — about a quarter of the most famous number in finance was never alpha at all, just compensation for risk the old camera could not see.[^2] Post-1996, it is not a quarter. It is essentially everything: 2.45 percent unconditionally becomes 0.43 percent conditionally, with a t-statistic of 0.15 — statistically indistinguishable from nothing. In France, in Germany, in Sweden, conditional risk accounts for effectively all of value’s alpha. The investment factor gives up about half. Momentum is more stubborn — only about seven percent of its alpha over the long sample, though closer to a quarter in recent decades — which will matter enormously in a moment. And the cost-of-capital gap for a value firm — our CFO’s five-versus-seven — is this paper’s own worked example.

Now, the reflex reading of this literature is: science progressing. Lewellen-Nagel measured crudely, Boguth et al. refined, Gormsen-Jensen perfected; the estimates are converging on the true alpha. That reading is wrong, and seeing why it is wrong is the whole point of this essay.

Gormsen and Jensen did not measure Lewellen and Nagel’s number more accurately. They computed a different number and gave it the same name. Short-window realized betas (measure beta afresh in each short window and watch it move), instrumented realized betas (predict beta using only what investors could have known at the time), and a tradable timing portfolio (fold the estimation into a benchmark-side factor that harvests what time-varying beta earns) are not three approximations of one quantity. They are three definitions — three operational answers to the question “what do we subtract before we call the remainder skill?” And no, these papers are not one identical experiment run under a shared design — the test assets differ, the samples differ, the information sets differ. That is not a defense of the edifice. It is the indictment: there is no shared design because there is no shared object. Take value, the factor this whole essay follows. Lewellen and Nagel, measuring conditional betas directly, find its alpha essentially untouched — conditioning explains almost none of it. Gormsen and Jensen, folding the conditioning into a tradable portfolio instead, find the alpha of the same strategy in the same market cut by a quarter over the long sample and erased entirely after 1996. Intact, or gone, depending on what you subtract. Nothing is converging, because there is nothing to converge to. “Value’s alpha” is not a property of value. It is a property of the pair — value, and the lamp you chose to hold over it. Move the lamp, move the shadow, and the field has spent thirty years publishing increasingly sophisticated shadow measurements while insisting the shadow belongs to the object alone.

Be careful to take the right lesson. Gormsen and Jensen have not discovered the correct benchmark — crowning their conditional model as truth would just be electing a new emperor, and their own full-sample estimate leaves value 3.33 percent of alpha, so the model that kills value post-1996 certifies it pre-1996. The lesson is one level up: adjust for risk a little differently — defensibly differently — and the most-studied number in finance runs from substantial to zero. A number that behaves like that is not a measurement of the strategy. It is a negotiation between the strategy and the model — and only one of the parties gets named in the abstract.

And let me concede the obvious before a referee does it for me: chosen does not mean arbitrary. Coordinates are chosen; the distance between two points survives the choice. Celsius is a choice; the fever is real. A benchmark can be defended — by mandate, by opportunity set, by liabilities, by what the investor could actually have held instead. A risk model, an index mandate, and a feasible alternative are in fact three different jobs, and part of the industry’s confusion is letting one applicant interview for all three positions at once. The scandal is not that finance chooses benchmarks. The scandal is that finance writes the choice into the regression and then deletes it from the prose.

At this point the practitioner in the back of the room — and reader, I have been that practitioner for thirty years — stands up and says: this is why nobody on a trading desk uses the CAPM. Academics can keep their conditional models. My benchmark is the S&P 500. Beat the index, you have alpha; trail it, you do not. Solid ground.

I have some news about the solid ground.

Act Two: The Benchmark That Is People

First, a confession delivered without irony: I love the S&P 500. I have traded it, been measured against it, and watched it embarrass geniuses for three decades. The people who maintain it do an excellent job, and nothing that follows is a criticism of them. It is a criticism of what the rest of us have built on top of them — because what the rest of us have built is a monument to objectivity resting on judgment calls.

Start with what the S&P 500 is not. It is not the 500 largest American companies. It never has been. It is a portfolio of roughly 500 companies selected by a committee — the U.S. Index Committee at S&P Dow Jones Indices, market professionals who meet in private. Do not take my word for it; take S&P’s. The current methodology document — June 2026 edition, publicly posted — says it in one sentence: “Constituent selection is at the discretion of the Index Committee and is based on the eligibility criteria.” There are criteria — a market-cap floor ($22.7 billion at this writing), liquidity, float, and famously a profitability screen: positive GAAP earnings in the most recent quarter and summed over the trailing four. But clearing the criteria makes a company eligible, not included. The criteria are the velvet rope. The committee is the bouncer.

And the turnstile only checks tickets on the way in. The same document: “eligibility criteria are for addition to an index, not for continued membership. As a result, an index constituent that appears to violate criteria for addition to that index is not deleted unless ongoing conditions warrant an index change.” A company that stops qualifying stays in — unless, in the committee’s judgment, conditions warrant. The gate has rules. Membership has judgment. (Whether it also has privileges is another question entirely.)

If that sounds like an active manager’s charter, you have company in that opinion, and the company is the man who ran the committee. David Blitzer, chairman of the Index Committee for two decades, wrote in 2014 — publicly, cheerfully, on S&P’s own blog — that “there are no rigid or absolute rules for the S&P 500.” His post is a small masterpiece of saying the quiet part out loud. He tells the story of 2008: Lehman collapses, Washington rescues AIG — the original terms handed the government a 79.9 percent equity interest[^3] — and AIG’s public float crashes through the index’s fifty-percent floor. The rulebook said: remove it. Removing one of America’s largest financial companies from the index, that week, with every index fund on Earth forced to sell — Blitzer’s own words — “would have sent the markets tumbling yet again.” So, his words again: “the Index Committee quietly set the 50% float rule aside.”

Read that twice. In the most dangerous week in modern financial history, the world’s most important benchmark looked at its own rulebook and decided the rulebook was wrong. And here is the uncomfortable part: they were probably right. It was good judgment. That is precisely the point. A formula cannot decide its own exceptions. People can, and did, and do.

Tesla, 2020, same lesson in slow motion. The profitability screen kept Tesla out while it grew into one of the largest companies in America. In the summer of 2020 Tesla finally cleared the screen — its fourth straight profitable quarter, which is more than the formal test even demands (a positive latest quarter, and a positive sum over the trailing four). Eligible at last, overqualified even, and enormous. In September, the committee passed it over anyway, adding three much smaller companies while the financial press gaped. In December, Tesla went in — at a valuation well over half a trillion dollars, the largest addition in the index’s history by float-adjusted market capitalization, forcing index funds to execute roughly $73 billion of trades in a single day. Eligibility is necessary. Inclusion is a decision. You could watch the gap between those two words trade, live, for three months.

How big is the discretion, in general? Adriana Robertson, a law professor with an empiricist’s temperament, counted. In her 2015-to-2017 sample, the committee’s own criteria left it roughly six hundred eligible companies for five hundred seats — more than a hundred discretionary substitutions available on a typical day, representing about five percent of the index’s then-value. Call it a trillion dollars, sitting in the space between “the rules” and “the portfolio.” And when she compared the actual index to the obvious mechanical alternative — just buy the 500 largest stocks, no committee — she found that on the average day from 1989 through 2017, roughly 140 of the S&P 500’s members, better than a quarter of the list, were not in the mechanical top 500. The index is not a formula’s output with a committee attached for ceremony. It is a committee’s portfolio with a formula attached for comfort.

Robertson’s conclusion deserves to be quoted in the register of dry understatement law reviews are made for: index investing is “passive in name only” — it is delegated management, where the party you have delegated to is not a fund manager but the index creator. Sit with what that does to the industry’s favorite word. Passive investing does not exist. Passive implementation exists — the fund really does follow the index, mechanically and admirably. Selection-free index construction does not exist. The discretion has been delegated upstream. So there are only two kinds of active management: the kind you pay eighty basis points for, and the kind you pay three basis points for because the stock-picking committee in Manhattan works for the index provider instead of for you. The three-basis-point committee has a better track record. But it is a committee.

Now — and this is where I am honor-bound to argue against the most seductive version of my own case — does that mean the committee’s judgment is why the index is so hard to beat? Blitzer’s post gleefully relays that Bill Miller, the Legg Mason legend who beat the market fifteen years running, used to argue exactly this: the S&P 500’s record proves active management works, because the Index Committee are good active managers. It is a wonderful line. I want it to be true. The honest answer is: not established. Robertson herself measured the return gap — her mechanical index tracks the real one at a correlation of 0.9992 — and pointedly did not test whether the committee’s choices earn risk-adjusted anything. The long-run performance gap between the S&P 500 and the mechanical alternatives — Russell 1000, top-500-by-cap — is basis points per year, not a skill signature.[^4] And most of “hard to beat” needs no genius at all: it is Sharpe’s pitiless arithmetic — within any specified market, the average actively managed dollar earns that market’s return minus fees, as a matter of accounting rather than economics. And note what even the field’s one bulletproof theorem demands before it starts: someone must specify the market. The benchmark enters before the proof begins. The committee does not have to be brilliant for you to lose to it after costs; the average manager loses to the aggregate that managers collectively are, and the index sits close enough to that aggregate to collect the trophy.

And no — before the closet indexers uncork anything — none of this rescues the fund industry from SPIVA’s annual beatings. Losing to a portfolio chosen in advance is still losing; the scoreboard stands. What dies is the scoreboard’s philosophy: the pretense that the thing every manager in America is measured against is a neutral, skill-free fact of nature. It is not a benchmark-free null hypothesis. It is somebody’s portfolio. There is no benchmark-free null. There never was.

Act Three: Turtles

So test it, says the empiricist who has read this far. Fine: the committee is a portfolio manager. Measure the committee’s alpha. I admire the spirit. Now do it. Alpha against what? The mechanical top 500? Says who — why is size the right sort criterion rather than liquidity, or float, or profitability, which the committee itself uses? Russell’s rulebook? Russell’s rules were also written by people, and rewritten by people, most recently after the 2020 reconstitution embarrassments. Whichever yardstick you pick to measure the yardstick, someone picked it, and the regress does not terminate. It is the oldest joke in epistemology: turtles all the way down. Except this time the turtles charge licensing fees.

Do not trust me; trust the chairman. In that same 2014 post, Blitzer ran the regress one loop himself, in public, and you can hear the vertigo: “Of course, the rules in a rules-based index can be re-written (and sometimes are). But who will re-write the rules? Are there rule-makers ready to act or does someone need to form a committee?” There is no bottom, and the man at the bottom said so. Robertson found the farcical limit case: in her end-2016 sample, 81 of 571 U.S. equity ETFs — about one in seven — tracked indices created by the sponsor or an affiliate. The ruler, measuring itself, for a fee.

One exit remains, and the theorists will already be reaching for it: forget indices and models, the true benchmark is the stochastic discount factor — the deep object that prices everything, of which all these CAPMs and committees are mere shadows. Correct in principle! But an abstraction is not a yardstick. To make the SDF operational you must commit to preferences, information sets, test assets — choices of exactly the kind the previous essay watched contaminating every “measurement” of market structure, where the field’s own trilemma formalized the trap for demand elasticities. The exit door is real. It opens back into the same room. This is not three separate scandals. It is one scandal with three organs: the discovered predictors do not forecast (Emperor 1), the structural parameters are not measurements (Emperor 2), and the yardstick is a committee, or a contested model, chosen by somebody, all the way down (this essay). The field built a cathedral of quantitative rigor and put a suggestion box at the altar.

What should a working investor do with this? The same thing I told you the trilemma implies, now with the third leg under it. Every published alpha — every backtest, every factor premium, every fund ranking, every SPIVA table, every “our strategy added 200 basis points” — is a joint bet on a strategy and a yardstick, and the second bet is the one nobody tells you you are making. So when someone quotes you an alpha, the first question is not “how big?” It is “against what?” And the second question is “who chose the what, and what were they optimizing when they chose it?” You will be amazed how many careers, product launches, and Nobel-adjacent reputations cannot survive the second question.

And with that, the Emperor trilogy is complete: no alpha, no identification, no benchmark. The predictions do not predict, the measurements do not measure, and the ruler is a committee. Alpha is relational. The field has always known this in the mathematics; it forgets it in the language; it has built institutions on the omission.

Naturally, therefore — in the grand tradition of Douglas Adams, whose Hitchhiker’s trilogy ran to five volumes — the fourth part of the trilogy arrives shortly. Because one wall of the cathedral is still standing: the factors themselves. And there is a loose end even sharper than that. Momentum was the one predictor that survived Emperor 1’s massacre — and it is also the one whose alpha will not sit still: on the order of a percent a month under Lewellen and Nagel, a fifth to two-fifths smaller once Boguth and coauthors instrument it properly, barely dented over the long sample yet down about a quarter in recent decades under Gormsen and Jensen. Something about momentum is being conserved across all of these lamps, and I do not think it is a factor loading. Next time.

Meanwhile, somewhere in America, the CFO picks a discount rate. She builds the factory, or she does not. Twenty years from now, someone will score that decision — against a benchmark that has not been chosen yet, by people who have not been hired yet, using a model that has not been contested yet. She is not measuring the cost of capital. She is choosing a model and reading a number off it — which is all anyone has ever done. And somewhere in Manhattan, a committee meets quietly, exercises excellent judgment, writes nothing down, and remains the closest thing to an objective standard that the entire quantitative edifice of modern finance has ever produced.

May they never retire.

[^1]: Lewellen and Nagel estimate the standard deviations of conditional betas at roughly 0.25 for value, 0.30 for size, and 0.60 for momentum — real variation, just several times too small in the dimension that would rescue the CAPM. Figures are from the working-paper version; treat magnitudes as approximate.

[^2]: Throughout, the Gormsen-Jensen figures are from the published Journal of Financial Economics version (2024). In their own alternative specifications, conditional risk explains as much as half of value’s unconditional alpha even in the long sample — so the quarter quoted above is, if anything, the conservative reading.

[^3]: Blitzer’s 2014 recollection puts the government’s AIG stake at “over 90%”; it eventually got there in later restructurings, but the September 2008 rescue terms specified a 79.9 percent equity interest. The float breach, and the waiver, are not in dispute.

[^4]: Verified directly: over 2004–2026, monthly total returns on the S&P 500 (via SPY) and the Russell 1000 (via IWB) differ by roughly one basis point per year, at a correlation of 0.998. The widest gap among the big mechanical large-cap alternatives — the CRSP US Large Cap index, via VV — is about 24 basis points per year, and most of that is the funds’ expense-ratio difference, not index construction. Whatever the committee is adding, it is invisible at the yardstick’s own resolution. Robertson’s in-paper measurement agrees: a 0.9992 correlation between the actual index and her mechanical top-500, 1989–2017.

This is the third essay in the Emperor series. Read the first, The Emperor Has No Alpha, and the second, The Emperor Has No Identification.

If you enjoy watching cherished stories audited down to the studs, my book, The Science of Free Will, asks an equally uncomfortable question about an equally cherished story. For more markets work, see Trading; the physics-minded will find its cousins in the Physics series. New here? Start with The Paradox of India and the full India series — and for related country-specific work, there’s Albion, Britain’s institutional decline, first spotted in 1979.

For more on the background behind this work, see my first and second interviews with Titans of Tomorrow, and Part 2 of my Algorithmic Advantage Podcast conversation with Simon M.

On June 2, Tyler Cowen gave a forty-six-page NBER paper a three-word review, and honestly it might be the most Tyler he has ever been. Luckily I was not drinking coffee at that moment else I might have choked on it!

The paper asks whether you can roll government debt forever. Cowen’s post reproduces two paragraphs of the abstract and links the thing — and then, for original commentary, offers exactly three words: “Big if true.”

That’s the whole review. No gloss, no “here’s why this matters” — he trusted you to do the work yourself. So let me do the disrespectful thing and do the work out loud, partly out of affection for the bit and partly because the gap between “big” and “if true” is where the entire argument lives. Tyler left it sitting there on purpose. So did the paper’s author.

The paper is by Stavros Panageas, one of the serious asset-pricing theorists working today; the acknowledgments run through Lars Peter Hansen, Ravi Bansal, Martin Lettau. When someone at that level writes a paper whose corollary is “maybe you can roll government debt forever,” it’s worth slowing down.

Here is the whole thing under the jargon.

Everything turns on a single comparison: is the risk-adjusted growth rate of the economy below the risk-free interest rate, or above it? By “risk-adjusted” I mean growth as the market prices it — the economy’s actual growth rate, marked down by a premium for the fact that growth tends to disappoint exactly when bad times make a dollar most precious. If that risk-adjusted rate is below the risk-free interest rate, the present value of the entire economy is finite — asset prices reflect their dividends, debt has to be serviced by surpluses eventually, and there are no bubbles on assets in positive supply. This is the world textbooks assume. If it’s above, the present value of the economy is infinite (this is just the Gordon-growth fact that discounting a cash flow at a slower rate than it grows gives you an infinite sum), and all of that breaks: debt can then be rolled over forever without ever running a primary surplus, and the rollover need not be a Ponzi scheme — it can be a stable equilibrium.

Beneath a surprisingly large part of the fight between deficit hawks and their opponents lies the sign of that one number. It isn’t the whole war — there’s inflation, real resources, who holds the debt, what it does to capital — but it’s a bigger front than either side tends to admit. And most of the combatants don’t know it’s the front they’re on.

Now, the tempting argument — the one Panageas sets out to kill — goes like this. The present value of the economy must be finite, because look around: there are finitely-priced assets, like land, that don’t depreciate and whose cash flows track aggregate output over the long run. Land has a price. The price is finite. Land’s rents move with the economy. So the economy’s present value must be finite too.

This feels airtight. It feels like an accounting identity. And it is wrong.

This is, I’ll note, exactly the failure mode I keep circling in these posts: someone spots a quantity that looks conserved, stops asking where it actually went, and builds a cathedral on the unexamined intuition. The land argument smuggles in a conservation law that isn’t there. Cointegration — the technical name for “these two things share a long-run trend” — means the two series move together. It does not mean they have the same present value. Panageas shows that two cash-flow streams can be cointegrated, can look perfectly correlated over any long horizon you like, and yet one can have a finite present value while the other’s runs to infinity. Comovement is not a price ceiling.

And there’s a clean way to see that the land argument was never as tight as it looked. Land doesn’t depreciate, but capital does — and a depreciating asset gets discounted at the interest rate plus the depreciation rate, at r + δ rather than r. So you can have a world where capital is finitely valued, its profits growing at rate g but discounted at the faster r + δ, while the aggregate economy it sits inside has infinite present value, those same goods growing at g but discounted at only r. Finite asset, infinite economy, no contradiction. You just had to pick the right asset to see the intuition drip.

So if the high-level theoretical argument doesn’t work, what does? Here is where the paper earns the word “big.”

You measure it. Directly.

The ideal data would be a security that pays off the aggregate endowment itself, so you could read its yield and see whether the market discounts the economy’s growth at a positive or negative rate. Those securities don’t trade in the United States and never have. But — and this is the kind of thing that makes empirical finance occasionally feel like archaeology — France came close. On June 1, 1956, the French government made a single fifteen-year bond issue whose coupon rose with the index of industrial production. The bonds then traded on the Paris exchange through 1971, leaving behind a long series of market prices. Industrial production isn’t GDP, but over the period the two track each other almost exactly, and that price series is enough to back out the quantity we want.

Here is what the price series says. Backed out of it, the implied yield on a claim to aggregate growth is negative — which is the whole ballgame, because a negative growth-strip yield means the risk-adjusted growth rate sits above the risk-free rate. Put in numbers: Panageas estimates the risk-adjusted growth rate at about 3.26 percent against a real risk-free rate of roughly 2.19 percent, a gap of a bit over a point, and he argues that gap is if anything conservative. Note the object that’s negative: it’s the inferred yield on a growth claim, not the ordinary yield-to-maturity on the French bonds themselves. This isn’t a story about a bond that traded at a negative rate. It’s a story about what the bond’s price reveals when you ask it the right question.

Sit with how unusual that kind of evidence is. Most of the r-versus-g literature observes realized r minus g — what the safe rate and growth actually printed — and then argues about what the risk adjustment should be, with that adjustment buried inside a model or inferred sideways from other assets. (Blanchard’s 2019 presidential address does marshal historical evidence on how often safe rates sit below growth; the risk-adjusted piece is the part that stays theoretical.) Panageas found a security whose price carries information about the risk-adjusted growth rate directly. He didn’t calibrate a macroeconomic model; he read the quantity off a market price. He found a thermometer that history happened to leave in a drawer. Now, even a thermometer needs calibrating — his reading comes through a bond-pricing formula, assumptions about volatility and interest rates, an imputed production series, and a nonlinear fit — so this is not quite reading a number off a dial. But it is a different and rarer kind of claim than one more calibrated model, because the market price is doing work the model usually has to do alone.

And here’s the part the abstract buries, which I think is the most interesting thing in the paper. What breaks the “finite” intuition isn’t anything exotic. No unbounded prices of risk, no fat tails, no infinite variance. The thing that does it is nonlinearity — specifically, nonlinear mean reversion in the cointegrating relationship. Keep the adjustment linear and the two strip yields stay equal and the intuition survives. Let the adjustment be nonlinear, even modestly, and the finiteness condition silently fails.

Here is the mechanism in one breath. The two yields differ only if a particular long-run average refuses to settle down, and whether it settles depends on the gap between the two cash flows reverting fast enough to a stable level. The trap is that “fast enough” has two different meanings. The mean reversion can be plenty strong where you observe the gap, so it looks tame and well-behaved in the data — but asset prices don’t live in the real-world measure, they live in the risk-adjusted one, and the same reversion that tames the gap in the data can be too weak to tame it under the measure that actually sets prices. There, the gap wanders off, the long-run average blows up, and the two yields come apart. Linear adjustment forces “tame in the data” and “tame under pricing” to be the same condition. Nonlinear adjustment lets them split. That split is the whole result.

I’m going to write a whole separate post about this, because it’s a cousin of a warning I’ve been sounding from the physics side. The Schrödinger evolution of a quantum state is exactly linear — superpose two solutions and you get another solution. And yet the probabilities you actually observe are quadratic in the amplitudes, the update when a measurement lands is not linear at all, and the effective equations governing macroscopic variables are routinely nonlinear. Linearity at the bottom layer does not buy you linear behavior in the thing you actually see. The finance case rhymes with this rather than repeating it: the economy looks tidy, the long-run relationship looks linear, and then the one term everybody wanted to drop — the nonlinear adjustment — turns out to be carrying the entire result.

The paper’s own history tells you something, too. It has worn three titles. It started as “Of Trees and Beanstalks: A Paradox of Co-integration and the Sharpe Ratio of Growth-Indexed Bonds” — a riddle about an anomaly, trees being the finite things and beanstalks the ones that climb to the sky. Then it became, more soberly, “Growth-Indexed Securities” — a paper about a data source. Now it’s “The Risk-free Rate and the Risk-adjusted Growth Rate” — a paper that has figured out it’s about the price of the whole economy. You are watching a researcher discover, across three filenames, that the strange instrument in his hand is a thermometer for the one quantity nobody could otherwise measure. I have spent thirty years pricing things for a living, and that is exactly how it goes: you sit down to study some peculiar little security and look up to find you’ve measured something fundamental.

The revision history carries a second signal. An earlier version put the comparison between risk-adjusted growth and the interest rate on the borderline of zero — roughly a tie, maybe a hair above. The current version drops the hedge: the risk-adjusted rate is above the interest rate, full stop. Ordinarily a claim getting more categorical between drafts is a reason for suspicion, not applause — bolder prose is not bolder evidence. But here the estimate itself moved in the right direction, and the growth premium came out roughly stable when the sample was split into subperiods, which is the kind of thing that earns a little extra confidence. The wording got bolder because the measurement did.

So what do we do with this?

Less than the excitable will want — and here Panageas is more disciplined than his louder readers will be. The paper is studded with sentences that begin “the paper does not claim.” It does not claim debt can definitely be rolled forever. It does not claim positive-supply assets definitely contain bubbles. It claims something narrower and harder to dismiss: you cannot rule those things out by waving at the finite value of land. The door you thought was bolted is merely closed.

Because here is the catch the optimists skip. An infinite present value today is a statement about a measure under today’s regime. Risk prices move. Interest rates move. The reading taken in France between 1956 and 1971 is not welded to all times and all places. “The risk-adjusted growth rate exceeds the risk-free rate” is not a conservation law; it is a measurement, and measurements are contingent on the regime that produced them. The binding constraint on rolling debt forever is the persistence of that inequality — and whether it persists is an empirical question, not a moral one.

Which is the part I find I cannot let go of, because I have watched the alternative for thirty years. Debt sustainability gets argued as a morality play: thrift is virtue, borrowing is sin, the prudent nation balances its books. But a practitioner who prices long-dated claims for a living develops an allergy to the phrase “obviously finite,” because the long end is precisely where “obvious” dies and the real work begins. The finiteness of the economy’s present value was never a moral fact about national character. It was always a measurement nobody in the room wanted to take, because taking it is hard and moralizing is free.

So: big if true. The precision of the review is the point. Big, because it takes a hidden assumption sitting underneath one of our largest fiscal arguments — that the economy has a finite price — and turns it into a quantity you can, at least in principle, go out and measure. If true, because the measurement comes from one historical instrument, one postwar market, and an extrapolation to infinite horizons that remains contestable. Tyler quarantined all the doubt into two letters and trusted you to feel their weight. Panageas spent forty-six pages doing the same — the whole paper is an exercise in living honestly inside that “if.”

That’s what a good argument looks like. Not one that demands your assent, but one you could actually lose to the evidence. We should want a great many more of them — and if you like watching a cherished story meet a measurement that doesn’t care about it, that happens to be the move I spend a whole book making below.

If you found this worthwhile, you might like my book, The Science of Free Will — which asks an equally uncomfortable question about an equally cherished story.

And if you’re new here: I write a lot about the paradox of India (whole series here), and about Britain’s long institutional decline, which I first saw coming in 1979.

Closing the Endless Frontier

Two weeks ago I published a battle cry: America Remains Great, Halfwits. I meant every word — the dollar is fine, the companies are the best on earth, and the people who have spent forty years shorting this country have spent forty years losing money. This is the sequel, and the sequel is a groan. ARGH, meet UGH — because undoing greatness is precisely what we are, right now, lining up to do. And if we finish the job, the groan gets shorter and dumber and final: the sound a great country makes a half-second after throwing away the one thing it could not replace. Oops.

I defended my dissertation in 1993.

That same autumn, the United States House of Representatives voted to take the most ambitious scientific instrument ever conceived and bury it in a hole in Texas. By a margin of 264 to 159 they refused to finish paying for it. Days later the President signed the bill that made it official. More than twenty kilometers of tunnel had already been bored under the prairie near Waxahachie. Two billion dollars were already spent. Years later the whole complex was sold off for six and a half million — pennies on the dollar — to a man who hauls freight. We dug the most ambitious hole in the history of science, let it flood, and flipped it for the price of a nice house.

The machine was the Superconducting Super Collider. It was designed to collide protons at forty trillion electron volts — three times the energy of the most powerful collider humanity has built before or since. We had it. We were building it. We killed it.

We killed it because Sherwood Boehlert stood on the House floor and called the Super Collider good science the country could not afford — “good science,” as he put it, “but not priority science.” I want you to know who got to make that call. Boehlert was a Republican congressman from upstate New York, and his degree was in public relations. He used to tell the story that when the party bosses put him on the House Science Committee, they looked at his résumé, saw the last science course he had ever taken was high-school chemistry — in which he earned a C — and decided he was “a natural.” He agreed.

And here is the part that should actually frighten you. Boehlert was the good one. He spent his career as one of the most pro-science members in the entire Congress. He fought for the National Science Foundation. He defended climate science against his own party. Eight years after he helped bury the Super Collider he became chairman of the House Science Committee. If the closest thing the United States Congress had to a friend of science was a public-relations major who got a C in chemistry, the most powerful microscope ever designed never had a prayer.

And before you decide Boehlert was the problem, count the rest of the room. Today there is exactly one physicist in the entire United States Congress — Bill Foster, a former Fermilab man — one out of five hundred and thirty-five. There are about twenty medical doctors. There are nearly a hundred and ninety with law degrees. The body that decides whether America keeps its seat at the edge of human knowledge is run, overwhelmingly, by people who have not solved a quadratic equation since high school, if they solved one then. This is not a partisan complaint. The Super Collider was killed by Democrats and Republicans together, holding hands. Anti-science stupidity may be the last genuinely bipartisan institution we have left.

Nobody on that floor said the science was bad. They couldn’t have told you what it was. People who did not have a prayer of understanding the machine voted, together, that the country could not afford to find out what the universe is made of.

And then they built a space station.

And then they built a space station

In the very same autumn they killed the collider as unaffordable — September 1993, weeks before the final vote — the United States and Russia shook hands on the project that became the International Space Station. Final price: about $150 billion. It is one of the most expensive objects human beings have ever built — more than the Large Hadron Collider, more than any aircraft carrier ever launched. Not Apollo money, but a mountain of money all the same. The machine that would have opened the frontier of physics was unaffordable at twelve billion. The space station was just fine at twelve times that.

So ask what the space station was actually for, because it was not science. By the honest reckoning of many scientists, almost all of its research could have been done by robots on the ground for a fraction of the cost. What it produced instead was geopolitics — a way to keep the collapsing Soviet Union’s rocket engineers employed and pointed at orbit instead of at Tehran and Pyongyang — and jobs. Lots and lots of jobs, in lots and lots of congressional districts.

And there, at last, is why one machine lived and the other died, and it has nothing to do with the science, because not one of them understood the science. To keep its costs down, the Super Collider had concentrated its spending in Texas. Efficient, and fatal: it meant the other forty-some states had no contract, no job, no ribbon to cut, and therefore not one reason to lift a finger to save it. The space station was the opposite — deliberately smeared across the entire map, a little money in everybody’s backyard. That is the whole difference. The only language that penetrates a skull that cannot read a blueprint is the language of pork. Not what will this teach us about the cosmos. Not what will our grandchildren build on top of it. Just: what is in it for my district?

Screw the country. Screw the future. What’s my pork?

It’s what’s for dinner.

What “not affordable” bought us

Here is what that bought us. Thirty-three years of silence at the frontier.

We eventually found the Higgs boson — in 2012, in Switzerland, at a smaller European machine that could barely reach the bottom edge of the range the Super Collider was built to explore. CERN announced it on the Fourth of July, which I have always assumed was either an accident or a taunt. Finding it was wonderful, and it was not a surprise. We were all but certain it was there; the theory had demanded something like it since 1964. The entire point of building a machine three times bigger was to see what lay past it. Supersymmetry. New particles. Whatever the universe is actually made of. We never looked. The frontier from fourteen to forty trillion electron volts is dark, and it will stay dark for the rest of my life.

My own thesis committee included Steven Weinberg, who won the Nobel Prize for the theory the Higgs completed. He fought for the collider to the end, and years after it was gone he warned that the search for the laws of nature itself might “slow to a halt, not to be resumed again in our lifetimes.” He was right. He was almost always right. He was ignored.

Particle physics pays out in decades, not quarters. So here is the honest question, thirty-three years on: how much do we simply not have? How many discoveries, how many technologies, how many entire industries do not exist because we filled in the tunnel? You cannot put a number on the thing you chose not to find. That is precisely why it was so easy to cancel. The bill never comes due on the watch of the person who refused to pay it.

The overruns are not a defense. They are the confession.

Now let me grant the other side its best card, because it had one. The Super Collider did run over budget. Badly. The early estimate was around four billion dollars and it climbed past ten. The magnets had to be redesigned. The accounting was a disaster. The Inspector General found spending nobody could document, and twenty-one thousand dollars of office plants. All true. So let me make the cost-overrun case for the people who killed it — and then let me show you why it convicts them.

First: of course it ran over budget. You are building the largest and most complicated machine in the history of the species, a thing that has never existed, at the absolute edge of what physics and engineering can do. Frontier science overruns the way water runs downhill. And this is not my opinion — it is the sworn position of the United States Congress. Eighteen years after they buried the collider, a House committee moved to kill the James Webb Space Telescope for the identical sin, calling it “billions of dollars over budget and plagued by poor management,” and in the same report conceded that significant cost overruns are commonplace on exactly this kind of project. They know. They have always known. Overruns are the toll the frontier charges, and they indicted the frontier for paying it.

Second — and this is the part that should put your fist through the drywall — when you find out that a once-in-a-generation instrument is over budget and badly run, there is an obvious thing to do, and it is not to set fire to two billion dollars and bury the future under a field in Texas. You fix the management. You bring in people who have actually run megaprojects. You descope. You slow it down. Or you do the thing the entire rest of the developed world does as a matter of routine, and that the United States, in its provincial arrogance, would not do: you bring in international partners to split the bill. A frontier collider is not unaffordable. CERN is the living proof — funded by some twenty nations, running the very machine that caught our Higgs, a machine that also ran over budget and broke down for the better part of a year, and not one person in Geneva suggested flooding the tunnel. They fixed it, because the people in charge could read the blueprint.

We know every word of this is true, because we ran the experiment a second time and passed. The James Webb telescope overran its early estimates by more than an order of magnitude. A House committee voted to zero it out. And then thirty-two Nobel laureates and one stubborn senator from Maryland fought to save it, and won, and the cancellation was reversed. The physicist Lawrence Krauss warned, at the time and in print, that killing Webb would be killing the Super Collider all over again. Webb launched on Christmas Day, 2021. It is now the most productive infrared observatory ever built, with astronomers begging for nine years of its time for every single year it has to give. Had that 2011 vote held, we would have pocketed a few billion dollars and erased every image, every discovery, every paper. That is exactly the trade we accepted in 1993.

So no — the overruns are not a defense. They are the confession. The men who killed the Super Collider were handed a problem with a dozen solutions, every one of which would have saved both the science and the two billion dollars already sunk into the ground. They reached past all of them for the only lever a man who cannot read a blueprint knows how to pull. They didn’t fix it. They couldn’t have told you what fixing it looked like. So they killed it — and they called it fiscal responsibility.

Three professions

I should confess a bias here, and then explain why it isn’t one.

Three professions — as professions, not the many fine people inside them, but the classes — have always seemed to me to produce the highest ratio of self-regard to foresight in modern life: politicians, lawyers, and MBAs. The Super Collider is where I learned exactly why, because the Super Collider is where all three of them met, shook hands, and walked away from the most ambitious hole in the history of science.

And the first thing to understand about this class is that it feels no shame. We run two completely different standards of embarrassment in this country, and almost nobody notices. Let a powerful adult say “Shakespeare — who’s that?” and we know instantly what we are dealing with: an ignoramus, a man to keep away from anything that matters. But let that same adult say “the photoelectric effect? Never heard of it,” and we simply nod. He’s “not a science person.” No shame attaches. None. He can chair the committee, run the company, govern the country, while being unable to give you one sentence on the physics that turns the sunlight on his own roof into the current in his walls — and he will feel no more embarrassment than if you had asked him to name a star.

And it is worse than that, because the photoelectric effect is not some obscure footnote. It is the thing Albert Einstein won the Nobel Prize for — not relativity, which the committee thought too speculative, but the photoelectric effect, in 1921. The most famous scientist who ever lived, the man whose face simply means genius, won his Nobel for the discovery that light arrives in packets that can knock electrons loose — the same quantum physics every solar cell on Earth runs on. Offer that one fact to the people we have put in charge of science and most will look at you the way you’d look at a baseball statistic from a sport you don’t follow. A bridge too far. Not “they haven’t read the journals” — they cannot tell you what Einstein got the prize for, and it has genuinely never crossed their minds that they should be able to. That is the level of idiocy we are discussing.

And these are the people who now propose to sit in judgment of science itself. Watch how that goes. The same official who cannot tell you what Einstein won his prize for will look you in the eye and announce that the cell tower at the end of the block is giving the neighborhood cancer — never once suspecting that the photoelectric effect he skipped is the exact reason it cannot. Here is the whole of it, the thing Einstein worked out in 1905: it is the frequency of radiation, not its intensity, that decides whether a single photon carries enough energy to break a chemical bond in your DNA. Below a threshold the photons are simply too weak, and you can pour on intensity until the cows come home without changing the energy any one of them delivers. Cell-phone and tower radiation sits hundreds of thousands of times below that threshold. You could bathe in what comes off a tower; the worst it could do is warm you by a fraction of a degree.

And the evidence has long since caught up to the physics, as it always eventually does. The “possibly carcinogenic” label the worriers love to wave around comes from a 2011 review that leaned on shaky questionnaires, and it files cell phones in the same drawer as pickled vegetables and aloe vera extract. The most comprehensive assessment since — a systematic review commissioned by the World Health Organization in 2024 — found no association between wireless radiation and cancers of the brain or head. And after twenty years in which nearly every human on earth has pressed a transmitter to the side of their skull for hours a day, the rate of brain cancer has not so much as twitched. There is no mechanism. Now there is no signal either. Cell phones cause cancer my ass.

My own town has no cell towers, because a zoning board of confident, science-free citizens decided they do. Try explaining E = hf to a zoning board. And there, scaled up to the size of a country, is the whole danger in a single image: we are now handing the keys to the national scientific enterprise to precisely that zoning board.

And yet the deepest indictment is not even this ignorance, and it is not that these people are stupid — most are clever, some are brilliant. It is structural, which is worse, because stupidity is at least randomly distributed. Each of these three professions is trained, selected, and lavishly rewarded to win the room it is standing in at this moment. The politician optimizes for the next election — two years, six at the outside. The lawyer optimizes for the verdict, the case on the desk today. The MBA optimizes for the quarter, the number due in ninety days. Not one of their incentive structures can so much as perceive a payoff that arrives in the year 2050. They are not short-sighted because their eyes are weak. They are short-sighted because short sight is precisely what their professions reward, and long sight is precisely what those professions punish.

Now look at what basic science is. It is the longest-horizon, least-legible, least-monetizable thing human beings do. The payoff comes in decades, in a form nobody can name in advance, often in a field nobody expected. It is, in plain terms, the single worst possible match for a politician, a lawyer, or an MBA — and we have handed all three the keys. It is like handing a lifelong sprinter a fistful of acorns and ordering him to bring you a forest by Friday. He will run himself into the ground. He will produce nothing — you cannot sprint an oak into existence — and he will not understand why. Then, with enormous confidence, he will explain that the forest was a waste of money in the first place. That is not a slander on sprinters. It is a slander on whoever put the sprinter in charge of the forest.

You have seen this before, outside of Washington entirely. Xerox built a research lab in Palo Alto, filled it with the best computer scientists alive, and let them build the future: the graphical screen you are reading this on, the mouse next to your hand, the windows, the desktop, the Ethernet wired into your office, the word processor that shows you the page before you print it. PARC assembled essentially all of modern personal computing into working machines. And Xerox — run by men who could recognize a faster photocopier and almost nothing else — captured nearly none of it.

And the man who built PARC and then presided over its slow bleed — chairman C. Peter McColough — was a Harvard MBA who had studied law and climbed the company through sales. His copier men took one look at the Alto, the first real personal computer, saw that it cost twelve thousand dollars to build, and could not imagine who would ever pay that for what looked to them like an expensive typewriter. The definitive history of the debacle gives that decision its own chapter, titled — with no irony available to the men in it — “Finance: The Rejection of the Alto.”

Let me not get personal. Oh, fuck it — let me. The man was the precise dictionary illustration of the dumb fuck out over his skis — someone who, in a properly ordered universe, would not have been trusted to run an Arby’s. Although, in fairness, Arby’s are generally run pretty well. So strike that; it’s unkind to Arby’s. A cashier, then. Somewhere with a register that beeps when the change comes out right, so he’d have a fighting chance of knowing.

Now — was he a conventional dumb fuck? Probably not. An MBA is mostly a license to bullshit with conviction, but a law degree takes some real wattage. That is exactly the trap, and it is why he is worse than a simple fool: the wattage a law degree certifies is the wattage to write a brief and work a room — to win the argument in front of you and charm the table at dinner. None of that is judgment. None of it is the capacity to look at a working future and recognize it. He carried two of the three cursed credentials stacked one on the other, doubly trained and doubly certain, and a man built like that — fluent in everything except the one thing on the table — will turn down the future every time and feel like the smartest man in the room doing it.

The one thing they did commercialize was the laser printer, because it looked like their existing business. Everything that didn’t fit inside their small model of the world walked straight out the door, to Steve Jobs, to the founders of Adobe, to anyone who could see. The researchers told the executives exactly what they were sitting on. The executives could not hear it.

That is the United States and its scientists right now. We are Xerox. We have the lab. We always had the lab. We have handed the keys to the men who can only see the photocopier.

Where are the customers’ yachts?

Let me make this personal, because I have skin in precisely this game.

I have spent more than thirty years in money management. I have dear friends in finance, some of them with MBAs, who love me and worry about me, and who cannot for the life of them understand three things about me.

The first is that I turned down the jobs. Real jobs, at famous funds, with numbers attached that I won’t print, because you’d stop reading the argument and start doing the math. “They offered you what, and you said no?” Yes. I said no.

The second is that I still do physics. Actual physics — I publish real papers, in real journals, the kind almost nobody reads and nobody pays for. “Isn’t that a waste of your time?” To them it’s self-evidently mad: time spent on a thing that returns no money is time set on fire.

The third is that I run my own firm without ever folding myself into any of the tidy little boxes the big allocators want you to sit in, and I do no sales and no marketing. None. Zero. This, I’m told, is why the business grows slowly. They’re right. I don’t care.

Here is the thing they keep waiting for me to feel — the shame, the regret — and it never shows up, because I’m running a different program than they are. My standard is a single question. Is it true? If the answer is yes, the thing is worth doing, and the money is a side effect or it isn’t. If the answer is no — if the only reason to do it is that it pays — then, with great respect and real affection: fuck you, and fuck the thing. And once you hold to that one standard, you start to see the industry I work in a little too clearly. Because here is what almost nobody will say out loud: the giant investment banks and the giant fund managers are not, in the main, in the business of making you money. They are in the business of making themselves money, using yours. They are sales-and-marketing machines with a trading desk bolted on. The product is the fee. You are the raw material.

And who trains the people who build those machines? The same well from which we draw the consultants who, when a drug company asks how to grow, sit down and engineer the answer. I want to be specific, because the specific is so much worse than the general. McKinsey — the single most prestigious destination on earth for a freshly minted MBA — was paid by Purdue Pharma to “turbocharge” the sales of OxyContin. That is their word. Turbocharge. While Americans died by the tens of thousands. They counseled pushing the higher, more lucrative doses. They strategized, in writing, about how to “counter the emotional messages from mothers with teenagers that overdosed” on the very product they were turbocharging. When it began to fall apart, a partner started deleting the emails. McKinsey eventually paid five hundred and seventy-three million dollars to settle with the states, and then another six hundred and fifty million to the federal government — which is roughly what it costs to make grieving mothers go away when you have that kind of money. This is not a flaw in the training. Maximize the client’s revenue, externalize the corpses, bill accordingly — that is the training.

In 1940, a man named Fred Schwed wrote a book about Wall Street under what may be the most honest title ever set in type: Where Are the Customers’ Yachts? A visitor is brought down to the harbor and shown the beautiful yachts of the bankers and the brokers. He looks around and asks the only question that matters: where are the customers’ yachts?There were none then. There are none now. So let me ask you, since you might keep an account somewhere, since you might be trusting some of these people with your retirement: where is your yacht? And while we’re here — how many doctors do you know who bought a boat with the gains some genius in a good suit earned for them? I’ll wait.

If you think I’m being unfair to my own industry, don’t take my word for any of it. Take Michael Lewis’s. Read Boomerang — it’s short, it’s genuinely, darkly funny, and it’s the clearest field guide I know to how this machine actually behaves once it’s loose on a continent. Watch what happens to Germany: the great German banks, trusting and rule-bound, certain that if every box was checked nothing could go wrong, became the world’s designated buyers for the toxic American mortgage paper Wall Street needed to get off its books. Lewis’s line is that the entire eurozone was, in effect, a CDO sold to the Germans. They checked every box exactly as their allocators demanded — and that is precisely how they got fleeced. Then watch Goldman Sachs walk into Greece in 2001 and run what Lewis calls “repellent deals designed to hide the Greek government’s true level of indebtedness” — the same laundering machine that had packaged the American subprime borrower, now aimed at an entire country so it could lie its way into the euro and detonate a decade later. The Greeks did most of the damage to themselves. Goldman’s job was simply to help them hide how broke they were until the door had closed behind them.

I tell you all of this not to settle scores with my own trade, but because it is the same disease in a different suit. The friends who can’t fathom why I do physics, the men who buried the Super Collider, and the machine that sells you funds you don’t need are all committing the identical error. They cannot see value that doesn’t arrive as a number, this quarter, with their name beside it. Truth that pays slowly, or never, is invisible to them — and basic science is the slowest-paying, least-legible truth we have. That is the entire story of what we are about to do to ourselves.

The bull case

Now let me cool down and be fair, because the case against panic really is strong, and I believe most of it.

Two weeks ago, in this space, I told the people forecasting America’s collapse to go home. I meant it then and I mean it now. Michael Cembalest, who runs market strategy for J.P. Morgan and is about as far from a flag-waving optimist as a careful man gets, has just published forty pages of charts — an Eye on the Market special edition he calls Semiquincententacles — that make my Fourth of July argument better than I made it. The “Sell America” trade everyone swore would punish us has lost money every single time it’s been tried. The dollar ignored the doomsayers. In nearly every major sector, American companies earn higher returns on equity than their European, Japanese, and Chinese rivals — across the board, not on the strength of one lucky technology bet. American productivity leads the entire G10. American capital markets are so dominant that the rest of the rich world is reduced to writing reports about why it can’t keep up. All true. All real. I will defend every word of it.

But notice what every one of those numbers is standing on, and then notice where Cembalest himself goes. After two dozen pages on why America is fine, the most cautious man in the building turns to the two risks that have nothing to do with the business cycle: first the unpredictability of the rule of law, and then, by page thirty, an entire section on the defunding of science. I will leave the rule of law for another day. The other one is the subject of this entire essay, and we are no longer talking about one collider in 1993. We are talking about the same mistake, made all at once, to the whole machine.

So let me show you the physics of it.

Talent is conserved

Here is a conservation law they don’t teach in business school. You cannot destroy scientific capacity by defunding it. It does not vanish. It relocates. Cut the grant, close the lab, and the knowledge in the heads of the people you just fired does not evaporate — it gets up, packs a bag, and goes to work for whoever is still paying. Talent is conserved. The only question a funding decision actually answers is whose talent it will be.

We know this is true with more certainty than we know almost anything in policy, because we have already run the experiment — in reverse, and in our favor, at the largest scale in modern history. In the 1930s a madman in Germany decided that some of the finest minds his country had ever produced were the wrong kind of people, and he threw them out. And we caught them. Einstein. Fermi, by way of fascist Italy. Von Neumann. Bethe. Szilard. Teller. Wigner. Born. We did not train a single one of them; we did not spend a dollar raising them; another country paid every cost of their education and then handed them to us out of pure stupidity and hate. The bomb, the architecture of the computer, the entire theoretical spine of the American century — built, to a degree Americans have largely forgotten, by the minds Europe drove out. We did not earn the golden age of American science so much as inherit it, and then we lived off the inheritance for three full generations.

Vannevar Bush gave that inheritance an institutional home. In 1945 he wrote a report for the President called Science, The Endless Frontier, and out of it came the National Science Foundation, the modern research university, the social contract under which the government funds the basic science that no company can justify and from which, eventually, every company profits. The NIH would go on to underwrite the research behind essentially every new drug approved in this country. The NSF would seed the internet, supercomputing, the foundations of artificial intelligence itself. The dominance Cembalest documents — the productivity, the earnings, the markets, the AI — is the dividend on that frontier. It is money invested between roughly 1945 and 2015, arriving now.

And the conservation law runs in both directions. That is the part nobody wants to say out loud. The same physics that filled our laboratories in 1940 will empty them just as faithfully if we decide, as we now seem to have decided, to play the role Germany played. We are no longer the country that catches the talent other nations throw away. We are becoming the country that does the throwing.

Look at Cembalest’s own pages — the meter reading on the very machine he is praising. American scientists submitted thirty-two percent more applications for jobs abroad in 2025 than the year before. Eighty-five rising and established researchers, including an NIH neurobiologist and a Princeton nuclear physicist, have already left for China. Enrollment in the PhD programs that produce our biomedical researchers, after years of growth, has flatlined. The European Union has stood up a half-billion-euro fund whose entire purpose is to buy our scientists away from us, and Europe does not spend half a billion euros on things it does not expect to work. And it is not only China and Europe: as I write this, Chen Wang — who builds the kind of superconducting circuits that may one day become quantum computers — is leaving the University of Massachusetts for the University of Toronto. Once that caliber of talent goes, it does not come back. The seed corn is not burning. It is boarding flights.

And it is boarding flights to precisely the competitor Cembalest also documents climbing the ladder behind us — China breaking into the ten most innovative economies, Chinese GPU self-sufficiency rising from ten percent to forty and aimed at eighty, Chinese frontier models now a handful of points behind our best. This is the detail that should end the debate. When you defund American science, you are not slowing the world down. You are slowing yourself down while handing your only serious rival the one input it cannot easily manufacture: trained human minds at the frontier. Germany’s madman handed us the future. We are now handing it back, across the Pacific, on purpose, and calling it a budget cut.

We are doing it again, and we are doing it bigger

The grant data tells the story plainly: sharp declines in competitive NIH funding for cancer, for stroke, for diabetes, for heart disease, for infectious disease; sharp declines in NSF funding for biology, for engineering, for computer science, for the physical sciences. By Cembalest’s own estimate, roughly six thousand NIH grants were terminated, about half for running afoul of political directives and the rest for, as far as anyone can tell, nothing at all. The administration fired the entire National Science Board — the body that governs the NSF — days before the board’s own Science and Engineering Indicators report, the one documenting how fast China is closing the gap on exactly the metrics that decide the next century: papers in the top one percent of the most-cited journals, total spending on research and development, and patents in artificial intelligence, quantum information, biotechnology, semiconductors, and nuclear technology.

And then there is the part that should make a market person sit up, because it is structural and it is permanent. The Office of Management and Budget has proposed a rule — one that would bind every federal agency by October of this year — that would dismantle the three pillars the entire postwar system rested on. Political appointees, not scientists, would conduct a “pre-issuance review” of every discretionary grant, explicitly forbidden from deferring to the experts, instructed to block anything that promotes “anti-American values.” Peer review — the mechanism by which independent scientists, and not commissars, decide what is good science — would become advisory, overrideable by a political appointee with no finding of cause. Active multi-year grants could be terminated mid-stream for being “inconsistent with agency priorities,” no fraud or failure required. Understand what this is in market terms. Peer review and open competition are the price mechanism of science — the closest thing the field has to a market that sorts good bets from bad. Replacing that mechanism with the political preferences of whoever holds power this year is not reform. It is central planning, imposed on the one corner of American life that beat the centrally planned competition precisely because it was free.

Shirley Tilghman — molecular biologist, former president of Princeton — put her finger on the part a builder should fear most. The intention, she allowed, was never to destroy American scientific capacity; it was to punish a handful of elite universities. But intention is not outcome, and the outcome is the destruction of the capacity itself. It is one thing to tear something down. It is another to tear it down with nothing ready to put in its place. That is the moment we are in.

Albion

I keep having the feeling that I have seen this film before. And I have. The country was Britain, and I first saw the rot in 1979.

Empires do not die on the ticker. Britain’s didn’t. The pound stayed convertible, the City stayed busy, the great names kept their offices long after the thing underneath had gone. An empire — and a scientific civilization is a kind of empire — dies first in the institutions and the laboratories, invisibly, in a way that shows up in no quarterly print, and the markets find out last. By the time the decline is legible in the numbers everyone watches, it is already a generation old and substantially irreversible. Britain led the world into the industrial age and the world into the foundations of modern physics and biology, and then it spent decades quietly underfunding and out-emigrating its own talent, and one day it looked up and the center of gravity had moved across an ocean — to us, partly on the backs of the scientists it failed to keep.

We are now running the British playbook on the one asset Britain would have sold its fleet to hold. The difference is that we are doing it faster, louder, and with a written rule that puts political officers over the peer reviewers. The dollar will be the last thing to tell us. It always is.

The objections, and why they fail

Let me take the strongest versions of the case against me, because there are three, and they deserve real answers.

“We are not gutting science. We are cutting bloat — the captured grants, the ideological projects, the waste.” Some of that is fair; some grants are junk, as some grants have always been junk. But the administration’s own numbers refute the scale of the claim: of the terminated NIH grants, only about half tripped the political filters at all, and the other half were killed for no stated reason that survives scrutiny. And whatever you think of trimming fat, the OMB rule does not trim fat. It puts political appointees above the scientific community’s judgment and lets any live grant die for “agency priorities.” You do not make an engine more efficient by firing the engineers and handing the controls to people who think the manual is propaganda. The market test for science already exists. It is called peer review. They are abolishing it.

“The private sector will fund it now — the AI labs, the pharma giants, the great corporate research budgets.” They will fund the applied layer that sits on top of the public base; they always have. What they cannot fund is the base. Cembalest’s own pages show the frontier AI labs incinerating cash, their own projections of when they will turn cash-flow positive sliding into 2028, 2029, 2030, which he rightly calls speculative. You cannot ask companies that must be profitable by 2028 to carry the twenty-year, no-revenue, who-knows-if-it-works research from which the profitable thing will eventually grow — research whose entire defining feature is that it has no buyer yet. The reason the government funds basic science is that it is the one bet rational on a civilizational horizon and irrational on every private one. Remove the public bet and the private bets have nothing left to stand on. It just takes twenty years to notice, which is exactly the interval these people are built to ignore.

“China will trip over its own government before it overtakes us. The Party will smother its own scientists.” Maybe. I would not bet against the possibility. But notice the shape of the argument: it asks you to stake the scientific future of the United States on the proposition that your rival’s government is more self-destructive than your own. And your own government has just fired the entire board that oversees its science agency, days before that board could warn it about the rival. When your single remaining hope is that the other side is even dumber than you are, you have already lost the thread. Cembalest’s charts show China closing, not collapsing. Plan for the China that climbs.

Where is your yacht

So let me end where I began two weeks ago. America remains great. Say it again, because it is true: the dollar is resilient, the companies are the best on earth, the doomsayers shorting this country have lost money for forty years and will keep losing it. If your plan is to bet against America on the dollar or the debt ceiling or a bad inflation print, you will be carried out, as you should be.

But there is exactly one way to make the doomsayers right, and it is not a tariff or a yield or a CPI surprise. It is to take the machine that produced every bit of the greatness Cembalest documents — the engine Vannevar Bush switched on in 1945, the one we built out of the genius another madman gave us — and hand the controls to people who think peer review is the enemy and a fleeing physicist is somebody else’s problem. To close the endless frontier, fill in the tunnel, and call it responsible. Elias Zerhouni, who ran the NIH under a Republican president, put it as plainly as it can be put: they are taking an axe to the system, they are destroying it, and they do not understand that when you tell them.

You want to short America? Here is how you actually do it. Don’t touch the dollar. Keep defunding the thing the dollar is built on. Drive the scientists to Shenzhen and Munich. Put the commissars over the peer reviewers. Then wait twenty years, and collect.

I would simply rather we didn’t hand them the trade.

We’ve buried one collider this way already. The tunnel is still out there under Texas, full of water, and nobody remembers it, and that is the whole problem: nobody ever remembers the discovery that was never made, the cure that never came, the industry that never started, the yacht the customer never got. The cost of closing the frontier is invisible by construction. So is the rot in a great house, right up until the roof comes down. Ask Britain. Or don’t — ask the one physicist we have left in Congress. He’s easy to find. There’s only one of him.

For the optimistic case, see the companion piece, ARGH! America Remains Great, Halfwits. On Britain’s long institutional decline — first seen in 1979 — see Albion. And for an equally uncomfortable question about an equally cherished story, my book, The Science of Free Will.

I wrote recently about a paper that should have embarrassed academic finance more than it did.

The finding was simple and brutal. Two hundred and twelve peer-reviewed stock-return predictors — the output of a five-year review gauntlet at the field’s most prestigious journals — predict returns out of sample no better than a computer mindlessly grinding through 29,000 accounting ratios in search of a t-statistic above two. Decades of theory. Decades of careful storytelling about risk and mispricing. And it bought you nothing a brute-force search wouldn’t have found anyway.

That paper was about prediction. Academic finance, it turned out, fails the prediction test.

I want to tell you now about a paper that does something worse.

It does not go after the predictive wing of the discipline. It goes after the structural wing — the half that doesn’t try to forecast next month’s return but claims, far more ambitiously, to measure the deep parameters of investor behavior. The demand curves. The elasticities. The numbers that tell you how much prices move when a central bank buys bonds or an index fund rebalances.

And it proves — from two assumptions that essentially no one in finance is willing to give up — that those measurements cannot be made model-free from the holdings-and-prices data they are usually advertised as coming from. Not “are hard to make.” Not “have wide error bars.” They are not measurements at all, unless you first smuggle in the very theory they were supposed to test.

The paper is “A Trilemma for Asset Demand Estimation,” by William Fuchs, Satoshi Fukuda, and Daniel Neuhann. The title gives the structure away. A trilemma is a menu of three items from which you may choose at most two. The three here are: no-arbitrage pricing; the assumption that investors value assets for their payoffs; and the ability to identify structural asset demand from holdings-and-prices data without first committing to a full model. Finance wants all three. It cannot have them. Something has to go — and if no-arbitrage and payoff-based preferences stay, what goes is model-free measurement.

A word on the authors. Two of the three — Fuchs and Neuhann — sit at the McCombs School at the University of Texas at Austin. That happens to be where I earned my PhD, under E.C.G. Sudarshan, with Steven Weinberg on my dissertation committee. The University of Texas mascot is a Longhorn, and its supporters salute it with a hand sign and a cheer: Hook ‘em horns.

What Fuchs, Fukuda, and Neuhann have just done to a research program in their own field is, in precise economic terminology, exactly that.

They put the horns through it.

The thing the paper kills

To see why this matters, you need to know what “asset demand estimation” is, and why a large and serious slice of modern finance has staked its credibility on it.

The idea traces to a 2019 paper by Ralph Koijen and Motohiro Yogo, “A Demand System Approach to Asset Pricing,” published in the Journal of Political Economy. It has been enormously influential. It spawned an entire literature, a small industry of follow-on work, and a vocabulary that has migrated out of the journals and into central banks.

The pitch is appealing. Treat investors the way an industrial-organization economist treats shoppers in a supermarket. Each investor has a “demand curve” for each asset. Estimate those curves from data on who holds what at what price. Out pops a number — the demand elasticity — that tells you how sensitive holdings are to price.

Why would anyone outside a finance department care about an elasticity?

Because the elasticities are load-bearing. They are the inputs to claims you have absolutely heard, whether or not you knew where they came from:

Quantitative easing compressed long-term yields by such-and-such basis points. Adding a stock to the S&P 500 raises its price by some measurable percentage. Markets are “inelastic” — a dollar of net inflow pushes aggregate value up by roughly five dollars — and therefore the rise of passive investing has mechanically inflated stock prices.

Every one of those sentences is an elasticity wearing a coat. Central banks lean on these numbers. The entire active-versus-passive shouting match leans on these numbers. They are not academic trivia. They move policy, and they move capital.

The Fuchs-Fukuda-Neuhann result is that the numbers cannot be identified, model-free, from the data they are supposedly estimated from. Whatever makes the number a measurement of structure, you carried in.

Let me show you why. It is not complicated — and, pleasingly, the argument is essentially a piece of physics.

The dictionary in the middle

Here is the entire engine of the paper, compressed into a single line. The matrix of asset demand slopes — call it A, the object the econometrician wants — can be written as:

A = (Y⁺)ᵀ D Y⁺

Three pieces. Take them one at a time.

D is fundamental demand. It is what the investor actually cares about. And what an investor cares about, in the standard asset-pricing setup, is not “shares of Apple” — nobody wants Apple stock for its own sake. What they want is consumption: money in good states of the world and, especially, money in bad ones. D describes an investor’s appetite for state-by-state consumption. It is the genuine object of preference.

Y is the payoff matrix. It records what each asset delivers in each possible future state of the world.

Y⁺ is the part that does the damage. It is the Moore-Penrose pseudo-inverse of the payoff matrix — and for our purposes you can think of it as a dictionary. When a matrix is square and well-behaved it has a true inverse, a clean way to run it backwards. Most payoff matrices are neither. The pseudo-inverse is the mathematician’s best available substitute: the closest thing to a reverse gear when no exact one exists. Here, that reverse gear translates the language investors think in — “I want this pattern of consumption across future states” — into the language markets transact in: “therefore hold this particular portfolio.”

Now look at the structure of the problem. Investors have preferences over consumption — that is D. The econometrician observes holdings and prices, and wants to infer A, the slope of asset demand. Sitting between the object investors care about and the object the econometrician wants is the dictionary, Y⁺.

To recover preferences from observed holdings — which is the entire point of the exercise — you need the dictionary.

And here is the paper’s first blow: you cannot have it. Not from the data alone.

Why the dictionary cannot be observed

The payoff matrix Y is not a record of what assets did pay. It is a description of what investors believe assets will pay, across every state of the world — including states that have not happened, and including the future resale prices of the assets themselves.

That distinction is the whole game.

Suppose you have decades of data. Every realized return, every holding, every price. Now I, sitting beside you, change the assumed payoff of some asset in a state of the world that has not yet occurred in your sample. Perhaps it never occurred at all — a market regime that simply didn’t show up.

What have I changed? I have changed Y. Therefore I have changed Y⁺, the dictionary. Therefore I have changed A — every estimated demand elasticity in your model.

What have I left untouched? Every single realized return in your dataset. Every holding. Every price. The data is identical.

Two economies, then, with genuinely different demand elasticities — different answers to “how much do prices move when a central bank buys” — and not one observation that could ever tell them apart. The historical record is mute on the question, because the historical record is a handful of states that happened, and the dictionary depends on all the states, including the ones that didn’t.

This is what an economist means by a non-identification result, and it is fatal to the model-free ambition. Any observed pattern of holdings is consistent with infinitely many combinations of preferences and dictionary. The data, by itself, cannot pick one. It cannot even narrow the field.

You can still write down a model and estimate numbers — people do, every day — but the model is doing the identifying, not the data. The number reflects the theory you chose, not the theory you tested.

The authors put it with admirable bluntness. Estimated demand elasticities, they write, are contingent, model-specific constructs which reflect — rather than validate — a priori assumptions on investor behavior.

Read that twice. The elasticity does not come out of the data. It comes out of the assumptions you carried in. The data’s only role is to sit there, politely, while you confirm what you had already decided.

“But surely a clever instrument...”

At this point every empirically-trained economist reading this has the same reflex, and it is a good reflex. It is the reflex that has rescued a thousand identification problems.

Find an instrument. Find some exogenous shock to the supply of an asset — a Treasury auction surprise, a mechanical index reweighting, a forced sale by a constrained institution — and use it to trace out the demand curve. Supply moves for reasons unrelated to demand; price responds; you read the elasticity off the response. This is the canonical move, and in ordinary markets it works.

In asset markets, the paper shows, it does not. And the reason it does not is no-arbitrage — the single most load-bearing assumption in all of finance.

No-arbitrage says, roughly, that you cannot get something for nothing; that two portfolios delivering the same payoffs must cost the same. It is the assumption underneath every pricing model anyone has ever taken seriously. Nobody is giving it up.

But no-arbitrage has a consequence. If you shock the supply of one asset, its price moves — and no-arbitrage forces the prices of every other asset with overlapping payoffs to move along with it. Overlapping payoffs are not the exception. They are essentially universal: almost every asset pays off in good times, almost every asset pays off alongside almost every other asset in some state. Clean, isolated, one-asset-at-a-time price variation is not something the no-arbitrage world permits.

So your beautiful exogenous supply shock does not nudge one price while the others sit still. It moves the entire interlinked system. The variation you needed — this price changes, all else equal — is precisely the variation no-arbitrage forbids.

And it gets worse, in a way I find genuinely elegant. The paper shows that the price changes a supply shock induces are not merely contaminated. Generically, they point in the wrong direction. The state-contingent prices that should rise, fall. Asking what would happen if only one asset’s price moved is, in an asset market, asking what would happen if no-arbitrage didn’t hold. It is not a hard experiment. It is an incoherent one.

“But surely with enough structure...”

One move remains, and it is the sophisticated one. If you cannot observe Y, impose structure on it. Assume returns follow a factor model — a few common factors plus idiosyncratic noise, the workhorse of modern asset pricing. Surely that pins the dictionary down well enough to make progress.

It does not. And the proof of why it does not is the most physics-flavored thing in the paper.

The authors reach for random matrix theory — the mathematics Eugene Wigner developed in the 1950s because the energy levels of heavy atomic nuclei were too complex to compute, and he realized you could ask what a typical such system looked like instead. Random matrix theory is what you use when the specific details are hopeless but the statistical geometry is tractable.

Applied here, it delivers a result I would call funny if it were not so damaging.

In a large market, the sign of any given entry of the dictionary Y⁺ — not its precise value, just whether it is positive or negative — is asymptotically a coin flip.

A fair coin. You impose your tasteful factor structure, you turn the crank, and whether the dictionary tells you to buy or to sell is, entry by entry, fifty-fifty.

Then comes the kill shot. Take two economies with the identicalfactor structure — same factors, same loadings, differing only in idiosyncratic noise — and their dictionaries agree on sign about half the time. Knowing the factor structure buys you nothing. It cannot, because the instability lives in the geometry of the payoff matrix, not in the part the factor model describes.

The authors then run the experiment. They take real S&P 500 stocks, build payoff matrices from realized quarterly stock payoffs — end-of-quarter prices plus dividends — invert them, and count. Theory predicts the entries of Y⁺ should be positive about half the time. The data says: 50.58 percent.

The coin is fair. The dictionary is noise. There is no factor model clever enough to fix it.

Two halves of the same disease

Step back and look at the two papers together — the first essay and this one.

The first essay went after the predictive wing of academic finance and found that its peer-reviewed forecasts forecast no better than mechanical mining. Theory added no predictive value.

This paper goes after the structural wing — the wing that always considered itself the more rigorous one, the more serious one, less vulnerable to the data-mining critique precisely because it wasn’t fishing for predictors, it was estimating deep parameters. And it finds that the deep parameters were never identified from the data in the first place. The central numbers are model outputs in measurement’s clothing.

Prediction fails. Measurement fails. Same disease, two organs.

And the disease has a name. Academic finance has spent decades borrowing the vocabulary of physics — “structural parameters,” “identification,” “equilibrium,” “elasticity” — without borrowing the part that makes physics work. In physics, the Standard Model committed in advance to a scalar particle with definite quantum numbers and couplings; once its mass was measured, every other property was fixed and falsifiable. General relativity predicted gravitational waves a century before LIGO heard one. The point is not that physics knows everything in advance. It is that real theory sticks its neck out: the data has a vote on whether the theory survives.

Academic asset pricing produces a number — “the demand elasticity is 0.4” — and when you ask what would have to be true in the world for that number to be wrong, the answer is: nothing the data shows. A different modeler, with a different assumed payoff matrix, gets a different number from the same data, and the data has no vote.

That is not a theory in the sense a physicist means the word. It is a very elaborate way of writing down your priors.

A physicist on the receiving end

I want to be personal for a moment, because this paper has changed something for me — and not in the direction you might assume.

For years, people have quoted these elasticity numbers at me.

It happens at dinners. It happens on panels. It happens in my comment section and in my inbox. I will make an argument — that the rise of passive investing is not the mechanical, price-inflating monster it is often made out to be, say — and someone will produce a demand elasticity from the Koijen-Yogo literature the way a lawyer produces an exhibit. The estimates show markets are inelastic. The number says you’re wrong. This is settled.

Until now, my pushback had to come from intuition. From three decades of watching markets do things no demand system anticipated. From a physicist’s itch that a quantity with a published standard error was being treated as a fact about the world when it smelled like a fact about a modeling choice. I believed the numbers could not bear the weight being placed on them. I could not prove it.

Now I can. The proof is this paper, together with a companion: “Demand-System Asset Pricing: Theoretical Foundations,” by the same three authors. In the companion, they take the workhorse logit demand system — the one Koijen and Yogo built the field on — and prove formally that it can report a measured elasticity near one when the true structural elasticity is near infinite.¹ Sit with that. Not off by twenty percent. Not off by a factor of two. The measured number sits near the low end of the scale — “not very responsive to price” — while the truth sits at the other end of the universe — “perfectly responsive.” It is the entire dynamic range of the concept, collapsed onto a single misleading number.

That is not measurement error. That is measurement theater.

The honest description of the situation is this: nobody knows the demand elasticity of the stock market. Nobody can know it from the data alone. And the published standard errors you have been shown describe the precision of an assumption, not the precision of a fact.

So the next time someone quotes one of these numbers at you as though it closes an argument, you may tell them, with a clear conscience, that it opens one.

Why this is how I update

Let me connect this to something I have said in these pages before, because the connection matters.

I have argued, more than once, that competing for alpha is a zero-sum game and mostly a poor use of a human life — unless you are willing to do what Renaissance Technologies’ Medallion fund does, which is to sweep computationally for every signal that exists, keep what survives, and never once require an explanation for why it works.

The trilemma does not formally prove that this is the only honest road to alpha. It would be too neat for any single paper to. What it does is something more useful to a working investor than a theorem: it tells me the other road — out-theorize the market on the basis of superior academic understanding of structural parameters — was never the road I thought it was, because the structural parameters were never what the field claimed they were.

If demand elasticities are model-contingent objects rather than measurements, then theory-driven alpha — the proud claim “I understand the deep structure of this market better than you do, here is my equilibrium model, here is my estimated elasticity, here is my edge” — is doing something other than what it advertises. You cannot out-theorize a market when the theoretical objects in question are not pinned down by the data anyone has access to. The supposed edge is an edge in choosing a model. Choosing a model is not, by itself, an edge.

That leaves, as a trader rather than a journal referee, exactly two roads I trust.

The first is to sweep harder than everyone else. This is the Medallion road. As I quoted Peter Brown of Renaissance last time: the durable signals are the ones nobody can explain — because the ones that can be cleanly explained get reasoned out and competed away, while the inexplicable ones survive precisely because no tidy story recruits a crowd to arbitrage them to death. If theory worked, those signals would not exist. They exist because it doesn’t.

The second is to find a corner of the market so uncrowded that the standard machinery has not arrived yet, and work it before it does.

Both roads are punishing. The first demands an infrastructure that costs a fortune and a tolerance for not understanding your own returns. The second has a low ceiling and a short shelf life — niches get found. But notice what is not on the menu. There is no third road, the one where you read the Journal of Finance very carefully, absorb the latest structural estimates, and find twenty-dollar bills lying on the sidewalk. That road was always an illusion. This paper is the document that explains, at the level of mathematics, why.

The deepest thing the trilemma tells a working investor is not “this or that elasticity is wrong.” It is that the entire enterprise of competing on the basis of superior academic understanding of structural objects rests on a foundation that was never there.

What to do with all this

If you are a retail investor: the next time you read that index funds have inflated stock prices by some authoritative-sounding percentage, ask one question — what payoff matrix did the author assume? They did assume one. They had to. And they cannot show you it was the right one, because no one can.

If you are a quant: “structural” is a word, not a guarantee. If a strategy’s edge depends on someone’s estimated demand elasticity, you are not trading a feature of the market. You are trading that person’s prior — and you are paying full freight for the privilege.

If you make policy: claims about the demand-channel transmission of QE that rest on model-free asset-demand elasticities are not measurements with merely wide error bars. They are model-contingent objects dressed up as empirical findings. I had a line ready comparing them to climate forecasts a century out — but that flatters them badly. A century-out climate forecast is anchored in conservation of energy and radiative physics. These elasticities are anchored in an assumed matrix nobody can see. Treat them with the humility that follows: not as theology, and certainly not as physics.

If you are an academic: a discipline that publishes structural estimates without flagging, loudly and in the abstract, that those estimates are not identified model-free is doing something. Whatever it is, it is not measurement. It is model discipline wearing a lab coat.

The Emperor’s wardrobe

In the first essay the Emperor had no alpha. His predictions did not predict.

Now we learn he has no identification either. The structural numbers — the elasticities, the demand curves, the deep parameters that were supposed to be the rigorous, grown-up part of the field — are not measurements at all, in the sense the word usually means. They are assumptions in measurement’s clothing.

A large slice of modern empirical asset pricing depends on an object, the latent mapping Y⁺, that cannot in principle be recovered from data alone. The discipline knows this now. It has been told, in a long and careful proof, by three of its own. Whether it changes anything is a separate question, and I would not bet the way the optimist bets.

Two of the men who delivered the message work at the school where I learned to do this. So I will permit myself the local idiom. The horns are in, and they are in deep.

There is a third paper, and it is the strangest of the three. It sets out to defend academic finance — to rescue a respectable, risk-based story for the famous anomalies — and ends up indicting the field harder than either demolition above. It shows that even the alpha you can measure, the headline number itself, the most-studied quantity in all of empirical finance, refuses to hold still: adjust for risk a little differently and value’s alpha runs from substantial to essentially zero.

The post I will write about it is called, inevitably, The Emperor Has No Benchmark. But that is for next time.

The first essay in this series, The Emperor Has No Alpha, is here. For more on markets and trading, see the full series. My book, The Science of Free Will, is available now. For more on the background behind this work, see my first and second interviews with Titans of Tomorrow, and Part 2 of my Algorithmic Advantage Podcast conversation with Simon M. on trading.

1. Koijen and Yogo have replied that their estimator identifies the relevant elasticities. The authors have responded in turn that this defense works only at a knife-edge symmetry case. That dispute is technical and ongoing; it does not bear on the point here, which is that without a maintained payoff structure the elasticity is not being read model-free from the data. ↩

I grew up in a country that decided, as a matter of high-minded national policy, that prosperity was vulgar.

This is not a figure of speech. From the 1950s to 1991, India ran what its critics came to call the License-Permit Raj — a system in which you needed a government permit to build a factory, expand a factory, import a machine for the factory, and then, once the factory existed, government permission to shed workers or close the thing. (Read that last part twice. In any sizable establishment — broadly, a hundred workers or more — the pathology wasn’t “you may not hire.” It was “hire, and then good luck ever adjusting” — economic flypaper.) The result was what the economist Raj Krishna in 1978 dubbed the “Hindu rate of growth”: an economy that grew at roughly 3.5% a year while its population grew at over 2%, which is to say, an economy that mostly stood still while telling itself it was on a journey. While India crawled, the Asian tigers ran — Singapore, South Korea, and Taiwan all compounding at around 9% a year over the same stretch. Not because they were laissez-faire fantasies; they ran plenty of industrial policy. The difference was that they forced their firms to face the world and compete, while India sheltered its firms from reality and called the shelter a plan. The nation that gave the world the zero spent four decades demonstrating what one looks like at national scale.

I lived in the aftermath of that. I watched, later, as the 1991 reforms — forced, not chosen; India liberalized because it was hours from defaulting and the IMF had it by the throat — unwound the machine and the economy finally began to grow the way an economy of a billion clever people always should have. Daron Acemoglu and James Robinson wrote a whole book about why some nations stay poor on purpose (Why Nations Fail; Acemoglu, Robinson, and Simon Johnson shared the 2024 economics Nobel for the underlying work on institutions and prosperity), and India across those four decades is very nearly the textbook case. The aftereffects will be felt for a century. The reforms are thirty-four years old and the country is still digging out — though “digging out” undersells the turn. The economy that had crawled at the Hindu rate surged after reform: 7.5% growth in 1994–97, above 9% in the 2005–08 boom, around 8% across the high-growth stretch after 2003; per capita income has since climbed past $2,600, many multiples of its early-1990s level. (The honest caveat, which I’ll make before a critic does: growth had already ticked up to ~5.8% in the 1980s — but on borrowed money, which is precisely why it ended in the 1991 near-default crisis that forced the reforms. The spree was the disease’s last fever, not its cure.) Four decades of standing still, then a sprint the moment the machine came off. That is the whole argument in one country’s national accounts.

So when I tell you I am not a socialist, understand that this is not a political preference. It is an immune response.

In the old days, before we had clean attenuated vaccines, you protected a child against smallpox by deliberately giving them a small live dose of the real thing — variolation. If it didn’t kill you, you never got smallpox. That’s me. I was variolated. I took the live virus of actual, functioning, on-the-ground socialism as a child, and the antibodies never went away. I cannot get the disease. I have tried to be fair to it, the way you try to be fair to a thing that has demonstrably tried to kill you, and I keep arriving at the same place: it does not work, it has never worked, and the people who insist it would work this time, with the right people in charge are describing a chemistry experiment that has been run several hundred times and always produces the same precipitate.

I bring all this up because America is currently being shouted at, from both ends of its political spectrum, by people who have never been variolated against anything.

This is a piece about those people. Both kinds. And the thing I most want you to take away is that they are, underneath the costumes, the same people making the same mistake.

The mistake, stated once, cleanly

Here is the mistake.

The United States is a working system. Not a perfect one — there are no perfect ones, and anyone who tells you they’ve found one is selling either a religion or a revolution, which historically have a way of becoming the same product. But a working one: a machine that has, for two and a half centuries, converted human beings of every conceivable origin into the largest and most innovative economy in the history of the species — and one of the most productive — and done it while remaining free enough that the people inside it can stand on a soapbox and call it a hellscape without being sent anywhere unpleasant for saying so.

The mistake — and it is bipartisan, it is the great unifying stupidity of our moment — is to look at a working system, notice that it is not an idealized system, and conclude that it should be torn down and rebuilt according to a theory.

The MAGA right has a theory. The theory is that there was a golden past — roughly 1955, soft-focus, the factories humming, the borders sealed, the trade balanced, the men employed and the women home — and that this past was stolen by foreigners and globalists and that we can tariff and deport our way back to it.

The socialist left has a theory. The theory is that there is a golden future — fully planned, fully equal, the rents frozen and the groceries municipal and the billionaires abolished — and that this future is being withheld from us by oligarchs and corporations and that we can tax and decree our way forward to it.

Look at those two sentences. Look at them. One worships a past that never existed; the other worships a future that has never once arrived. One blames the foreigner and the globalist; the other blames the oligarch and the corporation. Different villains — but watch what happens when the policy actually touches a foreigner. The right won’t trade with him because he’s foreign: he talks funny, he works cheap, he isn’t us. The left won’t trade with him either, only the left does it while professing to love him — trading with him is “exploitation,” the factory must be repatriated, the supply chain brought home. One bars the door out of contempt and the other bars the same door out of solidarity, and the foreigner, standing on the far side of an identical closed border, is invited to admire the difference in their reasoning — the difference that somehow justifies the exact same result.

And here is where I stop being polite, because this is the part that actually offends me. Who, exactly, are these people? In a society that calls itself free, who are you — tariff-writer, supply-chain-repatriator, either one of you — to stand between me and a man in Shenzhen and forbid the two of us from voluntarily exchanging goods to our mutual benefit? What business of yours is it? Now, there is a respectable answer: that my trade with Shenzhen has third effects — a laid-off worker in Ohio, a strategic dependency, a cost borne by someone who never sat at the table. Fine. But notice where that puts the burden. It is on you, the one reaching for the veto, to make the affirmative case that the third-party harm is real and large and not merely asserted — not on me to justify a peaceful exchange that is, on its face, nobody’s business but mine and his. And if you can’t carry that burden, then the question stands: if you may veto a voluntary trade between two consenting people, may I veto yours? Because if I may, then no one can transact and the word “free” is a fraud; and if I may not, if the veto runs only one direction, then you are not my fellow citizen in a free country. You are my sovereign, and I am your subject, and we should at least have the decency to call it what it is. The foreigner and I have the same two-word reply to that arrangement, and it translates cleanly into every language on earth, and it is not thank you.

And then — because the rage isn’t even the embarrassing part — there’s the arithmetic. Both factions want the trade deficit abolished. Both factions also want the world to keep pouring money into the country: into Treasuries, into equities, into the factories they claim to be defending. Here is the problem, and it is not an opinion, it is an accounting identity: those are the same thing with the sign reversed. A current-account deficit is, on net and by construction, mirrored by a capital-and-financial-account surplus — foreigners sending you goods and services now, and the balancing entry is their savings buying up claims on America. (Yes, the current account is broader than the trade balance — it folds in services and investment income too — but the identity holds all the same.) You cannot, on net, both ship the world more goods than it ships you and be the world’s net destination for its savings; the trade surplus you say you want would, by the same identity, make you a net exporter of capital, not the recipient of it. MAGA wants to kill the trade deficit while the world still finances American consumption on terms almost any chronic debtor in history would envy. The left wants to reshore the factories while foreign savings funds the federal deficit it depends on. Both are demanding the two halves of an identity and disowning the equals sign between them. There is an honest way to shrink the deficit — consume less and save more, at home, as a nation — but neither faction will say that out loud, because “spend less” wins no votes, so they reach for tariffs and decrees that fight the identity by fiat instead of changing the saving that drives it. It is the economic equivalent of insisting on a triangle with four sides, then blaming China when you can’t draw it.

And there’s a third freedom they’re both busy strangling, and it’s the one that should be hardest to get wrong, because it is the single most distinctive thing about this country. Almost every free nation on earth protects speech. Only America makes censoring it nearly impossible. That is not a small distinction; it is the whole difference between a right and a permission. The European Convention, the Canadian Charter, the German Basic Law — all of them guarantee free expression and then attach a carve-out, for “hate speech,” for “public order,” for “dignity,” and a carve-out is just a lever the government keeps within reach for the day it decides your speech qualifies. The First Amendment keeps no lever. Congress shall make no law is not a balancing test; it is a prohibition on the state itself. This is why the United Kingdom arrests people over tweets, why Germany sends police to homes over online insults, why Canada convenes tribunals over the wrong opinions — and these are not tyrannies, that is precisely the point. They are free countries that kept the lever. America is the one major country that constitutionally smashed it.

And both sides want it back. The right spent a decade as self-appointed free-speech absolutists and now pulls books from libraries, legislates against teaching the history it finds uncomfortable, menaces broadcasters over coverage it dislikes, and would jail you for burning a flag. The left built the speech-policing machinery on the campus and inside the platforms, coined the slogan “hate speech is not free speech” — which is verbatim the European carve-out the First Amendment exists to forbid — and leaned on private companies to throttle views it deemed dangerous. Each faction is utterly convinced it is the defender of speech while it strangles the speech it hates. And neither will say the actual principle aloud, because the principle is unbearable to both: free speech is only ever tested by speech you find vile. The speech you like was never in danger. The First Amendment is for the Nazi in Skokie, the flag-burner, the campus crank, the conspiracy peddler — and the moment either tribe gets the power, it discovers an emergency exception for exactly the speech that offends it, which is the only speech the Amendment was ever about. (Yes, even America draws lines — incitement, true threats, defamation, obscenity, and a few other narrow categories. But those are historically defined exceptions tied to concrete legal harms, not licenses to ban an opinion because the state finds it hateful or false. The gap between “no viewpoint may be banned” and “offensive viewpoints may be banned” is not a difference of degree. It is the entire ballgame, and America is the only major country standing on the right side of it.) An immigrant from a country with the lever does not experience this as a culture-war token. He experiences it as the reason he can publish a post calling the mayor a halfwit and not receive a visit. He knows what the lever is for. That is why it is so obscene to watch native-born members of both parties reach for it as though it were a normal instrument of government.

Strip away the costumes and what’s left is identical: a theory that has never once survived contact with reality, and a serene conviction that the problem is other people — that if we could just remove the right ones and install the right plan, the heavens would open.

This is the same person. He has simply bought two different hats.

What the immigrant knows that the theorist does not

Now. Here is the part that matters, and it is the reason I, specifically, get to write this.

All of which raises the obvious question: if the freedoms are this basic and the arithmetic this elementary, why do so many people get it wrong? And the answer points straight at the one group nobody thinks to ask.

The point of an immigrant is not that he is virtuous. The point of an immigrant is that he is working from data and not from theory.

The native-born American radical — of either flavor — has only ever lived in the working system. He has never seen the alternative run to completion. He can therefore afford to romanticize it, because romanticizing a thing you’ve never experienced is the cheapest hobby on earth. The MAGA voter mourning the lost factory town has never had to actually live in an autarky, with the empty shelves and the ten-year wait for a telephone line that economic nationalism reliably produces. The DSA member dreaming of the planned economy has never had to actually queue for bread under one. They are both, in the most precise sense, tourists of catastrophe. They’ve seen the postcards. They’ve never been.

The immigrant has been. That’s the whole thing. That’s the entire value proposition.

When I, or my fellow refugees from various failed utopias, choose to come to the United States, we are not voting for an abstraction. We are voting with our feet, having personally inspected the alternatives, against the very theories the native-born radicals find so seductive. We came here because the planned economy didn’t work, because the strongman didn’t deliver, because the closed border made everyone poor. We are, as a population, the single largest body of empirical evidence on the planet that the working system beats the theory. We are the experiment’s control group, and we got on a plane.

And — this is the part the nativists in particular cannot metabolize — the data on us is overwhelming. We come, and we behave better than the people already here.

Take incarceration, since the MAGA right has made crime the centerpiece of an entire political program. The single most thorough and current analysis we have — the Cato Institute’s March 2026 study of the 2024 American Community Survey — found that the incarceration rate for native-born Americans was 1,195 per 100,000. For illegal immigrants, it was 674. For legal immigrants, it was 303. Read that again slowly, because it is the sort of number that should end a debate and somehow never does: a legal immigrant is about 75% less likely to be sitting in a prison cell than a native-born American, and even an illegal immigrant is meaningfully less likely than the native. The thing the entire restrictionist program is organized around fearing is, statistically, a thing native-born Americans should fear about themselves.

Take the economy, since the socialist left has decided that capitalism is a machine for crushing the vulnerable. As of 2026, immigrants had founded or co-founded 59% — 455 of 775 — of America’s privately held billion-dollar startups, up from 55% in the prior reports. Nearly half the Fortune 500 was founded by immigrants or their children — a figure that leans on counting the second generation and that skeptics at the Center for Immigration Studies have contested on methodology. Fine; throw it out and keep only the unicorn number, where the founders are themselves the immigrants. It’s still a majority. The point survives a hostile audit, which is the only kind worth making. The country at the top of that founder list, incidentally, was India — 96 unicorns, ahead of every other nation — the same India that spent forty years proving the other system doesn’t work, exporting its most capable people to go prove it somewhere that lets them.

Now, the sharp objection — and it comes from the left, which is the half of the room that actually reads the methodology sections. “This is just selection,” they’ll say. “Immigrants aren’t better-behaved or more entrepreneurial because immigration ennobles them. They’re filtered. You screen for ambition and a clean record at the visa office and self-select for risk-tolerance before anyone boards a plane. The data say nothing about inoculation. They say something about the sieve.”

Yes. Exactly. Thank you. That is the mechanism, and naming it doesn’t weaken the argument — it is the argument, stated in the language of econometrics instead of immunology. Of course it’s selection. The whole point is what the cohort selected on. These are, by construction, the people who looked at the available systems — the planned one, the strongman’s one, the autarkic one, and the messy working one — and chose the working one at real personal cost, with their feet, against the romance. “Selection effect” and “revealed preference at scale” are the same fact wearing two lab coats. The leftist thinks he’s identified a confound. He’s identified the independent variable. The immigrant cohort is precisely the natural experiment in which people who have seen the alternatives sort themselves toward the working system — which is why their behavior, once here, looks the way it does, and why the native-born theorist who has selected on nothing but a seminar reading list is the worse bet.

So here is the immigrant, holding the data. Lower incarceration. Higher top-end entrepreneurship. A revealed, plane-ticket-purchasing preference for the working system over every theory on offer. And here is the native-born radical, holding the theory, explaining to the immigrant that the system is a fraud and the answer is the very arrangement the immigrant crossed an ocean to escape.

You can perhaps see why some of us find the lecture a little rich.

The hinge: but Mamdani is an immigrant

And now we arrive at the objection that is supposed to detonate this entire essay, and which instead — if you follow it carefully — turns out to be the load-bearing beam.

“But Samir,” you say, “the current mayor of New York City is himself an immigrant. Zohran Mamdani was born in Kampala, Uganda. He’s a democratic socialist. He won. So much for your theory that immigrants are inoculated against socialism. Your control group voted for the disease.”

Good. Excellent. Let’s take it seriously, because it is the most interesting thing in the room.

The error in your objection is the error in this entire national argument, and it’s worth slowing down for. It was never the passport that inoculates you. It was the disease.

Variolation only works if you take the live virus. You have to actually have the smallpox, in a small and survivable dose, for the antibodies to form. A child who grows up reading about smallpox in a beautifully appointed seminar room, who studies the epidemiology and writes a moving senior thesis on the structural injustice of the disease, does not thereby acquire immunity. He acquires opinions about the disease. These are not the same thing. One is built from antibodies; the other is built from footnotes.

Zohran Mamdani holds a passport stamped immigrant. But consider, with genuine care, the life. His father is Mahmood Mamdani, a chaired professor at Columbia, one of the most celebrated postcolonial theorists alive. His mother is Mira Nair, the internationally acclaimed filmmaker. He grew up between Kampala and the Upper West Side of Manhattan, the son of that household. He attended the elite Bronx School of Science and then Bowdoin College. This is not the biography of someone who fled a planned economy with two suitcases and a medical degree the new country wouldn’t recognize. This is the biography of the global academic aristocracy — the single demographic on earth most marinated in the theory of socialism and most insulated from its practice.

He didn’t immigrate from the disease. He immigrated into the faculty lounge. He is, in the precise technical sense I’ve been building this whole essay around, unvariolated. He has the elite seminar’s complete theory of the smallpox and not one antibody in his blood. He is the living, governing proof of my actual thesis, which was never “immigrants are magically wise” — that would be its own dumb romanticism, and I have no interest in trading the nativist’s fairy tale for a flattering one about my own kind. My thesis was always narrower and harder: lived experience of the failed system inoculates you against it, and nothing else does. Not the passport. Not the suffering of your ancestors. Not your sincerity. The disease, or nothing.

This is why the children of Cuban exiles in Miami vote the way they do and the children of Cuban theory in the academy vote the opposite way, and why this surprises only people who think the relevant variable is ethnicity rather than memory. It’s why my own inoculation is specific and personal and non-transferable: I cannot give it to my American-born children by telling them stories, any more than you can vaccinate a child by reading him the label. The antibodies don’t cross the placenta. Each generation has to either catch the live virus or, far better, be governed by people who did and who built the institutions accordingly.

Mamdani caught the footnotes. Now eight and a half million people get to find out the difference, in real time, with their rent and their groceries and their police department as the assay.

I would genuinely rather they didn’t have to. I lived downstream of that experiment once already.

A word to each idiot, in turn

Let me be even-handed about the contempt, because both of these factions have earned it honestly and it would be unfair to slight either.

To the MAGA economic nationalist: You have decided that the path to American greatness is to make Americans poorer on purpose, via a tax you’ve been told isn’t a tax. A tariff is a tax. It is paid by the importer, who is American, and passed to the consumer, who is American, and the whole exercise is a transfer from the broad American public to a narrow protected industry that has successfully lobbied to be spared from having to be good at its job. This is not a controversial finding. It is one of the few things in economics on which essentially every economist who has ever lived agrees, from Adam Smith forward, which is a sentence I do not get to write often. You have looked at the single most successful wealth-generating arrangement in human history — open, dynamic, immigrant-fueled American capitalism — and concluded that what it really needs is the economic policy of 1930 and the demographic anxiety of 1924. You want to wall off the country that the Why Nations Fail literature identifies as history’s great winner because it stayed open, and you want to do it in the name of winning. You have mistaken the lock on the door for the engine of the house.

To the DSA socialist: You have looked at the same arrangement — the one that took my variolated, capable, formerly-poor immigrant cohort and let us build companies and pay for the schools your program depends on — and concluded that what it really needs is the economic model of the country I left. You want rent control, which 95% of a University of Chicago expert panel agree reduces the supply and quality of housing — a near-unanimity rivaling the tariff consensus — and you have decided that the lesson of every place that’s tried comprehensive planning is that they simply lacked your sincerity. They didn’t. East Germany was sincere. Venezuela was sincere — it had the largest oil reserves on the planet and the most fashionable economic theories in the faculty lounge, and it engineered mass hunger and emptied its own supermarkets. Maduro meant well, if you ask the right professor. The graveyards of the twentieth century are landscaped with good intentions and the people they starved. You are not the first to be sincere. You are merely the latest to mistake sincerity for a mechanism.

And to both of you, the thing I most want to say:

You are surrounded by people who fled the exact experiment you are proposing to run. We are right here. We came thousands of miles and filled out your absurd immigration paperwork specifically to get away from your big idea. We are the most expensive focus group in the history of the world, assembled at our own cost, and we are unanimous. And you will not ask us a single question, because the answer is inconvenient to a theory you find more beautiful than the data.

That is the part that makes me want to put my head through a wall. Not that you’re wrong. People are wrong all the time; being wrong is the human condition and I have been wrong about plenty. It’s the incuriosity. It’s the serene confidence of a man who has never run the experiment lecturing the people who survived it about what the results will be.

America Remains Great, Halfwits

So, the title.

America remains great. Not perfect — I’ve said this enough times that you’ll believe I mean it — but great in the only sense that has ever mattered: it is the place that the rest of the world’s most capable and determined people are still, despite everything, trying to get into. Revealed preference is the only honest poll. Nobody is risking their life on a raft to reach the worker’s paradise. Nobody ever did. The boats have always pointed one direction, and it is the direction of the working system, and it has been pointing that way for two hundred years through every fashion of theory that has risen up to declare the system a fraud.

The halfwits, left and right, are people who have inherited the single greatest going concern in human history — the freedom to trade with whom you please, to reckon with the numbers honestly, to say the unsayable without a knock at the door — and would like to torch it because it fails to match a picture in their heads, a picture, in both cases, of a place that has never existed, drawn by someone who has never had to live there.

I have lived there. I have the antibodies to prove it. And the message from those of us who took the live virus and survived, to everyone currently demanding we all take a fresh dose for the principle of the thing, is short and it is heartfelt:

We’ve seen this one. It doesn’t end the way you think. America remains great. Stop trying to fix it with the thing we ran away from.

ARGH.

If you found this worth your time, you may enjoy my book The Science of Free Will, which asks an equally uncomfortable question about an equally cherished story.

For the long version of the India argument — what the License Raj actually did and how the country is clawing back — start with The Paradox of India, or browse the whole series.

And for the same disease in a different country, see Albion — Britain’s institutional decline, by someone who first saw the place in 1979.

Somewhere in India right now, a schoolteacher is putting two thousand rupees into an index fund.

She did not agonize over it. She set up an auto-debit from her salary account three years ago and has thought about it roughly as often as she thinks about her electricity meter. On the fifth of every month the money leaves, buys units, and disappears into the great anonymous machinery of the Nifty. It is the single most boring financial act a human being can perform — the personal-finance equivalent of flossing. No drama, no genius, no timing. Just a small, dull, disciplined transfer, repeated until it compounds into something that might one day be a house, or a daughter’s college, or simply the difference between a frightened old age and a calm one.

And a financial columnist has identified her as a threat to the national currency.

He does not put it that way, of course. He is far too careful and far too intelligent to put it that way, which is precisely what makes the argument worth taking apart. The crude version — the proles are gambling and it’s hurting the rupee— would be easy to dismiss and not worth a thousand words. The version he actually wrote is subtler, partly correct, internally sophisticated, and arrives, with the quiet inevitability of a man reversing confidently into a lake, at a conclusion that would have made Indira Gandhi proud.

His conclusion is that India should disincentivize its citizens from investing in stocks.

This is a real argument, made in good faith, by someone who has himself been a disciplined equity investor for twenty years and who signs off by bracing for abuse. I am not going to abuse him. I am going to do something he will find considerably more irritating, which is agree with him three times before I’m done.

What he gets right (and we must say so)

The series has a rule about this: you do not get to knock down a strawman and call it a victory. So let us build the man’s argument at full strength, because most of it stands.

He is right that foreign money leaving India demands dollars. When a foreign fund sells its Indian shares, it is paid in rupees, and those rupees must be converted into dollars to go home. Multiply that by enough funds over enough years and you get genuine, measurable pressure on the rupee. This is not a theory. The Reserve Bank’s own data on the repatriation of foreign direct investment shows the number climbing steadily, and he reproduces the chart. Concede it cleanly: the dollars are leaving, and it matters.

He is right that a great deal of that repatriation has been venture capitalists and private-equity firms cashing out of loss-making startups at the precise moment those startups listed on the exchange at valuations untethered from anything so vulgar as earnings. This, too, is real. We have all watched a company that has never made a rupee of profit price its IPO as though it were minting them, while the early investors quietly walk through the exit they’d been waiting eight years for.

And he is right — most damning of all, because it’s the part everyone gets wrong — that there is no selling without buying. He says so explicitly. He does not make the lazy error. When a foreign institution dumps a billion dollars of Indian stock, someone has to be standing on the other side of that trade with a billion dollars to catch it. Increasingly, the someone catching it is our schoolteacher and ninety-six million of her countrymen, arriving on the fifth of every month with their two thousand rupees. He has correctly identified that domestic retail money is the counterparty to the foreign exit.

So far, so good. Three concessions, all earned. He understands the plumbing. He is not an idiot, whatever his relatives may say at weddings.

Here is where he reverses into the lake.

The man had the right tool in his hand

Read the column carefully and you find that he has already diagnosed the actual problem and already proposed the correct fix — and then, two paragraphs later, abandons both for something else entirely. This is the part that should make you sit up, because it is the whole tragedy in miniature.

The actual problem, by his own account, is overpriced IPOs. The mechanism is specific and correct: existing regulations let a company list by floating only a small sliver of its equity. When demand for stocks is feverish, that small float lets investment bankers price the whole thing at insane multiples, which lets insiders cash out at valuations that “would have been difficult to justify under normal market conditions.” The disease is not that Indians are buying stocks. The disease is that a structural quirk in listing rules — small mandatory float plus euphoric demand — manufactures inflated exits for a few dozen well-connected sellers.

And he names the cure himself. Raise the minimum public float. Make companies sell a bigger slice when they list, so the price has to reflect something closer to reality and can’t be juiced by artificial scarcity. He even points the finger at the right address: the Securities and Exchange Board of India, which writes those rules and could change them on a Tuesday.

This is a scalpel. It is aimed at the actual tumor. It touches the schoolteacher not at all.

So what does he do with his scalpel? He sets it down on the tray, walks across the room, and picks up a sledgehammer.

Because the recommendation he actually lands on — the one in bold, the one he braces for abuse over — is not “fix the float.” It is: tax all investment income at the same rate. Capital gains on equity should be taxed identically to interest on fixed deposits, gains on gold, gains on real estate. A “level playing field across asset classes,” so that tax considerations no longer “disproportionately influence investment choices.”

He had a tool that hit the investment banker. He swapped it for one that hits the schoolteacher. And the swap was not an accident. It is the most revealing thing in the column, because it is the oldest reflex in the entire Indian policy cabinet, and he reached for it without seeming to notice he’d done it.

The gold trap, and why inert metal is the tell

The bridge he uses to get from the IPO problem to the universal tax is gold — and it’s worth dismantling the bridge, because it’s load-bearing and it doesn’t hold.

His logic runs: India imports nearly all its gold, paying in precious dollars (seventy-two billion of them in 2025-26, more than double three years prior). This is a real drain. Gold-buying has therefore been gently discouraged — higher duties, official tut-tutting, mutual funds restricting gold-fund inflows. So, he reasons, if we discourage gold to protect the rupee, why not discourage excessive stock-buying along the same lines?

Notice what he’s done. He has taken the logic that applies to gold and walked it across to equity. And here he trips over a distinction he himself raises and then waves away. He concedes — in writing — that “gold is largely a useless asset and equity isn’t.” Then he says the distinction “weakens significantly” because so much stock-buying is just funding overpriced insider exits.

But the distinction doesn’t weaken. It is the entire point, and it survives his objection completely intact.

Gold sitting in a locker is inert. It does nothing. It produces no goods, employs no one, builds no factory, funds no payroll. A bar of gold in 2026 will be exactly a bar of gold in 2046, having contributed precisely nothing to the productive capacity of the nation in between. Discouraging it is, at worst, neutral. The gold just sits there being sat upon.

Equity is the opposite of inert. A rupee of equity is a rupee of risk capital locked into a productive enterprise — a company that hires, builds, exports, innovates, and pays corporate tax on its profits before the shareholder is taxed again on his gains. The schoolteacher’s two thousand rupees does not sit in a locker. It becomes working capital. It is the single most useful thing a household can do with its savings from the point of view of the very country he claims to be protecting.

The “overpriced IPO” objection doesn’t change this, because it confuses the price of some equity at the moment of listing with the nature of equity as an asset. Yes, some IPOs are overpriced, and yes, that funds some bad exits. But the cure for an overpriced subset is to fix the pricing mechanism — which, again, he already proposed — not to tax the entire asset class as though it were metal in a vault. You do not discourage everyone from eating because some restaurants overcharge for the bread.

The gold parallel is the tell. The moment you find yourself reasoning “we discouraged the inert thing, so let’s discourage the productive thing by the same logic,” you have stopped thinking about economics and started obeying a reflex. And we know this reflex. We have its photograph.

The experiment already ran. We have the photographs.

Here is the part the column never reckons with, and it is not a matter of theory. It is a matter of what happened three months before he sat down to write.

In March 2026, foreign institutional investors sold a record 1.18 lakh crore rupees of Indian equity in a single month — the largest monthly foreign exit on record, larger than the entire foreign inflow of the boom year 2013. The rupee was scraping ninety-four to the dollar. Oil was over a hundred. By every page of the historical playbook, this is the setup for a rout: foreign money stampedes for the door, the market collapses, the currency follows, and India spends the next eighteen months in a defensive crouch composing wounded op-eds.

It didn’t happen.

Domestic institutions — overwhelmingly the mutual funds fed by those boring monthly SIPs — bought 1.16 lakh crore rupees of equity in the same month. They absorbed almost the entire foreign exit, rupee for rupee. The Nifty fell about eleven percent and then steadied. There was no systemic panic. Analysts reached for a new phrase: India had become a twin-engine market, no longer dependent on a single foreign engine that could cut out whenever a fund manager in London got nervous about the Strait of Hormuz.

Sit with that, because it is the exact inversion of his thesis. The SIP money he wants to disincentivize is the precise mechanism that absorbed the foreign exit he is worried about. He has looked at the airbag that just deployed and saved the car, and concluded that the real problem is all these airbags.

And then there is 1991, which this series has told at length and will not retell here, except for one image. In 1991 India had no domestic buffer at all — no twin engine, no schoolteachers, no SIPs. When the dollars ran out, the country airlifted forty-seven tonnes of gold to the Bank of England and another twenty to Switzerland to raise six hundred million dollars and avoid defaulting on its debts. That is what it looks like when a nation has no domestic capital base to lean on: you fly your jewellery to London in the dead of night. We have spent thirty-five years building exactly such a base, household by household, two thousand rupees at a time. It now arrives at the rate of roughly thirty-one thousand crore a month. The proposal on the table is to discourage it, for the rupee’s sake.

He also diagnosed the wrong disease

There’s a second, quieter problem, which is that the thing actually pressuring the rupee in 2026 has very little to do with SIP investors or startup repatriation, and the column would have known this had it looked sideways instead of down.

The dominant driver of the 2026 foreign exit was a global rotation out of Indian equities and into the artificial-intelligence trade. The chips money — the allocation that used to flow to Indian IT — went to Korea and Taiwan instead. That’s structural, and it has nothing to do with our schoolteacher. The second driver was valuation: Indian stocks trading at a forward price-to-earnings ratio around twenty-two while the broad emerging-market index sat near thirteen. A growth premium India has always enjoyed; a gap that size is harder to defend. The rest — rupee at ninety-five, oil over a hundred, the US ten-year yield testing four and a half percent — is cyclical macro weather that will turn when the geopolitics turn.

And the detail that detonates the framing entirely: through all of this, foreign investors were net buyers of the Indian capital-goods and capex story in every single month of 2026. They were not fleeing productive investment. They were trimming overpriced financials and buying the factories. The picture the column paints — foreigners cashing out of loss-making startups as the defining dynamic — was a 2021 story being narrated over a 2026 market that had long since changed the channel. He prescribed poison, and he prescribed it for the one organ that was perfectly healthy.

Now, the economics — and the part where he’s not even in the room

Set aside the plumbing and the photographs for a moment and ask the cleaner question: what does the actual discipline of economics say about taxing capital?

It says, roughly, as little as possible, and the experts can’t agree on how little.

The famous result belongs to Christophe Chamley and Kenneth Judd, who in the mid-1980s, working independently, reached the same startling conclusion within standard growth models: the optimal long-run tax on capital income is zero. The result was so counterintuitive that economists have spent forty years arguing about what it even means — and, this is true and I treasure it, they describe their own models as behaving “erratically” and producing results “of limited use for guiding policy,” which is the most honest thing any profession has ever said about its own flagship theorem. In 2020, Straub and Werning reopened the question and showed that under plausible conditions the optimal capital tax is positive after all — though still, in the serious cases, lower than the tax on labour.

So here is the honest state of the field, the version no economist can dock me for: the entire serious literature on optimal capital taxation lives in a band running from a floor of zero up to a ceiling of “positive, but generally below what you’d tax wages.” Reasonable people argue about where in that band to stand. It is a real, live debate.

His proposal is not in the band. Tax equity gains at the same rate as fixed-deposit interest, as gold, as real estate treats the formation of productive risk capital as identical to the interest on a savings account and the appreciation of metal in a vault. That is not a daring position at the high end of the debate. It is a refusal to attend the debate. He is not wrong within the discipline; he is standing in the parking lot, explaining to the valet why the building is on fire.

But here is the thing — and this is where I have to take the column behind the shed — even that band is too kind to him, because it is measuring the wrong thing entirely.

Every paper in that literature is, at bottom, about revenue: given that the state needs money, what is the least-destructive way to extract it. Which quietly smuggles in an assumption nobody examines — that the number worth maximizing is the government’s haul. And for a poor country sprinting to get rich before it gets old, that is the wrong number. The number that matters is growth. And growth, it turns out, rides a completely different curve.

I’ll keep this painless, because I develop the full version elsewhere and you have a life. Picture two curves. One plots government revenue against the tax rate — the famous Laffer curve, the one that bends back on itself because a 0% rate and a 100% rate both raise exactly nothing (at zero there’s no tax; at a hundred, nobody bothers earning anything to be taxed). Fine. But there is a second curve, which almost nobody bothers to draw, plotting economic growth against the tax rate. It too is pinned at two zeros: at 0% tax there is no government, no courts, no contracts, no roads, so growth is roughly nil; at 100% tax nobody gets out of bed, so growth is also nil. Which means it, too, has a peak — a growth-maximizing rate, sitting somewhere in the middle.

Here is the entire game in one sentence: there is no earthly reason the peak of the revenue curve sits at the same tax rate as the peak of the growth curve. They are two different functions. Expecting them to share an optimum is like assuming the fastest route to the airport must also be the most scenic. Occasionally, by dumb luck. Mostly, no.

And then the detail that turns a debating point into a sledgehammer: growth compounds. A revenue shortfall is a one-time number. A growth shortfall is a rate, and rates, left alone with time, do unspeakable things. Half a percentage point of extra growth, compounded across our schoolteacher’s working life, does not merely trim the alternative — it laps it, then laps it again, then wanders off to see if there’s anything else worth lapping. So once you take compounding seriously, the contest between the two curves stops being a contest. The rate that maximizes revenue in the long runsimply is the growth-maximizing rate, because the largest tax base any government will ever lay its hands on is the one bolted to the economy that grew the fastest.†

Which means the literature’s ceiling — “positive, but below labour” — was a revenue ceiling all along. For a capital-starved economy whose entire purpose in this decade is to accumulate the capital it does not yet have, the right rate sits lower still, shoved back toward the floor by the one force the revenue models forgot to invite to the meeting: compounding.

So our columnist is not merely standing above the band. The band was the wrong band. He has aimed a tax squarely at the single input — capital — whose accumulation is the growth he needs, in the single country that can least afford to discourage it, having reasoned his way there with great care along a curve he did not realize was the wrong curve. It is, in its way, an extraordinary feat of marksmanship. He has shot the one organ keeping the patient alive — from the parking lot, while telling the valet about the fire.

The conclusion he says he isn’t making

He will object, and loudly, that he proposed nothing so dramatic. A modest equalization of rates, he’ll say. A nudge. Hardly Indira Gandhi. I said no such thing.

He’s right. He didn’t. But he doesn’t get to keep the principle and disown its destination.

Because the principle he is actually endorsing is this: when citizens do something productive at a scale that creates demand for dollars, the state should tax it to discourage them. And once that is the principle, the only question left is the dosage — and there is no number printed on the bottle. If a mild tax is good because it gently discourages the rupee-draining behaviour, then a stiffer tax is better, because it discourages more of it. And a punitive tax is better still. Follow the logic where it actually goes and you do not stop at “the same rate as fixed deposits.” You keep climbing, all the way up, until you arrive — breathless, triumphant, brandishing a chart of FDI repatriation — at a 97.75% marginal rate.

We have been to that summit. India planted a flag there in 1973-74, when the top rate hit 97.75% and a citizen who earned a hundred rupees over the threshold was graciously permitted to keep two and a quarter. It was done for reasons that sounded, at the time, every bit as sober and public-spirited as this column: the prosperous must shoulder the burden, private excess must be curbed, the national interest demands restraint. The result was not restraint. The result was that tax evasion became the national pastime, revenue collapsed, and the policy was quietly strangled within a few years, because it turns out that when you tax something at 97.75% people do not line up to pay — they line up to hide.

The columnist is not Indira Gandhi. He is something this series finds far more interesting: a thoughtful, numerate, twenty-year equity investor who has, without quite noticing, picked up the exact instrument she used and is proposing to apply a gentler dose of it. But it is the same instrument. The principle that licenses his nudge is the identical principle that licensed her sledgehammer; they differ only in how hard the arm swings. And a principle whose logical culmination is 97.75% does not have a calibration problem. It has a direction problem. It is pointed the wrong way — and worse, it is pointed the wrong way down the single axis, growth compounded over a lifetime, that decides whether the schoolteacher’s grandchildren are middle-class or merely hopeful.

The bug that keeps shipping

He closed by bracing for abuse — “let the abuse begin.” But the abuse was never the danger, and it isn’t coming, at least not from here.

The danger is that he articulated, out loud, in a well-reasoned column, the single most durable bug in India’s operating system: the deep, autonomic conviction that when the citizenry starts doing something productive at scale, the state’s first instinct should be to find the lever that makes them do less of it. Not to fix the specific distortion — he found that fix and set it down on the tray. Not to target the actual bad actor — the investment banker pricing the garbage IPO sails through entirely untouched. But to reach, by reflex, for the instrument that disciplines the broad mass of ordinary people, because suppressing citizen behaviour is wired more deeply into the machine than targeting the culprit ever was.

This is the Licence Raj in a SIP-shaped costume. It is the same reflex that taxed toothpaste as a luxury, that made you wait eight years for a telephone, that threw out IBM and Coca-Cola to teach foreigners a lesson, that retroactively taxed Vodafone — and then, when the courts and the howls of foreign investors forced a humiliating repeal a decade later, learned precisely nothing from the episode — that demonetized 86% of the currency overnight to catch black money it did not catch. Each time, the costume is new and the reasoning sounds responsible. Each time, the instinct underneath is identical: the energy of a hundred million ordinary Indians, doing the dull and disciplined thing, treated as a leak to be plugged rather than as the entire point of the exercise.

He won’t get abuse. He’s too smart, and he’s mostly right about the diagnosis, and being mostly right is exactly how the dangerous ideas get past the bouncer. What he’ll get, if the bug ever ships to production, is a policy paper, and a committee, and possibly an award — and our schoolteacher, on the fifth of some future month, will find that the most boring and responsible thing she could possibly do with her two thousand rupees has been gently, fairly, neutrally made a little less worth doing.

For her sake, and the rupee’s, I hope SEBI just fixes the float.

†A caveat for the one reader already sharpening a reply: this holds cleanly where the tax simply discourages productive activity — which is precisely what is proposed here. If the revenue were spent building roads and schools that themselves raise growth, the very low end of the rate gets muddier. But “disincentivize” is not a plan to fund human capital. It is a plan to do less of the thing. The deadweight case is the one in front of us.

The growth-versus-revenue argument — and the four separate nonlinearities that quietly turn almost every tax debate you will ever hear into a confused argument about the wrong curve — gets a full chapter in The Science of Free Will (US · India · UK). Yes, it is a book about free will that contains a chapter on the Laffer curve. It makes sense in context. Mostly.

Elon Musk wants to put AI data centers in space, and the internet has decided this is stupid. The favored reason is thermodynamic: space is a vacuum, there’s no air, how do you cool a gigawatt of GPUs with no air to carry the heat away? The dunk writes itself. The man forgot about heat.

I’m a physicist. I don’t have a horse in this race, and I find the dunk-or-defend reflex tedious. There is exactly one honest way to settle whether orbital data centers make sense, and it’s the way Musk’s own engineers would do it: write down the physics, write down the costs, and turn the crank. The answer falls out — and it turns out to be neither side’s answer.

So let’s turn the crank. No thesis. Wherever the arithmetic lands is where we land. I’ll cite every number; the full table is at the foot of this post. And every time I hit a judgment call, I’ll round it in Musk’s favor — generous emissivity, best-in-class solar, the friendliest orbit — so that whatever the ground wins, it wins on the orbital case’s best day.

First, the heat — because that’s what everyone’s arguing about

A hot object sheds heat three ways: conduction, convection, radiation. Conduction needs contact. Convection needs a fluid — air or water — to carry warmth off. A vacuum offers neither, so in orbit a radiator has one channel: thermal radiation, governed by the Stefan-Boltzmann law,

where q is watts radiated per square meter, ε is surface emissivity (a good coating reaches about 0.9), σ is the Stefan-Boltzmann constant (5.67 × 10⁻⁸ W/m²K⁴), T is the radiator’s temperature, and T_env is what it faces. Temperatures in kelvin, since the fourth power only behaves measured from absolute zero.

The intuition that excites people is the cold of space — 2.7 K, the leftover chill of the Big Bang. Surely radiating into a 2.7 K void beats a warm room? Plug in the numbers. A radiator at 75 °C (348 K), emissivity 0.9:

Radiating into deep space roughly doubles the heat you shed by radiation, versus a room-temperature sink. Real, and worth having.

But that doubling is the entire radiative prize, and it’s smaller than it looks, because radiation isn’t the only way to cool something — it’s just the only one left in a vacuum. On the ground that same panel also has convection: air moving across it, carrying heat off directly. With a fan, convection alone removes several thousand watts per square meter — multiples of what radiation manages in space. So on cooling, the honest tally is: orbit gives you a 2× bump on the radiative channel and takes away the convective channel, which was worth far more. On cooling alone, the ground wins. (This is the part the critics get right; Sam Harsimony made the case cleanly: build one radiator, fly a copy, bolt the other to a warehouse floor — the warehouse copy sheds more heat and never paid for a launch.)

One honesty note on my own number: a real radiator in low Earth orbit doesn’t face a clean 2.7 K universe. It sees the Sun, sunlight reflected off the Earth, and the Earth’s own infrared glow. NASA’s thermal-control models treat spacecraft temperature as a balance of all those inputs plus what the radiator sheds. So 750 W/m² is an optimistic, deep-space-facing figure — again, rounding in orbit’s favor.

So the critics are right that orbit is a worse place to dump heat. But here’s the move that the whole debate misses: being a worse radiator is not the same as being uneconomic. The question was never “is space good at cooling.” It’s “does the whole machine pencil out.” For that we need the radiator’s mass, not its efficiency — and then everything else.

Now turn the crank on the whole system

Build the satellite, per megawatt of compute, and price it against a data center on the ground.

How much radiator? At 75 °C, two-sided, at a spacecraft-radiator areal density of 5 kg/m² (real flight hardware runs heavier), the Stefan-Boltzmann number above gives 3.3 tonnes per megawatt. Three honesty notes, because a thermal engineer will check. A 75 °C radiator implies running the chip junctions hotter still — around 100 °C, since you need a temperature gradient to push heat from junction to coolant to panel — which is aggressive but plausible for purpose-built silicon (unlike a terrestrial DGX, which wants 30 °C air). The 750 W/m² assumes the panel mostly faces deep space; a face turned toward the Earth, which glows in the infrared at around 255 K, sheds noticeably less — which means the satellite has to hold an edge-on-to-Earth, sun-avoiding attitude to hit that number, and that attitude constraint isn't free, because it also dictates where the solar array is allowed to point. And the 5 kg/m² panel is the lightweight end; many flight radiators are heavier. All three assumptions are the generous-to-Musk end, which means 3.3 tonnes is a floor on radiator mass — and even at the floor it's almost an afterthought. The critics were right that the radiator is large in area — a megawatt needs about 1,300 m² of radiating surface, or roughly 670 m² of physical two-sided panel: a quarter of a football field if you count both faces, an eighth if you count the deployed footprint — but in mass, the thing a rocket cares about, it barely registers. Costed honestly, the heat problem everyone's dunking on is a few tonnes.

What else flies? Compute, first. As a mass proxy, an NVIDIA DGX H100 is 130 kg and draws 10.2 kW, which is about 13 tonnes per megawatt. A real orbital machine wouldn’t be a DGX rack bolted to a satellite — it would need space-qualified packaging, shielding, thermal interfaces, and robotic-serviceable mounting, all of which add mass I’m ignoring. Then the solar array, using best-in-class flexible panels at 150 W/kg (typical space arrays manage a fifth of that): 6.7 t/MW in the best orbit. Add batteries to run through Earth’s shadow, plus structure and power electronics at a 30% allowance, and the whole satellite comes to roughly 30 tonnes per megawatt in the friendliest orbit, 41 in a generic one.

Now lift it. At today’s Falcon 9 price of about $2,700/kg, thirty tonnes costs $80M per megawatt — far above the roughly $11M/MW it takes to build a data center on the ground. The critics who say “launch is too expensive” are right about today. But Musk isn’t betting on today; he’s betting on Starship. SpaceX doesn’t publish an operational price per kilogram, so the figures everyone quotes — $100 to $300/kg — are projections for a mature, fully reusable system, not posted fares. Grant the optimistic end anyway. At a hypothetical $100/kg, thirty tonnes launches for $3M/MW — below the ground build. So the first launch is not the killer. On the narrow question “can you afford to put it up there once,” Musk wins the day Starship hits its targets.

Heat: not free, but no longer the obvious killer — 3.3 tonnes. Launch: also not the obvious killer, in a mature-Starship world. So what is? This is where the crank turns up the thing neither camp is shouting about.

The clock nobody priced

GPUs don’t last. NVIDIA ships a new architecture every 18 to 24 months, each generation improving performance per watt — though by how much is workload-specific and vendor-reported, not a clean law; NVIDIA’s own roadmap claims range from modest to 10× depending on what you measure. For the toy model I’ll assume frontier performance per watt doubles every D years and use D = 3 as a central case. That is not a law of nature; it’s the single parameter the whole result turns on, which is precisely why I’m flagging it. (I’ll call this “Moore’s law” colloquially, but the real variable is accelerator useful-work-per-watt, not transistor density.) Hyperscalers book six-year depreciation schedules, but that’s an accounting choice under pressure; Michael Burry has argued loudly the real economic life is closer to two or three years. Meanwhile the building that houses everything — concrete, power feed — lasts 15 to 30 years. Three clocks badly out of sync: chips obsolesce in a few years, the structure lasts decades.

On the ground this mismatch is boring and solved. When the chips age out, a technician wheels a cart down an aisle and swaps the cards. The expensive long-lived part — the building — stays put and amortizes over thirty years. You replace only the cheap, fast-moving part.

In orbit you cannot wheel a cart down an aisle. And this is where the entire economic question lives — but it’s subtler than “can you send a repair robot,” because there are two failure modes, not one.

The crude version: relaunch the whole satellite each refresh. If upgrading the chips means flying a fresh satellite every few years, then over a 30-year facility life you pay the full launch bill five to ten times. In this capex-only comparison, that only beats the ground if launch drops to roughly $25 to $76 per kilogram depending on refresh interval — below even the optimistic Starship projections. Possible in principle. But you’re betting the entire concept on launch prices nobody has demonstrated.

The sophisticated version: service in orbit. Suppose a robot can swap the GPU modules while the bus, radiator, solar array, and batteries all stay in place. This is the case the optimists are implicitly counting on, and it’s a real saving — you stop relaunching 30 tonnes of satellite. But it is not free, and here’s the trap the optimists miss: the replacement chips still have to reach orbit. You don’t relaunch the radiator and the solar panels, but you relaunch the compute payload — about 13 tonnes per megawatt — every single refresh cycle. And the compute is the part driving the refresh in the first place. Serviceability saves you the radiator-and-solar fraction. It cannot save you the compute fraction, because that’s the fraction that keeps going obsolete.

Run the corrected numbers. Best orbit, service in place, replacing 13 t/MW of compute every cycle, at a hypothetical $100/kg: a 3-year refresh costs $14.5M/MW (above the ground build), a 6-year refresh $8.1M/MW (below it), and at $300/kg the 3-year case balloons to $43.5M/MW.

The red line is the crude case — relaunch the whole satellite each refresh — and it sits far above the ground line everywhere worth discussing. The green lines relaunch only the chips: better, but the 3-year cycle still doesn’t beat the ground until launch falls near $80/kg. Serviceability helps — it stops you relaunching the radiator and solar — but it cannot stop you relaunching the compute, because the compute is the thing going obsolete.

Which is where it gets interesting, because I’ve been smuggling in an assumption that’s wrong, and correcting it is the most important move in this whole exercise.

You don’t throw old chips away. You demote them.

I assumed the orbital data center must refresh on the same clock as the ground — swap chips every three years to stay current. But that’s not how compute retires, on the ground or anywhere. A two-generations-old GPU isn’t garbage; it’s worth less. It still computes, just at lower performance per watt. So hyperscalers cascade hardware: newest silicon takes frontier training, last-gen drops to inference, older still to cheap batch work. Chips retire by demotion, not deletion.

So the orbital operator doesn’t have to match the ground’s cadence at all. It can run its silicon for years past the frontier, keeping aging chips busy on lighter, latency-tolerant work, and relaunch far less often. That breaks the punishing symmetry — and it’s a real win. Stretch the refresh from three years to nine and the recurring launch term collapses.

But — and conservation laws always have a but — old chips don’t compute as well, and here’s the precise asymmetry, because it’s the place a sharp reader will push hardest. Both the ground and orbit run aging chips; demotion happens everywhere. The difference is the cost of the power and space that an old chip occupies. On the ground, a demoted chip draws cheap grid power and sits in an already-built rack — the marginal cost of keeping it is basically its electricity. In orbit, that same old chip occupies solar array and radiator and structure that cost $100–300 a kilogram to lift and cannot be repurposed for anything else — so its true cost is the amortized launch of that fixed power budget, whether the chip running on it is frontier or four years stale. It’s an opportunity-cost argument, not a “space is worse at computing” one: orbit’s power is expensive and fixed, so filling it with obsolete silicon wastes capacity you paid a fortune to put there.

Here I also have to switch units, and I want to be explicit about it, because the switch quietly does real work. Up to now "per MW" has meant a megawatt of electrical IT load. From here it means a megawatt of *frontier-equivalent* compute, because an aging chip still draws its full orbital power but delivers less useful work relative to the current best silicon. Performance per watt improves over time (doubling every D=3D = 3 D=3 years, generously); your orbital power budget is fixed, sized once at launch. So a chip three doublings old — nine years, in this toy model — delivers maybe an eighth of the frontier compute per watt, on infrastructure you're still paying full freight to keep aloft. And here is the asymmetry I'm granting the ground for free, stated plainly so it isn't hidden: the orbital cost gets divided by this aging output fraction, but the ground's $11.3M does not, because cart-down-the-aisle refresh keeps terrestrial silicon at the frontier continuously. That's a real thumb on the scale — and it favors the ground, which is the direction I've been rounding throughout. The whole point is that orbit *can't* do the cheap continuous swap, so it's the side that pays the obsolescence tax.

(You might ask why the power budget is fixed — why not just launch more solar as the chips improve? You can. But that’s more launch mass every cycle, which is exactly the cost the whole optimization is trading against. Fixed power per deployed satellite is the right model; growth means more satellites, more launches, and the math below already captures that.)

Put both effects together and the cost per unit of actual frontier-equivalent compute stops being monotonic. Refresh too fast and you relaunch chips constantly. Refresh too slow and you’re paying orbital prices to run silicon that can barely keep up. In between, there’s an optimum:

In a smooth continuous-cycle approximation, with performance per watt doubling every three years, the optimal refresh lands around nine years — and if you force integer launches over a 30-year life (deployments at years 0, 10, 20), the minimum nudges to roughly ten years with these inputs. The exact year isn’t the point, and it isn’t even sensitive to my assuming a smooth exponential decay: I reran it with the chip value falling linearly, and with a lumpy generation-by-generation step function, and an interior optimum near a decade shows up every time. That’s because the structure is general — launch cost falls as you refresh less often, while useful output also falls as the fleet ages, and any value curve that decays toward zero relative to the frontier produces that tradeoff. The precise optimum depends on the shape; its existence and its rough magnitude do not. And the point that matters is robust: the orbital optimum is much longer than the ground’s frontier-chip cadence — three times longer or more.

This is also the answer to the sharpest objection a cost analyst will raise: you optimized launch, but at $100/kg launch is the minority cost — satellite hardware has never been built for $100 a kilogram, so manufacturing dominates, and you optimized the wrong line item. True about the dollars, and beside the point about the structure. Changing the launch price scales the entire cost curve multiplicatively (it looks like a vertical shift only because the chart’s axis is logarithmic); it does not move the optimum. And the same is true of any per-kilogram cost of putting refresh mass on orbit — manufacturing, integration, insurance, all of it. Fold a hardware cost-per-kilogram into the launch term and the curve rises, but its minimum stays put. The refresh optimum and its dependence on the pace of chip improvement survive whatever the true all-in cost of a kilogram in orbit turns out to be. Only the absolute dollar figures move; the shape of the argument, and its crux, do not. The optimum is set by the tension between the durable orbital infrastructure you keep and the rate at which lifted silicon falls behind the frontier — not by what a kilogram costs.

At that optimum, orbital compute at $100/kg costs about $14M per megawatt of frontier-equivalent capacity — just above the $11.3M ground build on launch alone. But the ground also pays a grid-electricity bill, and orbit does not — it pays up front instead, in solar, batteries, radiators, and launch mass: thirty years of grid power adds $13–31M/MW to the terrestrial side, putting the ground build-plus-electricity comparator in the $24–42M range. So on launch cost alone, orbit at its optimal refresh comes in below the ground’s build-plus-electricity band — but that gap is what has to pay for everything I’ve left out of the orbital side: spacecraft manufacturing, servicing operations, insurance, stationkeeping, radiation-hardening, networking, and disposal. Below the band on launch alone is what a real business case would need as its starting margin, not its conclusion. At $300/kg, the optimized launch-only figure is about $42M/MW frontier-equivalent — landing right at the top of the ground build-plus-electricity band before a dollar of those orbital costs is counted. That’s a warning light, not a win. (Worth noting that the most-cited independent estimate lands in the same place: aerospace engineer Andrew McCalip’s calculator, written up in IEEE Spectrum, puts a gigawatt-scale orbital facility plus five years of operation north of $50 billion against about $16 billion on the ground — roughly 3×. Different method, same neighborhood as the arithmetic here, which is reassuring: when a back-of-the-envelope and a detailed model agree, both are probably close. Tellingly, McCalip gets even that far only by assuming the orbital fleet runs efficient, radiation-tolerant chips rather than frontier GPUs — which is the demotion argument arriving from the other direction.)

You might also note that you needn’t relaunch the whole compute payload each cycle — swap only the highest-value boards and keep the trays and interconnect — which helps the orbital case further, by an amount that depends on packaging choices nobody has made yet. I’ve left that on the table, conservatively, for Musk.

(And the sun fraction, the thing I was careful to pick favorably? Second-order. A generic 60%-sunlight orbit makes the satellite a third heavier and barely moves any of this. The refresh optimum dominates.)

Where the crank stops

Neither side had it right, and the truth is more interesting than either.

The critics are wrong: the heat can be rejected, the radiator is 3.3 tonnes, and “you can’t cool it in space” — though space is indeed a worse radiator — is not the argument that decides anything. They reached a reasonable instinct through the wrong variable.

Musk is right about the parts he’s loud about: heat is manageable, a single launch is competitive at projected Starship prices, and — once you account for the fact that you demote old chips rather than discard them — there’s an optimal way to operate an orbital fleet that becomes launch-and-electricity competitive with the ground in this toy model, at the optimistic end of his own launch-cost projections. He is not being stupid. Run the numbers his way, generously, and there’s a real case in there — one that still has to cover the spacecraft and servicing costs the model leaves out, but a real case.

But it lives in a specific corner, and the corner is defined by two numbers that are neither thermal nor propulsive. The first is launch price: the case looks interesting near $100/kg and becomes uncomfortable by $300/kg — where the launch-only figure already reaches the top of the ground build-plus-electricity band before orbital hardware and servicing costs are counted — and SpaceX has not yet flown anything at either price. The second is subtler and, I think, the actual crux: the pace of semiconductor improvement. The whole optimization turns on how fast orbital chips fall behind the frontier. If performance per watt keeps doubling slowly — every four or five years, as the leading edge arguably trends — old chips stay useful longer, the optimal refresh stretches out, and orbit looks good. If improvement reaccelerates to the historical two-year pace, the optimum tightens, the relaunch term comes roaring back, and the ground wins comfortably.

So an orbital data center is not a cooling bet. It’s barely a launch bet. It is, underneath everything, a bet that Moore’s law stays slow — that the chips you lift won’t be embarrassed too quickly by the ones you didn’t. That’s the number to watch. Not the temperature of space, and not even the price of a rocket. The rate at which silicon improves.

One honest boundary on all of this: it’s a launch-and-infrastructure model, not a full lifecycle one. I use two ground comparators — a capex-only baseline of $11.3M/MW, and that same figure plus a 30-year electricity band of $13–31M/MW — and the orbital side counts launch and durable hardware mass but not spacecraft manufacturing, servicing operations, insurance, or disposal. So the electricity band helps the orbital case; the omitted orbital hardware and operations hurt it; the two pull opposite ways. I’m also excluding the data-movement problem entirely: orbital compute only makes sense for workloads whose inputs can be staged ahead of time and whose outputs are compact or latency-tolerant, because you are not getting a fat low-latency fiber to low Earth orbit. And I’m setting aside radiation: leading-edge accelerators aren’t built on radiation-hardened processes, so a real design pays for shielding mass or accepts higher failure and faster degradation — either of which makes the orbital case harder, not easier, and tightens the very refresh clock the argument turns on.

Two of my “durable” components aren’t, and both cut the same way. Batteries cycling through Earth’s shadow see thousands of charge-discharge cycles a year; even space-grade cells are spent in five to ten years, not thirty, and solar arrays lose a percent or few of output annually to radiation. So in any eclipsing orbit the batteries and part of the array join the compute on the recurring-relaunch side of the ledger — they’re wear items, not failures. That only deepens the conclusion: more recurring mass per cycle makes the refresh cadence matter even more.

And the largest caveat is one my arithmetic simply can’t price. This model assumes the ground can buy electricity and a grid connection on demand at $0.05–0.10/kWh. The actual bull case for orbital compute is that it can’t — that transformer lead times, interconnect queues, and permitting mean the binding ground constraint isn’t the price of power but the availability of it, in which case scarcity rents push the ground comparator upward by an amount no one can yet quantify. That may well be the real argument for building in orbit. But notice it’s a bet on terrestrial grid bottlenecks, not on physics or launch cost, and it lives outside what a back-of-the-envelope can settle. I’m pricing power as available; if it isn’t, the orbital case looks better, for reasons this calculation doesn’t capture.

But the structure survives all of these corrections, because the refresh optimum — and the chip-progress bet behind it — governs the recurring side of the ledger no matter what. If someone tells you orbital data centers obviously do or don’t work, ask them two things: what launch price, and how fast they think accelerators will keep improving. Everything else is the temperature of space.

The calculation, every input cited

All figures are per megawatt — of electrical IT load through the mass and launch sections, and of frontier-equivalent compute in the demotion section (an aging chip draws the same orbital power but does less useful work). This is a launch-and-infrastructure model, not a full lifecycle one. I use two ground comparators: a capex-only baseline of $11.3M/MW, and that same figure plus a 30-year electricity band of $13–31M/MW. The orbital side counts launch plus durable hardware mass but excludes spacecraft manufacturing, servicing operations, insurance, failures, and disposal. Chip purchase cost is excluded, not cancelled: it would cancel only in a same-refresh comparison, but here ground and orbit refresh on different cadences, so treat the result as launch-and-infrastructure, not a full chip-capex model. The exclusion isn't neutral, and it leans the generous way: at the orbital optimum you deploy roughly 2.4 MW of fresh silicon every nine years (the oversizing needed to offset the aging fleet), about 0.26 MW of new chips per year, versus the ground's 1 MW every three years, about 0.33 — so the ground actually buys more new silicon per year, and leaving chip purchase out of both sides mildly favors the ground.

• Stefan-Boltzmann constant σ — 5.67 × 10⁻⁸ W/m²K⁴ — physical constant.

• Radiator emissivity ε — 0.90 — high-emissivity coating (generous).

• Radiator temperature — 75 °C / 348 K — chip-junction-limited assumption.

• Radiator areal density — 5 kg/m², two-sided — aggressive lightweight-radiator assumption; many flight systems are heavier.

• → Radiator mass — 3.3 t/MW — computed; radiating area ~1,300 m²/MW, physical panel ~670 m².

• Compute mass (proxy) — 12.8 t/MW — NVIDIA DGX H100: 130.45 kg, 10.2 kW — terrestrial hardware, mass proxy only.

• Solar specific power — 150 W/kg (best-in-class) — NASA Small Spacecraft SOA (typical ~30); UltraFlex up to 150 W/kg EOL.

• Battery specific energy — 150 Wh/kg — NASA: commercial Li-ion 150–270 Wh/kg; space-grade trades energy for life.

• LEO orbital period — 1.5 h — standard.

• Structure / electronics — +30% — engineering allowance (my assumption).

• → Total satellite mass — ~30 t/MW (best orbit) to ~41 t/MW (generic) — computed.

• Ground build cost — $11.3M/MW (shell + core) — JLL 2026 Data Center Outlook; AI fit-out can add up to $25M/MW.

• Ground electricity (sensitivity band) — $13–31M/MW over 30 yr — computed at $0.05–0.10/kWh, PUE 1.0–1.2.

• Falcon 9 launch — ~$2,700/kg to LEO — Orbital Radar, May 2026.

• Starship (hypothetical, mature) — ~$100–300/kg projected — SpaceX publishes >100 t to orbit, no $/kg; $/kg are projections.

• GPU architecture cadence — new generation every 18–24 months — Introl.

• GPU economic life (contested) — 2–3 yr (Burry) vs 5–6 yr (hyperscaler books).

• Building life — 15–30 yr — three-clock mismatch.

• In-orbit servicing status — demonstrated in GEO (Northrop MEV undocking); not yet industrial (NASA cancelled OSAM-1).

The model. Sun fraction f sets two coupled masses: solar sizing scales as 1/f (oversize to charge through the dark side), battery mass as (1−f) (store energy for the eclipse). With m_refresh the compute mass relaunched each cycle and R the refresh interval in years over a 30-year life:

where L is launch dollars per kilogram, treated as delivered to the target LEO or sun-synchronous orbit, with inclination and orbit-insertion penalties ignored — another assumption that favors the orbital case. The crude case relaunches the whole satellite each cycle; the serviced case relaunches only the compute. The ground comparators are the $11.3M/MW build, optionally plus the $13–31M/MW electricity band. The serviced case is the relevant one, and it still carries the full compute mass on every refresh — which is why the refresh cadence, not the radiator, governs the result. (The effective-cost optimization in the demotion section divides this launched mass by the cycle-average frontier-equivalent output

which is what produces the ~9-year continuous, ~10-year integer optimum.)

For more conservation-law thinking — where a cost you thought you’d eliminated reappears somewhere else in the ledger — see The Conservation Law of Tax Alpha. My book, The Science of Free Will, asks an equally uncomfortable question about an equally cherished story.

In 2005, a very famous pod shop wanted me to come work for them. The recruiter was friendly. The compensation would have been excellent. The infrastructure on offer — capital, technology, operations, the entire institutional machinery a systematic trader would otherwise have to build for himself — was the kind of thing you don’t normally walk past twice.

Then they sent over the IP form.

It was black and white. Every rule I wrote, every model I built, every strategy I developed while in their employ would belong to them. The recruiter, to his credit, tried hard to talk me through it. There was nothing to talk through. The contract was clear, and it was non-negotiable. I said no and walked away.

Twenty years later, every knowledge worker on earth is about to be handed a version of that contract.

The form that landed on my desk in 2005 was unusual for its time — directed at the small number of people, like systematic traders, whose work product was legible enough to be copied. Discretionary traders didn’t get that contract, because nobody could actually extract what was in their heads. Lawyers didn’t get it, because their judgment didn’t fit on a page. Doctors didn’t get it for the same reason.

That’s about to change. Everyone in knowledge work is on the verge of producing exactly the kind of legible, portable, codified artifact that did land me that contract two decades ago. The agents are coming. The skills folders are coming. And the contracts are coming after them. This post is about why.

A different deal, same year

Later that year, I made a different deal.

Joe Ritchie was a legend. At one point in the 1980s, his firm Chicago Research and Trading was doing more options business than any other firm in the world. CRT pioneered the use of computers in options pricing — Joe famously programmed Black-Scholes into a Texas Instruments calculator before most of the floor knew what Black-Scholes was. He sold CRT to NationsBank in 1993 for $225 million. He came to trading by way of a Wheaton College philosophy degree, a stint as a Chicago bus driver, and a job as a guard at the Cook County Jail; he learned the markets from a book a friend handed him. Self-educated in the only sense that matters. Genius in every sense.

He was also a straight shooter — devout, principled, and scrupulously honest in business. Before we agreed on anything, I asked if my wife could come meet him. Joe said not only that she should — she must. A few weeks later, there we were, in his boardroom in suburban Illinois, my two kids coloring on the boardroom table, aged four and seven. He was happy to have the whole family. We shook hands on the deal shortly after.

I traded Joe’s prop capital for three years. We shared ideas freely; he gave me ideas, I gave him ideas, and neither of us laid claim to the other’s IP. There was no contract that swallowed what one of us brought to the table. We simply trusted each other not to do the thing that the pod shop had asked me to sign away in writing.

He was, and remains, the only person I have ever met who saw markets in exactly the way I did. We had different approaches and different personalities — every system has to match the personality of the person running it, which is a different post — but on every question that mattered, we agreed. What a rule is. Whether a rule should be traded. Whether a rule is real or robust. The value of systematization. The definition of systematization. The value of what Joe called “trader logic” over the academic frame of mind that produces, for instance, the Barron’s columns advising readers to use options to take directional positions. (You don’t. You trade options to bet on volatility. If you want to bet on direction, you bet on the underlying. That this is not obvious to a class of professional commentators is itself a useful piece of evidence about what “trader logic” is and isn’t.)

That kind of agreement is what makes a handshake deal possible. It is not enough to trust someone’s character. You also have to share their definition of the thing you are trusting them about. Two people who disagree about what a rule even is can sign every contract in the world and still end up suing each other. Two people who agree about what a rule is can shake hands and never need a contract. Joe and I agreed about what a rule was.

The contrast matters because it isolates the variable. Joe and I were doing exactly the work the pod shop wanted me for — developing systematic rules and applying them with capital. We were sharing IP in both directions, which the pod shop’s contract would have made formally impossible. The thing that was different wasn’t the work. It was the structure of the relationship. Joe was trading his own money. He had no reason to own the rules — only to use them. Pod shops are running other people’s money on behalf of investors who are paying for institutionalized, transferable IP. The contract follows from the business model.

Joe passed away during the pandemic. I still miss him.

What systematic traders knew first

The reason I bring up the pod shops isn’t aesthetic. I have specific, technical reasons for thinking they’re not very good at the thing they claim to be selling.

In November 2025 I published a paper in The Journal of Portfolio Management called “The False Promise of Drawdown Rules: New Evidence and a Better Framework.” The paper used ETF and long-short data from 1993 to 2022 to test the standard pod-shop “risk management” rule — the one where if you’re down X%, you’re cut. I showed that those rules, even in perfect-hindsight scenarios, turn certain profits into losses more often than not. They don’t manage risk. They manufacture it. Pod shops succeed in spite of their drawdown protocol, not because of it. The whole apparatus is incoherent on its own terms, and I think anyone who has worked under it long enough knows this.

There’s a more general lesson hiding in this finding. Externalized rules in the wrong hands actively destroy value. The pod shops’ own written-down risk rules make them worse traders, not better ones. If they can’t codify their own risk management without manufacturing losses, what reason is there to think they’d handle a codified version of someone else’s IP any more competently? The technical problem with their risk protocol and the structural problem with their IP grab are the same problem manifested twice — an institution organized around the wrong theory of where the value lives.

So when I tell you I refused to sign their IP form, understand that I had also looked carefully at the rest of what they were offering and concluded it wasn’t worth giving up the rules for. The contract was the most visible problem, but the building behind it had structural issues, too.

But the IP problem was the deepest one, and it’s the one that’s now generalizing.

Here’s what systematic traders knew first, and what’s now becoming a fact about every white-collar field. Trading IP, when it lives in someone’s head as discretionary judgment, is essentially impossible to extract. The pattern recognition of an experienced market-maker, the gut-feel of a macro trader who has lived through three crises, the spatial intuition of a vol trader who has watched skew misbehave in real time — none of that fits on a page. You can’t put it in a contract because there’s nothing to put in.

But the moment you write down explicit rules, all of that changes. Now there’s a thing to take. There’s a piece of paper. There’s a file. There’s a contract that says it’s mine.

The current decade has produced a tidy little series of legal cases that show exactly how this plays out, and the trajectory is escalating, not stabilizing.

In 2019, Citadel Securities sued the British hedge fund GSA Capital over an automated trading model called the “ABC Strategy.” Citadel said the model had cost more than $100 million to develop and was so closely held that just 15 of the firm’s roughly 3,000 employees had access to its strategic logic. The dispute was triggered by GSA’s attempt to recruit a Citadel quant. Citadel argued in court that GSA “couldn’t unsee” what its hire had brought with him. The case settled confidentially in 2021. Whatever the precise terms, the precedent was that you could litigate over the contents of a trader’s head if those contents had once been written down.

In April 2024, Jane Street sued Millennium Management and two former traders, Doug Schadewald and Daniel Spottiswood, over an India-options strategy that had generated about a billion dollars in profits for Jane Street in 2023 alone. Jane Street alleged that the strategy’s profits dropped by half in March 2024, the month Millennium reportedly began using it. The two traders, in their defense, argued that the strategy wasn’t a trade secret at all — that it was based on their experience and expertise rather than any protected algorithm or signal. That defense is the entire post in one sentence. Where does the trader end and the strategy begin? When the rules are explicit, they belong to the firm. When they’re implicit, they belong to the trader. The case settled in December 2024, but only after Jane Street had been forced to disclose sensitive details in court — a small demonstration of what employment lawyers now call the “enforcement paradox”: the harder you fight to protect a trade secret, the more of it you have to reveal in court.

Earlier this year — February 2026 — Citadel went after Daniel Shatz, a former portfolio manager who took global credit at Marshall Wace. Citadel alleged that Shatz had accessed proprietary documents — recruiting lists, strategic roadmaps, candidate evaluations — while still employed and during a fifteen-month garden leave that preceded his formal departure. Fifteen months. Pause and consider that. The firm was paying him to do nothing for over a year because they were afraid of what he would do if he were free to act sooner. The IP problem has become so acute that the cost of avoiding it now includes paying senior people to sit on their hands for over a year.

The arc here is the part that should worry the rest of you. In 2009, the IP fights in finance were over code and information (Sergey Aleynikov leaving Goldman for Teza, Citadel suing Teza’s founders). In 2024, they were over strategies. In 2026, they are over the garden-leave intervals between strategies. The thing being protected is getting more abstract, the protection mechanisms more elaborate, the legal stakes higher. And until now, this has been a peculiar little drama playing out in finance, watched by the rest of the world with a mixture of curiosity and contempt.

It’s about to stop being peculiar.

The smartphone bit, briefly

The cliché version of what I’m about to argue is the smartphone analogy: by 2027 everyone will have AI agents the way everyone has a phone now. That’s true and it’s trivial. The same model will be available to everyone. The same chat interface, the same general capabilities, the same default toolkits. There will be no scarcity in the thing.

The smartphone analogy nails one thing: the social-expectation treadmill. Right now, “I’ll get back to you next week” is normal. In two years, that response will sound the way “let me check my answering machine” sounds now. Why didn’t your agents have it ready Tuesday? You’ll get used to that question, the way an entire generation got used to “why didn’t you text back.”

But the smartphone analogy fails on the part that actually matters. Phones are isomorphic across users. Yours and mine do the same things and produce the same value for both of us. Agents won’t work that way. The capability of an agent depends entirely on what you’ve taught it to do for you, in your domain, on your problems. The tool is generic. The configuration is everything.

That configuration lives in a folder.

What a skills folder is — without showing you mine

A skills folder is the file system in which an expert tells an agent how their domain actually works. It contains documents, written in plain language, that codify procedures, conventions, decision rules, common errors, edge cases, and the specific corrections that turn the agent’s enthusiastic competence into actual usefulness. It is not a collection of one-off prompts. It is a permanent, evolving body of instruction.

I have skills for backtesting trading strategies, for generating code against my own specifications, for supporting my physics research, for building and querying my reference library, for editing prose. The publishing piece — the part that helps me write Substack posts — is the smallest tail of what’s there. The rest is the real work.

I’m not going to show you any of those skills. That’s the most important sentence in this post.

I’ll tell you what category of failure each of them addresses, because that’s enough to understand what’s at stake. The trading-strategy skills exist because agents will, by default, write you a backtest with subtle look-ahead bias and report it confidently. They will use future information to predict the past. Every quant who has ever shipped a paper-trading system before discovering this knows the feeling. A skill, written carefully, catches it. The physics research skills exist because agents will, by default, get computational details in particular calculations exactly wrong in ways that look right to a non-physicist. The code generation skills exist because agents will, by default, ignore conventions specific to the codebase they’re working in unless you write down what those conventions are. The reference-library skills exist because agents will, by default, hallucinate citations.

The general principle: a skill is a codified record of where a generic agent fails in a specific expert domain, plus the correction. It is the externalization of tacit knowledge. It is the thing that takes me from being one person with an agent — which is what everyone will have — to being one person with a configured agent that knows how a particular kind of work actually has to be done.

But here’s the thing the skills also are: they are the sort of artifact that can be lifted. They are exactly the kind of thing the pod shop’s contract would have said belonged to them. They are the same form, in a different domain, of the artifact I refused to externalize in 2005.

This is why I’m not going to show you mine.

It’s also why this post is, in a small way, an enactment of the argument it’s making. I’ve told you the moat exists. I’ve told you what shape it has. I’ve explicitly refused to walk you through the contents. That refusal is the model, and the first thing anyone with a working skills folder should understand is that nobody who has one is going to show it to you.

The mechanism

The reason a skills folder is uniquely valuable — uniquely yours, structurally — has nothing to do with the cleverness of any individual skill. It has to do with who can write one in the first place.

A skill cannot be written by a non-expert, because writing one requires knowing which agent failures matter and which are noise. The trading skill that catches look-ahead bias is only writeable by someone who has shipped a system that contained look-ahead bias and learned the hard way to recognize it. The physics skill is only writeable by someone who knows that a particular calculation has a sign error in textbooks A through D and a different sign error in textbook E. The legal skill is only writeable by someone who has personally been on the receiving end of a discovery review that went sideways.

The skill is, in this precise sense, a record of accumulated scar tissue. The agent provides the muscle. The expert provides the scars. You can give an agent to anyone, but you cannot give scars to anyone, which is why the folder is the real asset.

A skill also cannot be written by the agent itself. The agent doesn’t know what it doesn’t know. The whole reason a skill is needed is that the agent’s confidence and the agent’s correctness diverge in places only an expert can see. Asking the agent to write its own skills is exactly the same epistemological category as asking a freshman pre-med to write the residency-training curriculum. There’s no path from inside.

Tacit knowledge, in the philosophy-of-science literature, was always understood as the part of expertise that couldn’t be verbalized — the irreducible human residue. What’s happening now is that, for the first time, we have a tool — the agent — that creates pressure to verbalize as much of it as possible. And the people doing the verbalizing are the experts themselves, in their own voices, in their own folders.

This is a good thing, in the abstract. It is the most efficient knowledge-transmission mechanism ever invented. It is also the conservation law that defines the new labor market.

In knowledge work, productivity gains and IP defensibility are now conjugate variables. You can have one or the other at full strength. You cannot have both. This is the systematic trader’s old dilemma, and it now belongs to everyone.

There is exactly one way out, and we’ll come to it. Everyone who doesn’t take it lives with the dilemma.

The trap

Here is what’s about to happen — and is already happening — in fields that have nothing to do with finance.

A junior transactional lawyer at a big firm — the office lawyer, not the litigator who’ll be in court — builds a 200-line skill that captures her firm’s contract-review process. The specific clauses to flag in M&A diligence, the formatting conventions, the standard redlines, the way to summarize a target company’s liabilities for a partner’s memo. The skill saves her 20 hours a week. Her billables look extraordinary. Her partners notice.

Six months later, she gets an offer from a competitor. Her current firm sends her a letter reminding her of the IP clauses in her employment agreement. The skills she wrote, the firm asserts, were written on firm time, using firm matters, under firm supervision, and constitute work product belonging to the firm. She is welcome to leave, but the folder stays.

A radiologist builds a skill that encodes her department’s reporting conventions — the phrasing, the diagnostic priors that fit the patient population, the specific failure modes of the imaging equipment that produces 80% of her cases. The skill makes her twice as fast and improves her accuracy. The hospital, which paid for her training, the equipment, and the patient relationships, has a view about who that skill belongs to.

A management consultant builds a skill that captures the way her firm sells engagements — the diagnostic frameworks, the standard slide structures, the negotiation patterns that work and don’t. She is, in a real sense, more productive than any consultant before her. The firm has thoughts about whether she gets to take the folder when she leaves.

Each of these people has just produced, in 2026 or 2027, exactly the kind of artifact I was asked to produce inside the pod shop in 2005. And each of them is about to discover that the law of intellectual property does not care about the medium. A trade secret is a trade secret whether it’s an options-pricing algorithm or a discovery-review checklist. The contract clauses that protect employer ownership of work product are not going to politely exempt skills folders from their reach. The Jane Street–Millennium case turned, partly, on whether explicit trading rules counted as a protected trade secret. The next decade’s version of that case will turn on whether explicit legal rules, medicalrules, consulting rules count.

They will. The legal infrastructure for this fight already exists. The Defend Trade Secrets Act, the body of state-level non-compete law, the standard work-product clauses in every white-collar employment agreement — none of this was designed with skills folders in mind, but all of it applies. The asset is portable, the asset is valuable, the asset was created during employment using the employer’s resources. The fight writes itself.

Not every white-collar role is equally exposed. Therapists, salespeople, executives whose work is primarily relational — and litigators, whose work is performed live in adversarial courtrooms — produce less codifiable output than a transactional lawyer or a radiologist or a quant. The trap applies to the codifiable portion of knowledge work — a large and growing share, not a total one. One implication worth naming: the discretionary work that has historically been the higher-prestige half of several professions — the trial lawyer over the office lawyer, the master clinician over the junior diagnostician, the partner running the room over the associate running the spreadsheet — is about to become more valuable, not less. The trap closes hardest on the work that can be written down. Work that can’t be written down stays safe.

And whether the firms will win those fights even within the codifiable share is itself a live question. The Jane Street defense — strategy as experience and expertise, not protected algorithm — is exactly the argument workers in every codifiable field will start making. If the courts buy it, the trap stays porous. If they don’t, it closes hard. We don’t yet know which way the law lands. What we know is that the fight is coming.

And here is what makes this particularly cruel: the better you are at your job, the worse this trap is for you. The most productive lawyer at the firm has the most valuable folder. The most productive radiologist has the most valuable folder. The most productive consultant has the most valuable folder. The structural pressure of agent-augmented work concentrates value into a portable artifact, which means the workers most worth fighting over are also the ones most exposed to the contract. The pod shops always knew this. It’s why their contracts were drafted the way they were. Now everyone gets to learn.

Three positions

When the trap closes, the labor market stratifies into three positions, each with real costs.

The first position is the W2 employee on the sanctioned firm stack. The firm provides the agents, the skills folder lives on the firm’s drives, the IP clauses are tight, and the productivity is real but rented. You are more productive than the workers of 2024, but the productivity is captured by your employer. Most large institutions will end up here for the same reasons they ended up with company-issued laptops and company-managed email: compliance, audit trails, IP protection. This is not a bad position. It is just not a position with a moat.

The second position is the 1099 contractor, the consultant, the solo practitioner with their own stack. Your skills, your folder, your IP, your risk. You eat what you kill, but you also keep what you build. This was not the systematic trader’s typical path. The typical path was the first position — at firms whose contracts captured everything the trader learned. The IP cases I described earlier are exactly what the typical path produces on exit. What’s new is that fields where the W2 career was the stable default — law, medicine, design, research, consulting — are about to face the same fork systematic traders faced, and a higher fraction will take the solo path: partly because the IP grab pushes them out, and partly because the agent makes solo work newly scalable.

The third position is the loophole — the one exit from the conjugate-variables trap I flagged earlier. It’s the one I took, and it’s still available. Externalize the rules — but only for yourself. Build the folder, refine it, depend on it, and refuse to transfer it. The rules live in writing; they just live in writing you own. This is what walking away from the pod shop’s IP form actually meant in 2005, and it’s what it has meant since. The skills exist. They simply don’t belong to anyone other than me.

This stance used to be expensive because solo work didn’t scale. The cost of keeping the folder yours was approximately the cost of staying small. Agents change that math. A solo practitioner with a mature private folder can now do work that previously required a team — without ever turning the folder into something an employer can capture. The third position used to be a niche. It’s about to be a real option.

What you still give up: exit-by-acquisition, because acquisition requires the folder you refuse to hand over. Formal institutionalization, because nothing has been formally transferred to anyone. The ability to train juniors using your own materials, except in arrangements where you can actually enforce confidentiality — which most arrangements aren’t. And the appearance of belonging anywhere obvious — there’s no firm name on the business card. Many will look at that trade and decline it. That’s part of the story, not a weakness in it.

If you want a single image for the new labor market: the LinkedIn profile of 2028 reads “Samir Varma + 14 agents. References available.” And the agents have references too. The W2 worker’s profile reads “Samir Varma at FirmName. Agents licensed.” And the third worker’s profile, the one who built the folder and kept it private, just reads “Samir Varma” — because there’s nothing on offer. The folder is fine. It’s just not for sale.

The strongest counter, killed

The strongest objection to all of this is what I think of as the frontier-eats-the-stack worry. If the next-generation model is sufficiently smarter than the current one, your carefully written skills folder becomes obsolete. The plumbing-level skills — the ones that compensate for current-generation failures — get absorbed into the model itself. The moat is a mirage that recedes as the frontier advances.

Some of this is true. Current skills folders contain a lot of brittle, model-specific scaffolding that will indeed be obsolete in two years. Anyone writing skills now should expect a real maintenance burden as models change. There is no permanent skill, the way there is no permanent quant strategy.

But the structural argument survives. The skill is not a static asset; it is a continuously updated record of where this generation’s agents fail. As models improve, the failures shift, but they don’t disappear. The agent that no longer makes look-ahead errors in backtests will start making subtler errors in regime detection. The agent that no longer hallucinates citations will start misattributing them. The agent that no longer ignores codebase conventions will start over-applying conventions where they don’t fit. The work renews. It does not finish. And the orchestration and taste layer — the part of the skill that decides which failures are worth catching, which outputs are worth using, which corrections are worth codifying — does not depreciate at all. That layer is where the expert lives. It’s irreducibly yours.

To make the distinction concrete: the skill I wrote that catches look-ahead bias in backtests will become obsolete when models stop making look-ahead errors. The judgment that look-ahead bias was the failure mode worth writing a skill for does not. The first kind of skill is plumbing. The second is expertise.

The frontier eats the plumbing. It does not eat the expertise.

I won’t show you mine

Here is what I have told you. There is a thing called a skills folder. It is the externalization of expert tacit knowledge into a portable, configurable artifact. It is what makes one human-plus-agent more valuable than another human-plus-agent. It is also exactly the kind of artifact that creates the IP problem I walked away from in 2005, and that finance has been litigating over for the last decade and a half.

Here is what I am not going to tell you. Anything specific about what’s in mine.

The instinct to refuse, in 2026, looks like a personality trait. In two years it’s going to look like a labor-market strategy, and in five it’s going to look like the only sane labor-market strategy for anyone who built their folder themselves. The systematic traders saw this first. The lawyers and the doctors and the consultants are about to see it. There are excellent essays on the internet teaching you how to write skills, and the people writing them are doing good work. But look at who isn’t among them. There are no systematic quantitative traders out there walking you through their skills folders. There are no senior options market-makers publishing their factor libraries on Medium. The people who learned twenty years ago what externalized IP costs them are not the people teaching skills today. That absence is itself the lesson.

The first thing you are going to have to learn — same as I did, two decades ago — is that nobody who has one is going to show it to you.

You will have to figure it out yourself.

That, too, is the point.

If you found this essay worth your time, you may enjoy my book The Science of Free Will, which asks an equally uncomfortable question about an equally cherished story.

If you want to know more about India, start with The Paradox of India, the first essay — or browse the whole series.

Related country-specific work: Albion — Britain’s institutional decline, by someone who first saw the place in 1979.

Matt Levine has an intuitive model in Money Stuff. Bitcoin, he correctly keeps observing, has no cash flows. There’s nothing to discount, no earnings to capitalize, no rent to collect. The price is just… whatever people are willing to pay for it: it’s monetized attention. I personally find this hilarious and the joke seemingly writes itself roughly once a week with a different setup.

I’ve been reading Money Stuff for years, and at some point the joke stopped feeling like a joke and started feeling like a research design. Because here is the thing about “no cash flows”: if there really are no cash flows, then you cannot model Bitcoin’s price by capitalizing future cash flows. So either the price is unmodelable, or you have to model it using something else entirely. And the something else, when you sit down and think about what’s actually generating prices in a continuous limit-order market, is: arrivals and exits. Buyers showing up. Sellers showing up. Pressure accumulating. Pressure dissipating.

That is a queueing problem.

Which is convenient, because a Danish telephone engineer named A.K. Erlang worked out the math in 1909.

Erlang was trying to figure out how many telephone lines the Copenhagen telephone exchange needed. Calls arrive at random times. Each call lasts for some random duration. How many simultaneous calls do you have to be prepared to handle? Erlang’s answer turned into an entire branch of applied probability called queueing theory, and the workhorse model — the M/G/∞ queue — is exactly what you want for Bitcoin. M for Markovian (Poisson) arrivals. G for general holding-time distributions. ∞ because there’s no capacity limit; you can always have more pressure show up.

The mapping is direct. Replace “phone calls” with “latent directional pressure on the order book.” A long arrives — someone is leaning to buy — holds for some random time, then exits. A short does the same on the other side. The price moves with the imbalance. That’s it. No discount rate, no terminal value, no narrative about institutional adoption. Just arrivals, holding times, and exits.

(One clarification before any quant readers throw their coffee at the screen: these are not literal queues of customers, and they are not the visible limit-order-book queues. They are latent pressure states — the venue’s hidden buy-side and sell-side lean — inferred from trade prints under the standard M/G/∞ bookkeeping. The paper is explicit about that.)

(I’m aware this sounds suspicious. It sounded suspicious to me too, which is why the rest of the paper is 41 pages of seeing whether it actually works.)

The new paper, just out in the Journal of Risk and Financial Management, does exactly this for Binance spot BTC/USDT from 2020 through 2025 — five and a half years of the world’s deepest crypto venue, spanning the 2021 bull run, the FTX collapse, the ETF launch, and everything in between. The paper treats the latent buy and sell pressure as two independent M/G/∞ queues and asks what observable diagnostics fall out.

Three things do.

Variance per traded BTC. Call it R. It tells you how much price variance the venue is delivering per unit of trading activity. When R rises, you’re getting more volatility per BTC traded — meaning liquidity is thinning. It’s a microstructure tape-reader, in one number.

Effective mean-reversion rate. Call it θ. It tells you how fast directional pressure unwinds. Translate it into a half-life and you have a concrete answer to the question: after a shock, how long does it take for half the price displacement to bleed off?

A signed-flow proxy. Are buyers chasing recent up-moves? Are sellers leaning into down-moves? Is the flow contrarian or momentum-following? One coefficient gives you the sign and rough magnitude.

These three diagnostics share one bookkeeping system — they live on the same stock-flow accounting — which is the whole point of organizing the problem as a queue. Reduced-form market microstructure measures (Amihud illiquidity, Kyle-style impact) are essentially separate alarms wired to different bells. The queueing organization gives you a dashboard.

Now to the numbers.

Bitcoin’s resiliency half-life is roughly 10 to 25 hours.

Read that twice. Not seconds. Not minutes. Hours — and not few hours. On the deepest crypto venue on the planet, the half-life of price recovery from a directional shock is comparable to the time it takes to fly from New York to Tokyo.

For a venue whose marketing pitch is essentially “instant global liquidity, 24/7,” this is striking. Bitcoin doesn’t bounce back so much as slouch back. (The same diagnostic on a major equity index would be measured in minutes at most. The recovery profile here looks more like a thinly traded small-cap than a benchmark asset.)

Worse: when I run the same diagnostic on rolling 30-day windows across the five-and-a-half-year sample, only about 39% of windows are clearly mean-reverting under a modest statistical buffer. Roughly half are statistically ambiguous, and the remaining 12% have a point estimate that’s outright non-positive — meaning the diagnostic cannot even rule out a momentum regime. The venue moves between regimes; resiliency is not a venue constant.

This is exactly the picture you would expect from an asset that has no fundamentals to anchor it. Equity prices that get dislocated have an earnings stream to drag them back. Bitcoin’s price has order flow to drag it back, and order flow can decide to do anything it wants.

What about the obvious objection: “you’ve just renamed familiar liquidity measures.”

I expected this objection, so I spent a section addressing it. The standard trade-based liquidity benchmarks are Amihud illiquidity (price impact per dollar of volume) and a Kyle-style minute impact slope. I ran all three on the same 30-day rolling design over the same days and asked which one best sorts next-day tail risk — the probability that tomorrow’s largest hourly move lands in the top decile of the sample.

The queueing diagnostic separates next-day tail days at a high-minus-low spread of 8.6 percentage points. Amihud gets 4.7. The Kyle-style benchmark gets 1.1. The queueing-organized R is positively correlated with both — it isn’t unrelated to them — but it is clearly not the same number, and it sorts forward-looking tail risk roughly twice as sharply as Amihud and nearly eight times as sharply as Kyle.

I think this is because R is anchored in a coherent stock-flow accounting rather than a reduced-form ratio. The same M/G/∞ scaffolding that produces R also produces the resiliency and signed-flow diagnostics, and they all read off the same latent occupancy state. That’s the queueing tax. The dividend is that the diagnostics talk to each other.

(A small parametric overlay handles the heavy tails, because of course the symmetric core cannot fit Bitcoin’s 47-times-Gaussian excess kurtosis. The overlay improves the 99% Value-at-Risk breach rate from 1.80% under a rolling Gaussian to 1.13%. It still does not quite hit the 1% target — Bitcoin’s tails are a separate war — but it is a useful reduction.)

What this implies for normal markets

The cash-flow anchor in an equity is doing real work, and the half-life I measured in Bitcoin gives us a way to quantify how much. For a major equity index, the half-life of price recovery from a directional shock is measured in minutes — because every minute the price stays dislocated, there is an arbitrageable cash-flow value being violated. For Bitcoin, the only thing dragging the price back is more order flow, which is slow. That is why the half-life is 10-to-25 hours instead of single-digit minutes.

Read this as a quantitative measurement of how strongly fundamentals are anchoring a price. The half-life is the answer.

There is an obvious follow-on. In periods when fundamentals are not well-defined for an equity — earnings blackouts, news vacuums, crisis episodes when cash-flow estimates lose their informational value — the equity should start to look more like Bitcoin in its recovery profile. The framework predicts that. I do not have the cross-asset tick data sitting on my desk to run the experiment, but it is the obvious next paper. Anyone with the data, please go.

If you trade Bitcoin

Three things worth thinking about, with the usual caveat that these are diagnostics, not strategies. Use them to size; do not use them to enter.

First, rolling R is a warning system. When the variance-per-BTC moment is in its top third, your next-day probability of a top-decile hourly move is roughly 2.3 times what it is in the bottom third (15.1% versus 6.5% in the sample). When R is elevated, size down. If you market-make, widen.

Second, mean-reversion strategies are a regime trade, not a venue trade. Only ~39% of rolling windows are clearly mean-reverting; another ~12% are clearly not. A flat reversion strategy across all regimes is going to feed your edge to the periods when the model is wrong about the venue. The signed-flow proxy is the regime indicator: when it is high and θ is weak, you are in a momentum window. Do not fade it.

Third — the execution one — the 10-to-25-hour half-life tells you that if you move the price with a large directional order, do not expect the venue to “absorb” it in minutes. You are paying real impact for hours. Plan child orders, hedges, and stop placement accordingly. The 24/7 marketing slogan is true about the clock; it is not true about the absorption.

So: Matt’s model is also a research design. Bitcoin has no cash flows; therefore you cannot model it by discounting cash flows; therefore the entire price-discovery process has to live somewhere else; and the somewhere else turns out to be a queueing system that a Danish telephone engineer wrote down 117 years ago for a piece of analog infrastructure that no longer exists in the form he was studying.

I sent Matt a thank-you note when the paper came out. Intellectual debt to Money Stuff. A confession, while we are here. I personally find the entire edifice hilarious. Bitcoin has no cash flows, no terminal value, no rent — nothing but coordinated belief that the next person will also coordinate. I cannot quite believe this rubbish has value. But I am a libertarian: if people want to trade something with no fundamentals, they should be allowed to. That is what markets are for. I will not stop you. I will, however, do the math on your order flow, because the order flow is real even when the asset is absurd. The math doesn’t care whether you find the substrate ridiculous. It cares about the flow.

This is, I think, the patron-saint problem for finance writing done well. Matt’s job is to make sense of all of finance in several thousand words a day, every weekday, for years. He cannot stop and write the technical paper every time he notices the math underneath. He doesn’t have to. The observation does the work of identifying where the technical paper should be. Then it’s somebody else’s job to actually go write it. I am always going to try to be that somebody else, where I can. The physics PhD was supposed to be the boring substrate I left behind for three decades of trading. It turns out the substrate had things to say.

The paper is open access, free to read, and the Python that reproduces every number in it is on Zenodo. If you want to fight about it — and the Venn diagram of “people who hold strong priors on Bitcoin” and “people who hold them loosely” is a famously empty intersection — please at least bring the math.

Paper: https://doi.org/10.3390/jrfm19050372

If you found this essay worth your time, you may enjoy my book The Science of Free Will, which asks an equally uncomfortable question about an equally cherished story.

If you want to know more about India, start with The Paradox of India, the first essay — or browse the whole series.

Related country-specific work: Albion — Britain’s institutional decline, by someone who first saw the place in 1979.

A footnote on the timing of this post. The Preprints.org editorial team is celebrating their 10th anniversary with a #MyMostMemorablePreprint campaign, and Ellie Gao there nudged me to share the story behind a preprint. This is that story. The fact that the most memorable preprint (and now published paper) I have written started with a Matt Levine observation about Bitcoin having no cash flows is, I think, exactly the kind of origin story they were hoping to hear. #MyMostMemorablePreprint #PreprintsOrg10

I. The Building Floor

My former business partner invested in China.

His Chinese partner took the money. He protested. His Chinese partner pointed out the window.

“See that building? I built it with your money. The top floor can be yours.”

My partner, gently, said that wasn’t really why he had wired the funds. The Chinese entrepreneur was unmoved.

“Yes, and so?”

So my partner did what any grown adult in possession of a contract and a few good lawyers would do: he sued. In a Chinese court. And here is the part I want you to sit with for a moment: the court agreed that all the facts were in his favor. Every one of them. The court congratulated him on the quality of his case. And then the court left the money with the Chinese entrepreneur anyway, on the theory that the Chinese government did not want the capital flowing out of the country.

So: the facts were yours. The verdict was, in some sense, yours. The money was not yours.

You got a floor. In a building. In a city you didn’t live in.

My partner has since passed away. He died knowing he had been cheated. He died knowing the legal system had agreed he had been cheated. He died knowing that none of that mattered, that nothing was ever going to come of it, and that there was no further recourse, in this life, against the people who had taken his money. That is a very bitter pill to swallow at the end. It is the kind of bitter pill that I want any reader of this post to take seriously, before we get to the part where we laugh at hedge fund managers. Because everything that follows in this post — the data, the comedy, the booster-class roast, the civilizational thesis — sits on top of a small number of specific individual human beings who lost specific individual sums of money and never got them back, and who, in some cases, died waiting.

I have heard a lot of stories in my life, but this one has stayed with me, because it perfectly compresses something that took American capital roughly thirty years and several trillion dollars to discover: in China, the courts can rule for you and still rule against you. You will be told you are correct. You will be congratulated on the elegance of your arguments. And then you will be invited to look at the lovely building. And then, if you are unlucky, you will die before anything else happens.

For most of my adult life, the smart money on Wall Street believed the smart play was China. India was viewed as the slow uncle at the wedding — well-meaning, voluble, eventually charming, but somebody you certainly weren’t going to entrust with serious cash. China was the cool young cousin with the MBA and the gym membership. China was the future.

I have, for thirty years, told anyone who would listen one simple piece of investment advice: invest anywhere you want, but not in China. I was, for most of those thirty years, considered a crank. Slightly cracked. Probably resentful of Chinese success for ethnic-Indian reasons that I had not fully examined.

So let us now examine the people who were not considered cranks. Let us examine the smart money.

II. The Numbers Behind the Mistake

The data, before the comedy.

At the end of 2023, the US Bureau of Economic Analysis reported that the American direct-investment position in China stood at $126.9 billion. The corresponding figure for India: $49.6 billion. A ratio of 2.6 to 1. American capital had committed roughly two and a half times as much to the country actively designed to keep capital from leaving as it had to the country whose laws permitted capital to come and go through the front door.

This made no sense even on the simplest metric, return on capital. US-owned affiliates in India earned $7.9 billion in 2023, against $11.3 billion in China — meaning India produced about seventy percent of China’s income on thirty-nine percent of the invested stock. In 2023, and on the available evidence for a good while before it, India was simply the better trade per dollar deployed. Not by a little. By a lot.

Then, in 2023, the arrow finally turned. In the BEA data, new US direct-investment financial flows into India ($6.0 billion) exceeded new flows into China ($5.1 billion) for the first time. Separately, on China's own balance-of-payments measure, inward direct investment went negative in the third quarter of 2023 — minus $11.8 billion — the first negative quarterly reading in roughly twenty-five years of that series. Different official datasets measure FDI differently, and they do not always agree on the magnitude; China's Ministry of Commerce reports a higher "utilized FDI" number than the balance-of-payments series does. But the direction of travel is no longer ambiguous on any of them. The China premium is becoming an exit discount. India, meanwhile, crossed a cumulative trillion dollars in inward FDI in October 2024, and reported roughly $81 billion in inflows for the 2024-25 fiscal year.

Capital is supposed to be efficient. Capital is supposed to know things. Capital is supposed to vote with its feet. So what does it mean when the smartest capital on earth spent twenty-five years voting with its feet into a country whose system, on inspection, was specifically designed to prevent capital from ever voting with its feet again?

It means, technically speaking, that the smart money was full of shit.

Let us examine some of the relevant shit.

III. The Booster Class

The roll call of public China bulls over the past two decades reads, in retrospect, like a tour through the wing of an asylum reserved for people who got everything wrong but were paid extremely well to be wrong in expensive suits.

Jim Rogers. Co-founder of the Quantum Fund with George Soros. One of the most celebrated investors of his generation. In 2007, he moved his family from New York to Singapore specifically so his daughters would grow up speaking Mandarin — the air pollution in Beijing and Shanghai having put those cities out of contention as the actual destination. In the same year, he published A Bull in China: Investing Profitably in the World’s Greatest Market. I am not making the title up. There is also a sequel of sorts called Street Smarts, in which the bullishness continues unabated.

On India, Rogers was less effusive. In 2001:

“India as we know it will not survive another 30 or 40 years.”

In 2015, exiting the Indian market, he offered this elegant farewell:

“One cannot invest based on hope.”

On a CNBC interview in 2008, Rogers explained that he had moved to Singapore over Shanghai only because of pollution. Not because of the one-party state. Not because of the capital controls. Not because of the legal system that, as we shall see, regards foreign investors as a kind of voluntary tribute. The dictatorship was fine. The smog was unacceptable.

To his credit, by April 2024 Rogers had finally pivoted. He is now bullish on India. He cites Modi. He cites the convertibility issue. He is at long last on board with what some of us have been arguing since 1991. Welcome aboard, Jim. The party started thirty-three years ago. We saved you a chair.

Mark Mobius. Calls himself “the Indiana Jones of Emerging Market investing,” which is the kind of self-description that should disqualify a person from managing their own checking account, let alone other people’s. Ran Franklin Templeton’s emerging-markets desk for thirty years. Was, in his own publicity material, known for his bullish views on China.

Then, in March 2023, on Fox Business, Mobius reported the following discovery:

“I have an account with HSBC in Shanghai. I can’t take my money out. The government is restricting flow of money out of the country.”

This sentence is so good I want to needlepoint it onto a throw pillow. Notice the genuine surprise. I can’t take my money out. Mark Mobius has just been informed, in his late eighties, after a forty-year career as a celebrated emerging-markets specialist, that the People’s Republic of China — a one-party communist state that has imposed capital controls since approximately the Truman administration — is restricting the flow of money out of the country.

He went on:

“I can’t get an explanation of why they’re doing this... They’re putting all kinds of barriers. They don’t say: ‘No, you can’t get your money out.’ But they say: ‘Give us all the records from twenty years of how you made this money.’ This is crazy.”

This is crazy. Yes, Mark. This is crazy. This has been crazy for thirty years. The thing you wrote books about and earned management fees on for three decades was exactly this kind of crazy the entire time.

And then, having discovered — like Columbus, except with substantially less excuse — that the country was not arranged the way he had been telling everyone it was arranged, Mobius announced his new favorite emerging market:

“India... You’ve got a billion people, they can do the same thing that the Chinese do.”

Welcome aboard, Mark. We saved you Jim’s chair.

Mark Mobius died last month, on the fifteenth of April, in Singapore, at eighty-nine. He went out, by all accounts, bullish on India. The chair we saved him, in retrospect, was saved a bit late.

Stephen Schwarzman. Co-founder of Blackstone, currently $1.3 trillion in assets under management. In 2007, the Chinese sovereign wealth fund, China Investment Corporation, bought a non-voting minority stake in Blackstone's IPO — roughly 9.3 percent, carefully kept under the ten percent threshold that would have triggered a US national-security review. In 2013, Schwarzman announced a $100 million personal gift, plus a $200 million fundraising campaign, to build Schwarzman College at Tsinghua University in Beijing — a sort of Rhodes Scholarship for the China century. Its premise, in Schwarzman's own words:

“In the twenty-first century, China is no longer an elective course, it’s core curriculum.”

The advisory board includes Henry Kissinger, Condoleezza Rice, Hank Paulson, Robert Rubin, Colin Powell, Richard Haass.

Pause on what Schwarzman College actually represents. The largest internationally-funded philanthropic project in the history of the People’s Republic of China — over five hundred and seventy-five million dollars raised, with that advisory board reading like a roll-call of the postwar US foreign policy establishment — was created by an American private-equity billionaire to teach the next generation of Western elites that China was core curriculum. The college sits inside Tsinghua University. Tsinghua University is the campus where Xi Jinping went to school. The Schwarzman College buildings, designed by Robert A.M. Stern, sit on land that the Chinese state allowed to be allocated to a Western donor for the explicit purpose of credentialing future Western leaders inside a Chinese intellectual environment. The premise of the project was that the smart Western response to the rise of China was to deepen our understanding of it.

And while this enormous institutional bet was being placed in Beijing — over a decade, by some of the most credentialed figures in American public life, with the full and enthusiastic blessing of the Chinese state — the actual money was quietly moving in the opposite direction. Blackstone’s own internal reports were saying India. Not as a hedge. Not as a diversification play. As the best-performing market in the world.

“India has given us the strongest results across the world. And [I] am optimistic this [will] continue going forward.” — Stephen Schwarzman, Mumbai, March 2020.

In March 2025, Blackstone announced it intends to double its India exposure from $50 billion to $100 billion. By that point Schwarzman was personally announcing, in Mumbai, that he intended to put twice as much capital into India as Blackstone had ever put into China. The man who built the Rhodes Scholarship for the China century turned out, on inspection, to be a quiet bull on India. Possibly the largest one in the room. The hands and the mouth, working independently, possibly to the surprise of both.

And then there is Ray Dalio.

IV. Interlude — Ray Dalio Explains His Own Ideas

A short field trip.

Ray Dalio is the founder of Bridgewater Associates, the world’s largest hedge fund. He has written six books on Principles. He has, on multiple occasions, explained that he loves China — not metaphorically, but as a stated, formal reason for his investing decisions. In an April 2024 LinkedIn essay titled “To Answer the Question of Why I Invest in China,” the first reason Dalio listed was, in summary, love for China. Reason number two was that the country and its people were great. This is the man Wall Street took seriously on China for two and a half decades.

In December 2021, Dalio sat down with the economist Tyler Cowen — one of the most generous and well-prepared interviewers alive, a man who has a documented soft spot for India and who, when he asks you a question, has read every book on the table, including yours. What followed is, on the public record, instructive.

Exhibit A. Cowen asks the simplest possible question. The finance literature shows excess returns are very hard to predict. Dalio’s entire worldview is built on the idea that big macro forces — debt cycles, polarization, the rise of China — do predict where markets are going. So: do they? If yes, why doesn’t the research literature show that?

Dalio: “There are so many people who write finance papers, and then there are people who make money in the markets. I can’t speak for those who are writing the finance papers.”

Cowen presses. Dalio talks about debt cycles. Cowen presses again — if it’s publicly available information, why isn’t it already in prices?

Dalio’s final answer, in its entirety:

“Some can and some can’t. I guess you look at the track records over long periods of time, and you decide who can and who can’t.”

That is the response. The founder of the world’s largest hedge fund, asked to articulate his epistemic edge over the market, says: you look at the track records. The man who has written six books on Principles cannot, when politely asked, articulate a single Principle.

Exhibit B. Cowen, gently, hands Dalio what should be a softball:

“What is your favorite Chinese dynasty and why?”

Dalio has written a 500-page book that leans heavily on Chinese dynastic cycles. He has love for China. Surely he has a favorite.

Dalio’s answer runs over six hundred words. Across those six hundred words, he names zero favorite dynasties. He explains that he doesn’t really have a favorite. He gestures vaguely at “the Tang Dynasty, the Song Dynasty, the Ming Dynasty” without picking one or saying what he admires about any of them specifically. The closest he comes to substance is the observation that healthy dynasties had “education and civility” — a banality so spectacular it could be inscribed over the entrance of a community college lobby and would be considered, by the trustees, possibly too generic.

This is the man whose entire China thesis rests on a deep understanding of Chinese civilizational cycles. Asked to name his favorite Chinese dynasty, he cannot.

While Ray Dalio was being the smart money, Bridgewater’s flagship Pure Alpha fund underperformed the S&P 500 from July 2012 to August 2021 — nine consecutive years. In 2019, the leveraged version of Pure Alpha lost half a percent in a year that the S&P 500 returned 31.5 percent including dividends. Ray Dalio talks like the sage of the changing world order. Ray Dalio invests like a slightly underperforming index fund with substantially higher fees.

That is the level of intellectual seriousness Wall Street brought to the question of whether to send your retirement savings to Beijing.

V. What They Walked Into

Now, the system itself. Because this is where I want to be Milton Friedman for a minute, before going back to being Dave Barry. The Chinese system, on inspection, is not subtle. It was never subtle. The capture mechanisms have been hiding in plain sight for the duration. They are written in the law.

For most of the period from 1994 to 2022, a foreign automaker that wanted to build cars in China was required to do so through a joint venture in which the Chinese partner held at least half. GM-SAIC. Volkswagen-FAW. BMW-Brilliance. Toyota-FAW. Honda-Guangzhou. The ownership caps were lifted in stages — electric-vehicle makers freed in 2018, commercial vehicles in 2020, passenger cars in 2022 — but for the quarter-century that mattered, the rule held. The Western automakers brought the engineering, the manufacturing expertise, the brand, the supply chain knowledge, and the capital. The Chinese partners watched, took notes, and were given a quarter-century in which to absorb every transferable thing they could absorb. The five state-owned champions that emerged — SAIC, FAW, Dongfeng, BAIC, and Chang’an — now dominate the Chinese auto market alongside an indigenous ecosystem of competitors like BYD, Geely, Nio, and Xpeng that grew up in the shadow of these joint ventures, often staffed by engineers trained inside them. Today, in 2026, several of the original Western JV partners are being squeezed out of the Chinese market entirely, by Chinese champions raised on Western milk.

The same broad pattern appeared in aerospace. COMAC’s C919 is a Chinese aircraft, and the Chinese state is rightly proud of it — but its supplier stack leaned heavily on Western and Western-joint-venture components: engines from a GE-Safran venture, avionics from Honeywell, systems from Collins Aerospace. Boeing’s own joint venture with COMAC was a separate 737 completion-and-delivery center, not the C919. The point is not that Boeing secretly built China’s airliner. The point is subtler and more familiar: China built its competitor aircraft inside an ecosystem that Western firms had spent decades helping to assemble, and the Western firms handed over the components one purchase order at a time. Former US Deputy National Security Adviser Matt Pottinger told CBS News the C919 “looks like a knockoff.” It looks like a knockoff because a great deal of what is inside it came, by invitation, from the West.

Then there is the case of American Superconductor Corporation, which deserves a paragraph all to itself.

AMSC was a Massachusetts company that designed wind-turbine control software. Its largest customer was Sinovel, China’s biggest wind-turbine maker. In 2011, an AMSC engineer based in Austria, having been bribed by Sinovel, handed over the source code. When the theft surfaced, AMSC’s stock collapsed; the company lost the great majority of its market value over the following months, and laid off seven hundred people — more than half its global workforce.

In January 2018, a federal jury in Wisconsin convicted Sinovel of trade-secret theft. The court found AMSC’s losses exceeded $550 million. The maximum statutory fine the court was empowered to impose was $1.5 million. AMSC eventually settled for $57.5 million in restitution — roughly a tenth of what it had lost in a single trading session, and itself a recovery rate, on the court’s own damages finding, of about ten percent.

According to reporting from the time, Sinovel’s CEO Han responded to the early investigation by allegedly describing software as “like cabbage.” In China, software is like cabbage. You don’t pay for cabbage. You just take some.

This is the country into which Wall Street poured, by some estimates, eight hundred billion dollars in new investment between 2017 and 2020 alone.

And then there is the actual machinery of getting your money out. The People’s Republic operates a closed capital account, which is a technical term meaning that the State Administration of Foreign Exchange — SAFE, in a piece of regulatory nomenclature so on the nose it could have been written by a Jon Stewart writers’ room — must approve every outward flow of capital, dollar by dollar, audit by audit, year by year. Foreign companies wishing to repatriate profits must first pay twenty-five percent corporate income tax, then submit to a mandatory annual audit, then permanently allocate ten percent of their after-tax profits to a “Surplus Reserve Fund” — locked away in renminbi, in China, until that fund reaches half the company’s registered capital — and finally secure SAFE approval, which can take months. Dividends may only be remitted once per fiscal year.

If you try to route around this with intercompany service fees or royalties, you trigger an anti-avoidance audit. Anti-avoidance audits take years to resolve. While they are being resolved, you may not be permitted to leave the country.

Which brings us to exit bans. A 2024 Stanford Journal of International Law paper documented 128 foreigners — twenty-nine Americans, forty-four Canadians, the rest from elsewhere — placed under exit bans in commercial disputes between 1995 and 2019. The number is now believed to be much higher. The civil-society group Safeguard Defenders reports that exit-ban mentions in China’s official court database rose from fewer than five thousand in 2016 to thirty-nine thousand in 2020, and have doubled annually since.

A partial list of recent named cases, all of them foreign business executives:

• Chenyue Mao, a managing director at Wells Fargo, banned from leaving China in July 2025. Allowed to return only in September 2025, after high-level US-China negotiations.

• An Orange County businessman, the so-called “Horowitz” case from 2011, banned from leaving China after rejecting a defective Chinese supplier’s product. Released only after his wife wired $250,000 to the supplier whose product he had refused.

• Richard O’Halloran, an Irish executive, held in China for three years between 2019 and 2022 over a Chinese partner’s fraud that he had not personally committed and could not have prevented.

There are many more. Bankers from Nomura. A US Patent and Trademark Office employee, on personal travel with his family. The five Chinese staff of the American due-diligence firm Mintz Group, detained for two years between March 2023 and March 2025. The firm itself was fined $1.5 million for “unauthorized statistical investigations” — an offense which, by the standards of jurisprudence, has the same precise definition as “vibes the state did not like.”

A commercial-law professor at Columbia, writing on the school’s Blue Sky Blog in 2024, summarized the operational reality this way: an exit ban can be enforced against an individual executive of a US company when an ordinary commercial dispute arises. The Chinese company involved can ask a court to impose the exit ban through a formal legal process that occurs unbeknownst to the party whose exit is to be barred.

You only discover you cannot leave when you try to leave. Or, as the Eagles put it in a song that, on reflection, may have been about the People’s Republic of China all along: you can check out any time you like, but you can never leave.

And then there was Evergrande. The Chinese property developer, formerly the country’s largest and a perennial fixture in JPMorgan’s emerging-market bond indices, defaulted in 2021 with $300-plus billion in liabilities. Foreign bondholders held about $19 billion of the offshore exposure. The bonds traded at 2.25 cents on the dollar in late 2023. A Hong Kong court issued a winding-up order in January 2024. The order has “limited recognition” in mainland China — where ninety percent of Evergrande’s $242 billion in assets are physically located. Effective recovery for foreign creditors: approximately one cent on the dollar.

If you lent Evergrande a million dollars, you got back enough for a moderately priced lunch in midtown Manhattan.

This is what Wall Street, for a decade, called China’s investment-grade bond market. This is what Ray Dalio called China being tremendously successful in many different ways.

VI. The Country the Smart Money Missed

Now, for contrast, India.

India’s bureaucracy is real. India’s tax authorities are aggressive in a way that produces, on a recurring basis, international incidents and full-blown investor revolts. India self-inflicted, in 2012, a retrospective-tax amendment that effectively reached back five years to tax Vodafone’s already-completed acquisition of Hutchison’s Indian assets. The amendment also caught Cairn Energy, the British oil company. The international investor community lost its mind. Both companies took India to international arbitration. This is the kind of self-inflicted regulatory wound that, in many countries, would simply end the conversation about foreign investment for a generation.

And then something happened that has not happened in the People’s Republic of China in living memory.

Vodafone won at the Permanent Court of Arbitration in 2020. Cairn won $1.2 billion plus interest at the PCA in December 2020. India initially refused to pay. Cairn began enforcement actions, seizing Indian government-owned property in France. And then, in August 2021, India did what no closed-capital-account state ever does. The Government of India repealed the law. The Government of India paid. Cairn received roughly ₹7,900 crore — approximately a billion dollars — back in its bank account. Seventeen companies caught in the retrospective tax dragnet were absolved.

The system bent. The system paid. Because the system, however maddening on a day-to-day basis, is one in which international arbitration awards eventually have to be honored. India is a country in which, after the tax authority has tortured you for nine years, the courts may eventually rule for you. And when they do — and this is the part that should be carved into a wall somewhere on Wall Street — you actually get your money.

Compare this to Evergrande. Compare this to AMSC. Compare this, especially, to a man who died in his own bed in his own country knowing that a Chinese court had agreed he was right and that none of that had ever come to anything.

In India, you are slowly audited but can fight and get your money. In China, you are slowly eaten and will have nothing left to get.

The roster of foreign-corporate India operations that have run, without ever having a Mintz Group experience, is now extensive. Walmart bought seventy-seven percent of Flipkart for $16 billion in May 2018 — the largest e-commerce M&A in Indian history. The Competition Commission of India approved it, despite vocal opposition from domestic trader associations. Imagine, for a moment, the Chinese state regulator approving a $16 billion American acquisition of the leading Chinese e-commerce platform. Spend a minute with that mental image. Take your time.

Walmart has since spent another $3.5 billion to take its stake to roughly eighty percent. Amazon has invested approximately $11 billion in India to date, with $15 billion more committed by 2030. Google announced a $10 billion India Digitization Fund in 2020. Microsoft is in the middle of a multi-billion-dollar Azure expansion. Apple's revenue in India grew 36 percent year-over-year to roughly $8 billion in FY24, making India Apple's fastest-growing major market. Foxconn is scaling iPhone assembly in Tamil Nadu at a pace that has begun to alter Apple's global manufacturing geography.

None of these companies has had to wire $250,000 to a domestic supplier to recover a missing executive.

There is one more observation worth pausing on, because it is the consumer-side expression of the same civilizational difference.

Nike’s revenue in China is down twenty-eight percent from where it was five years ago, even as China’s overall sportswear market has boomed. Nike’s stock just hit a decade low. The company has laid off fourteen hundred people and jettisoned its longtime China leadership. Starbucks has sold a majority stake in its China operations to a local partner. Guess closed all of its more-than-one-hundred-fifty Chinese stores this past March. American automobile brands have been hammered by Chinese EV companies. Anta, formerly a domestic Chinese shoe brand, is now the largest shareholder in Amer Sports — which is to say, the Chinese athletic conglomerate owns Arc’teryx and Wilson. The high-end Western brands Americans buy at REI are, beneath the logos, Chinese. (The Nike data, and a great deal more, comes from a recent Wall Street Journal investigation — well worth reading in full.)

The proximate cause is a Chinese consumer-side phenomenon called guochao — a state-amplified wave of cultural-pride-driven preference for domestic over foreign brands, occasionally weaponized into formal boycotts when a Western company says something the state does not like about Xinjiang or Tibet or Taiwan. Foreign brands in China now operate in a market where loving the domestic brand is patriotic and buying the foreign brand is, depending on the political weather, mildly suspect.

In recorded Indian history, with one exception detailed below, no government — including the current Modi government, which is the most assertively Hindu-nationalist administration modern India has produced — has ever told Indian consumers that buying foreign brands is unpatriotic. Apple’s revenue in India is growing thirty-six percent a year. KFC operates extensively across the country. McDonald’s, Domino’s, Subway, Starbucks (yes, the same Starbucks that just sold its China business), Levi’s, Zara, H&M, Uniqlo, Nike, Adidas — all of them operate, profitably, in Indian cities. Mumbai is full of Mercedes-Benz dealerships. Delhi is full of BMW showrooms. India certainly has nationalist economics — swadeshi sentiment, “vocal for local” campaigns, periodic boycotts of Chinese goods in particular. What India has never produced is a durable, state-amplified consumer movement of the Chinese kind: the kind that can turn a global brand into a patriotic liability overnight and keep it that way. There is no durable Indian guochao, there has never been one, and there is unlikely ever to be one, because Indians want the best product available, do not particularly care where it comes from, and have been consistent on this for a very long time. The Indian state can do many things. It cannot turn off the Indian consumer’s millennia-old willingness to buy whatever is good, from whoever is selling it, at whatever price clears.

The one period in modern Indian history when Indians were not allowed to buy foreign brands was the era between roughly 1969 and 1991 — when the Indian state, briefly persuaded that Soviet-style autarky was the path to prosperity, expelled Coca-Cola and IBM in 1977, restricted imports through the Foreign Exchange Regulation Act, and tried to manufacture everything domestically behind a wall of tariffs and the License Raj. The result was a national catastrophe. Indians spent twenty-two years standing in line for badly made domestic substitutes they did not particularly want, while their cousins overseas drove imported cars and drank imported soft drinks. The 1991 reforms — which I have covered at length in Posts 19A through 19C — stuck, and have stuck for thirty-five years now, against the explicit ideological preferences of every successive Indian government including the present one. They stuck not primarily because economists won the argument but because the Indian population had personally experienced the consequence of being forbidden to buy foreign brands and had no interest in repeating the experience.

The political mechanism by which Coca-Cola was kicked out of India is worth a brief detour, because it illustrates the breadth of the failure across the Indian political class. The Foreign Exchange Regulation Act — the law that required foreign companies to dilute their Indian operations to forty percent local equity, and that demanded, in Coca-Cola’s case, that the secret formula itself be handed over to an Indian state-monitored body — was passed in 1973 by Indira Gandhi’s Congress government. Coca-Cola refused to comply and continued negotiating. Two years later, in 1975, Mrs. Gandhi declared a national Emergency and suspended Indian democracy outright. Two years after that, in 1977, the Indian electorate threw her out in favor of a coalition — the Janata Party — that had been specifically organized to reverse her authoritarian excesses. The new Prime Minister, Morarji Desai, and his Industry Minister, George Fernandes, then proceeded to take office and not undo FERA but, instead, to enforce it more strictly than Mrs. Gandhi had ever bothered to. Coca-Cola and IBM were given ultimatums. Both refused to comply. Both left.

The timeline, summarized, reads: the ruling party writes the rule, suspends the constitution, loses the next election to the people who had organized themselves around the explicit promise of fixing her, and the new government — having taken office on a wave of national revulsion against Mrs. Gandhi’s overreach — then keeps one of her economically dumbest rules and uses it to expel Coca-Cola. This is not one party’s blunder. This was the entire Indian political class, across every available faction, displaying a remarkable bipartisan inability to operate within their own civilization’s five-thousand-year-old operating system. The civilization had run the largest trading economy on earth. The civilization had been operationalizing Ricardo since the Mauryan period. None of that information appears to have penetrated to the Lok Sabha in either 1973 or 1977. Both parties, in succession, failed at being India.

Indians remember what closure tastes like. It tasted like Campa Cola. It could, alternatively, have tasted like the urine of the Prime Minister of India — Morarji Desai, who expelled Coca-Cola from the country in 1977 and who, in his spare time, drank his own urine daily for health and recommended the practice to Dan Rather on 60 Minutes in 1978. The official position of the Government of India in the late 1970s was, on the available evidence, that piss was healthier than sugar. Indians, alone in the world in this assessment, respectfully disagreed. They have been disagreeing, in increasing volume, ever since. They are not eager to find out what closure tastes like again.

That is, in significant part, why foreign capital makes money in India and is permitted to leave with it. The Indian consumer demand for foreign products is itself a civilizational artifact older than the Indian state. The state cannot suppress what it did not create — and the one time the state tried, the population eventually voted, in 1991, to never let the state try again.

The Chinese state, by contrast, can and does turn consumer behavior on and off like a faucet. Capital in China cannot leave because the state will not let it. Foreign consumer brands in China cannot earn because the state has decided they should not. The two phenomena are, civilizationally, the same phenomenon — and a five-thousand-year-old open trading society would never tolerate either.

VII. The Pattern Recognition Problem

Here is the thing I think actually explains the whole story.

The smart money on Wall Street has not been dealing with the Chinese for very long. A few decades, in earnest. Long enough to write some books with foolishly confident titles. Not long enough to recognize a pattern.

The Indians have been dealing with the Chinese for a very long time. The Indian merchant class has been trading along routes that touched Chinese trading partners since the early days of the Silk Road — at least the second century BCE, well before there was a country called the United States, before there was an England, and roughly contemporaneously with the Roman Republic. Indian traders along the Silk Road, Tamil merchants in Southeast Asia, Marwari and Gujarati traders moving silk and spices across the Indian Ocean, Parsi merchants in Hong Kong and Canton in the eighteenth and nineteenth centuries — there is no generation of Indian commercial life in living civilizational memory that has not, at some point, encountered the Chinese system.

Here are two facts that the modern China booster ought to know and, on the available evidence, does not. The English word Mandarin comes, through Portuguese mandarim and Malay menteri, from the Sanskrit mantrī — “counselor,” “minister of state” — and entered English in the 1580s. The English word China has a more contested path; most authorities trace it ultimately to the Qin dynasty, but the form that carried that name westward, into Persian and onward to Europe, was the Sanskrit Cīna. Both words are roughly five centuries old in English, and both reached English through the channels of Indian and Indian-Ocean commerce rather than through any direct European contact with China. The Westerner who today announces that he is bullish on China, and that he is having his children learn Mandarin, is — without knowing it — describing his enthusiasm in vocabulary that came west along Indian trade routes. India did not need Goldman Sachs to introduce it to China in 1997. India had been the West’s interpreter of China since before the East India Company was chartered. If you want to understand how China treats the outsider who arrives to do business, the people to ask are not the ones who showed up in the 1990s with a Bloomberg terminal. The people to ask are the ones who supplied the vocabulary.

And the Chinese system, on every one of those encounters, has done what the Chinese system does. It captures. It absorbs. It treats foreign commerce as a temporary accommodation to be terminated at the convenience of the state.

Consider the Canton System, which the Qing imposed from 1757 until the First Opium War in 1842. Western traders — British, Dutch, French, eventually Americans, and a substantial Parsi contingent from Bombay — were restricted to a single port, Canton (today’s Guangzhou). Within that port they were confined to a small physical compound on the riverbank, the “Thirteen Factories.” They were forbidden from learning Chinese, from traveling inland, from staying in the country during the off-season, and from doing business with any Chinese merchant who was not a member of a small state-authorized cartel called the Cohong. The Cohong set the prices. The Cohong arbitrated the disputes. The Cohong reported, when asked, to the imperial bureaucracy. The structural parallel to today’s foreign-investment regime — the negative list, the SAFE controls, the exit bans, the requirement to do business through politically vetted intermediaries — should be obvious to anyone willing to look. The People’s Republic is not inventing anything new. It is running a playbook that is roughly two and a half centuries old, with better software.

The Parsi merchants of nineteenth-century Bombay — the Wadias, the Sassoons, the Jejeebhoys — traded with China extensively, opium for tea, cotton for porcelain, silver for silk. Sir Jamsetjee Jejeebhoy went to China five times. He made one of the largest private fortunes in pre-modern India. He was knighted for it. And he came home with the money. Please note: he came home with the money. He did not buy a building in Canton. He did not set up a permanent Jejeebhoy Office in Shanghai with a five-year lease and a politically connected local partner who would, when the lease expired, point at a building. He went, he traded, he left. Every generation of Parsi merchants for the next century repeated this lesson. The Sassoons did the same. The Wadias did the same. There is a reason these families’ fortunes still exist in Bombay today and not in Guangzhou.

It is also worth pausing on what those fortunes funded. The Bombay Stock Exchange was founded in 1875, which makes it the oldest stock exchange in Asia. It predates the Tokyo Stock Exchange by three years. It predates the Shanghai Stock Exchange by one hundred and fifteen years — Shanghai’s modern exchange opened in 1990, when the People’s Republic of China decided, after forty years of revolutionary anti-market posturing, that perhaps capital markets were not entirely a bourgeois invention. The Shenzhen Stock Exchange also opened in 1990. China’s national capital-markets infrastructure is, in calendar terms, younger than my own children. The BSE is older than my great-grandfather.

And the 1875 founding date understates the case considerably. The BSE was the formal institutional crystallization of merchant-banking and credit arrangements that had been running, in continuous operation, on the Indian subcontinent for centuries before that. The hundi system — the indigenous bill-of-exchange network operated by Marwari, Gujarati, Multani, Bohra, and Chettiar bankers from Kabul to Rangoon — was a fully developed transnational credit instrument by the Mughal era and traces, in its earliest forms, all the way back to the Mauryan and Gupta periods. Which is to say: Indians were running written bills of exchange before England existed as a country. Surat was a recognized banking center in the 1600s, when London was still figuring out joint-stock companies and Amsterdam had only just invented the modern stock exchange. The Jagat Seths of Bengal financed armies and emperors in the eighteenth century; the East India Company itself was, for substantial periods, dependent on Indian banking houses to clear its own transactions. India did not learn capital markets from the British. The British learned a great deal of practical credit management from doing business inside India’s existing one.

Ricardo did not invent comparative advantage. Ricardo described what the Indian merchant class had been operationalizing for two millennia: the principle that everyone gets richer when each region produces what it produces best and trades for the rest. By the early eighteenth century, this had made India approximately twenty-five percent of world GDP — alternating in pole position with China across the imperial centuries, and at points the single largest economy on earth. (I have written about that twenty-five-percent figure in earlier posts.) That share was not extracted from anybody. It was the cumulative output of an open trading civilization doing what open trading civilizations do: selling textiles and spices and steel and gemstones and ships to the world, importing what the world had that it did not, and getting richer over the course of fifteen hundred years. The British did not come to India because India was poor and they wished to develop it. The British came to India because India was rich and they wanted some.

Which means: when the Wall Street fund manager of 2007 looked at India and saw “an emerging market” and looked at China and saw “the future,” he was operating under a fairly stunning misunderstanding of which of the two civilizations had the longer continuous tradition of running actual capital markets. The answer is India, by a margin of approximately three centuries and not really close. The Indian merchant class did not need to be taught what a security was. The Indian merchant class was already doing this in turbans, by hundi, in the seventeenth century, while Europeans were still debating whether interest constituted usury. The Indian capital market is not a 1991 invention or a 1991 recovery. It is a thing India has been doing continuously, with brief state-sponsored interruptions, for roughly four hundred years. The 1991 reforms simply removed the regulatory bandage that the Nehru-Indira state had put on top of the existing system, and the existing system — which had never actually died, only gone underground into informal networks of money-changers, hawala operators, and family-based credit lines — promptly resurfaced and went public.

The Chinese state has, by contrast, no equivalent continuous tradition. The Qing dynasty did not have public capital markets in the modern sense. The Republican period (1912-1949) produced a brief and chaotic flowering of one in Shanghai, which Mao closed in 1949 as a matter of explicit ideological commitment. Forty years later, when the People’s Republic decided it needed capital markets again, it had to build them from scratch — with no living memory of how to run one, no continuous merchant-banking tradition to draw on, and no civilizational disposition that regarded foreign profit-taking as anything other than a temporary indulgence. China’s capital market is forty years old. India’s is, depending on how you count, between one hundred and fifty and four hundred. Both countries have stock exchanges now. Only one of them has a civilizational understanding of what they are for.

The Indian merchant class has absorbed this pattern across generations the way Inuit children absorb the names of the kinds of snow. It is in the muscle memory of the civilization. Any Marwari trading family of any standing would have, in living oral tradition, multiple cousins-of-cousins who had tried to do business in China and come back with a building, or come back with nothing, or not come back at all. The thing the smart money discovered in 2023 — that there is a meaningful, structural, civilizationally-rooted difference between “letting foreigners come and trade” and “letting foreigners come and leave with their winnings“ — was known to my grandmother. It was known to her grandmother. It was known to a substantial fraction of the merchant subcastes of pre-colonial South Asia.

I have, since the early 1990s, told anyone who would listen that this was the wrong trade. I was, for most of those years, treated as a man with an ethnic axe to grind. The axe, it turns out, was a useful piece of inherited equipment.

The thing about pattern recognition, which is what civilizations are for, is that it does not scale through books. Stephen Schwarzman cannot fix his thirty-year-old pattern-recognition problem by endowing a college at Tsinghua. Ray Dalio cannot fix it by reading five hundred years of European empire history. Mark Mobius cannot fix it by being good at cocktail-party storytelling about Indonesian copper mines.

The pattern is in the bones. You either grew up in a civilization that has been on the receiving end of this particular operating system for the better part of two thousand years — in which case the move is obvious from across the room — or you did not, in which case you are going to send your money to Shanghai and a man is going to point at a building.

The East India Company is the standard answer to “well, but India did get colonized, didn’t they?” Yes. India did get colonized. By a British trading company that started in 1608 as a small commercial concession and ended, two and a half centuries later, as the de facto government of the subcontinent. India did not see that one coming. India was, that time, the slow learner.

But India did learn one specific thing. The lesson was not “be open to foreigners,” because India was open to foreigners well before the East India Company arrived and remained open to them well after the British left. India had been open to Greeks, Scythians, Huns, Arabs, Mongols, Persians, Portuguese, Jewish refugees, Zoroastrian refugees, Tibetan refugees, and a substantial number of confused British backpackers, for two millennia before 1608, and continued to be open to all of them after 1947. The openness was not a lesson. The openness was, and is, the operating system.

The thing India learned from the East India Company episode was narrower and more specific: do not let a foreign commercial entity capture the state. Take their investment, tax their profits, make them obey your laws, but do not — under any circumstances — allow a private foreign corporation to acquire its own army and start collecting your revenues. The Indian state in 2026 has many features that an outside observer might reasonably find objectionable, but the one feature it does not have is foreign-corporate sovereignty. Apple cannot raise a regiment. Walmart cannot collect the land tax. The mistake from 1757 to 1857 was not openness. The mistake was openness without sovereignty. India fixed the second part. The first part was never broken.

India then proceeded, in its characteristic way, to make other enormous mistakes — the forty-year socialist autarky described above being the most consequential. Through all of that, the civilization was open. The state was, for a while, deranged. The civilization waited the state out. In 1991, the state corrected itself, partially because the foreign exchange reserves had run down to two weeks of imports and there was no further choice. The civilization always was, and is, open. The state’s forty-year closure was a self-inflicted exception, now corrected.

China’s trajectory has no equivalent of this pattern. China did not open in 1608, get colonized, close down in 1949, briefly re-open in 1979, and remain genuinely open thereafter. China was a closed civilization before the Opium Wars. China was forcibly pried open by Western military violence in the nineteenth century, which it never accepted as legitimate. China resealed itself under Mao. China then announced a “reform and opening” in 1979 that was always — always — designed to be reversible, controlled, sectorally bounded, and revocable at the discretion of the state. The negative list is not new. The capital controls are not new. The exit bans are not new. The hostility to foreign profit-taking is not new. They are continuous with five centuries of Chinese statecraft, briefly interrupted by the period in which the West was strong enough to force the door open and is now strong enough no longer.

China was not open and is not open. China was open in the way a department store is open: you may come in during business hours, browse, purchase under the supervision of staff, and leave through the marked exit. You may not move in. You may not take inventory home. You may not stay past closing. The store reserves the right to refuse service and to keep your coat.

India was open and is open in the way a port city is open: people arrive, people leave, some stay, some get rich, some get robbed, the harbor accepts all comers and the city absorbs whatever the harbor delivers. The British arrived and stayed too long and India had to evict them. That eviction was a correction, not a closing. The door was open before they arrived and the door was open after they left.

One civilization is, civilizationally, a port city. The other civilization is, civilizationally, a department store. Both have been what they are for centuries. The state’s behavior in each case is downstream of what the civilization is.

To put it without the metaphor: the difference is not that one society likes money and the other does not. Everyone likes money. China likes money ferociously. The difference is whether the foreigner is understood as a guest who may leave with his winnings, or as a resource to be managed until the state decides the game is over. India, for all its bureaucratic cruelty, has consistently treated the foreign trader as the first thing. China, for all its reform-era hospitality, has consistently treated him as the second.

The point of this entire post is that the consequences of those two civilizational dispositions are completely predictable, that they have been predictable for centuries, and that a generation of fund managers paid millions of dollars a year to predict things spent thirty years not predicting this one. They were not making a hard call wrong. They were making an easy call wrong. The signal had been transmitted, continuously, for half a millennium. They were not listening for it.

The smart money was not the smart money. It was the inexperienced money, with a Bloomberg terminal.

VIII. The Fifty-Year Bet

Here is what I actually think.

In fifty years, India will be a richer and more prosperous country than China. Not slightly richer. Meaningfully richer. Not in the conventional emerging-markets-catch-up way, in which a low-base country runs faster than a higher-base country for a few decades and then converges. In a structural, civilizational, this-is-how-the-two-countries-are way. I will not be alive in fifty years to be proved right or wrong about this. I am, nevertheless, willing to be on the record now, in 2026, because I have been on the record about everything else in this post since the early 1990s and the only thing that has changed is that the rest of the world is, slowly, catching up.

The inversion has already happened, mechanically. The 2023 numbers tell you the truth in the antiseptic language of the Bureau of Economic Analysis: new US flows to India (6.0billion)exceedednewflowstoChina(5.1 billion) for the first time in the data series. On China’s balance-of-payments measure, inward direct investment turned negative in the third quarter of 2023 and the full-year figure collapsed from its 2021 peak to a fraction of it — the steepest sustained decline in the history of that series. The American Chamber of Commerce in Shanghai’s 2025 survey reported that twelve percent of US firms now rank China as their headquarters’ top investment destination, the lowest figure in the survey’s history. The AmCham China business climate survey for 2025 found only forty-nine percent of US companies expected to be profitable in China that year. Blackstone announced in March 2025 that it would double its India exposure to $100 billion. Even Jim Rogers, in April 2024, finally turned bullish on India.

When Jim Rogers has turned bullish on you, you can be reasonably confident that the trade is no longer contrarian. But that is not the trade I am describing.

The trade I am describing is louder and longer and considerably more lucrative for whoever has the nerve to take it seriously. India will, by 2076, be a richer country than China. Let me be precise about how bold that is. As I write, China’s GDP per capita is roughly five times India’s. China is, today, far ahead — richer, better-built, with newer airports, faster trains, and a manufacturing base India cannot presently approach. Anyone betting on India over a fifty-year horizon is betting against an enormous current lead. I am making that bet anyway, on the record, in 2026. India’s economy will, by the middle of this century, almost certainly be the third largest in the world by nominal GDP, and possibly the second. India’s per-capita wealth will, by 2076, exceed China’s. The reasons are exactly the reasons this entire series has been laying out for twenty installments: a young population, a working democracy, a functional (if maddening) court system, an English-speaking professional class that has placed Indians at the head of a remarkable number of America’s largest companies, an open civilizational disposition that absorbs and metabolizes foreign capital and talent rather than caging it, and a state which — for all its absurdities — does not as a matter of policy disappear foreign executives who write the wrong email. China’s current lead is real, and it is vast. My wager is that it is also, on a fifty-year clock, perishable.

China has, by contrast, an inverted population pyramid that is now in irreversible structural decline, a one-party state under Xi that has spent five years systematically destroying its own most productive sectors, a real estate bubble whose final accounting has not yet been fully done, a youth unemployment rate that the state has, at various points, simply stopped publishing, and a capital regime that has, as this post has documented at length, made it clear to the entire planet that money sent into China is no longer money that may, with full assurance, leave China.

The “smart” money is, in 2026, beginning to act on all of this. Quietly. The Apple manufacturing build-out in Tamil Nadu. The Foxconn factories. The Blackstone India fund. The Walmart-Flipkart hold. The Google Digitization Fund. The Microsoft Azure expansion. The Amazon billions. Every one of these companies is, right now, this quarter, voting with very real money for the thesis that India is the next quarter-century’s growth story. Almost none of them are saying so on television.

This is the actual smart bet. And it is being placed, in size, by the people whose money it is — even when the people whose mouths it is do not yet say so. The question I find myself wondering about, watching this from a desk in Connecticut, is not whether the bet is being placed. The bet is being placed. The question is: how long is it going to take before the smart money gets loud about it?

Because the day the smart money gets loud about India is the day the trade is over for whoever waited for the loudness. The trade is happening now, while Ray Dalio is still writing essays titled “To Answer the Question of Why I Invest in China.” The trade is happening now, while Schwarzman College is still graduating cohorts of Westerners credentialed in the China century. The trade is happening now, in 2026, while the cover stories of the major business magazines are still about Xi Jinping. The capital that gets there before the noise is the capital that captures the run. The capital that arrives after the cover stories arrive late.

I have, since 1991, been on the record. Quietly, in dinner conversations. Loudly, in this Substack, for the past two years. The smart money is, in some sense, finally catching up — which is to say, finally acting. The remaining gap is between what the smart money is doing and what the smart money is saying. I am writing this post, in part, to try to close that gap by a few weeks. The act and the word should move together. The act has moved. The word is dragging.

I will not be alive in 2076 to see India become a richer country than China. I would like, very much, before then, to be alive to see the people who run the world’s largest hedge funds say out loud the thing their own internal investment committees are already, this quarter, voting for. I would like that to happen on a stage. With cameras. I would like it to be the moment Ray Dalio is asked, on a podcast, what his favorite Indian dynasty is, and is not able to come up with a single one, because he hasn’t done the homework yet, again.

We will probably get there. I just hope we get there fast enough that some of the people who have been right about this for thirty years are still around to enjoy it.

Not all of us will be.

IX. The Building Floor, Revisited

My partner, as I mentioned at the beginning of this post, is no longer with us. He died knowing he had been cheated. He died knowing the Chinese court had agreed he had been cheated. He died knowing that none of that was ever going to be made right. Somewhere in some Chinese city, there is a floor of a building that was, in some technical and entirely useless sense, his. He never saw it. He never wanted to see it. He flew home empty-handed once and he never went back. The capital he had wired into the country stayed in the country. The principle of the thing stayed with him until the end. That is a long time to carry a bitter pill, and it is the kind of pill that does not stop being bitter just because the person carrying it has stopped being alive to taste it.

Compare him to Cairn Energy. Cairn lost $1.2 billion to the Indian retrospective tax. Cairn sued in international arbitration. Cairn won. Cairn began seizing French government property held by the Indian state. India, having been thoroughly embarrassed, repealed the law and wrote a check for approximately a billion dollars within a year. Nobody at Cairn died waiting.

This is the choice the smart money spent thirty years getting wrong.

One civilization will offer you a floor in a building, and then it will outlive you, and the floor will still belong to the man who took your money.

The other civilization will, eventually, after enough screaming, write you a check while you are still alive to cash it.

If you want to know which country is going to be rich in fifty years, do not look at where the smart money is currently parking. Look at where the smart money eventually escapes back to with whatever is left of its principal. The smart money, in retrospect, has not been very smart. But the dumb money — the bureaucratic, exasperating, infuriating, court-clogged, audit-happy Indian capital market — has, quietly, been giving people their money back at the end of the dispute. Sometimes after nine years. Sometimes after a dozen. But back, while the people are still here to receive it.

That is what a civilization that wants foreigners to make money looks like, even when it is simultaneously doing everything in its power to make them sweat for it. The sweat is real. The exit is also real. The check, eventually, clears.

For thirty years, I have told anyone who would listen: invest anywhere you want, but not in China. I was a crank. I am, in 2026, slightly less of a crank. The Mark Mobiuses of the world have, belatedly, joined the chair-shortage in the room where the cranks meet. The room is more crowded than it used to be.

There is one chair, in that room, that will always be empty. It belongs to a partner of mine, who did not get to be in the room when the rest of the world figured out what he had figured out the hard way, with his own money, in his own lifetime, in a Chinese courtroom. He is, in some final sense, the reason this post exists. The capital he lost was, in the grand scheme of global FDI flows, not very much. The principle of the thing was, and is, a great deal.

We have a few extra chairs. Henry Kissinger is no longer with us. Mark Mobius, sadly, joined him last month — he died in Singapore, bullish on India to the end, which is the way one would want to go out. We cannot save him a seat. But Kissinger’s blurb-recipients are increasingly welcome. Larry Summers may show up any time. Ray Dalio, if he is reading this — and one assumes that a man who has written six books about his own Principles has a Google Alert for his own name — Ray, the door is open. We have a chair for you too.

Just leave the book on the way in. We have read it.

And on the way in, perhaps, walk past the empty chair. Spend a moment with it. Think about the man who should have been sitting in it. Think about the building in the city he didn’t live in. Think about what it means that the brightest financial minds of our generation sent enormous sums of other people’s money into a system that does this to people, and that some of those people died waiting for the system to be fair to them, and that the people who sent that money are still on television.

Then sit down. We have a lot to talk about.

If you found this essay worth your time, you may enjoy my book The Science of Free Will, which asks an equally uncomfortable question about an equally cherished story.

New to The India Paradox? Start with The Paradox of India, the first essay — or browse the whole series.

Related country-specific work: Albion — Britain’s institutional decline, by someone who first saw the place in 1979.

People have been arguing about whether UAP sightings are aliens. Most of the arguments are not very good — blurry videos, appeals to authority, arguments from incredulity. But two very serious thinkers have made the case that the probability is appreciably nonzero, and their arguments are not easy to wave away.

(A note on terminology: I use “UAP” for what people report seeing, and “craft” only for the hypothesis I’m testing. The question is not whether people see things they cannot identify. They do. The question is whether the craft interpretation survives the physics.)

Tyler Cowen — the George Mason economist who runs Marginal Revolution — frames it as a conditional: if UAP sightings are alien, the most plausible model is unmanned drone probes with generalized software instructions — “seek out major power sources, send information back, run away if approached.” He treats it as a question about how advanced civilizations allocate discretionary surplus when costs are low and time horizons are infinite. He raises his own counterarguments. He honestly concedes the sightings “probably are not of alien creations.” But he puts the probability at around 10%.

— Tyler’s George Mason colleague, whose work on prediction markets and “grabby aliens“ has made him one of the most original thinkers in the rationalist sphere — has gone much further: less than 20% probability on all UFOs being illusions, an elaborate structure involving stellar-nursery relatives, seeding, no-colonization rules, a 75-year military conspiracy. Robin’s version is bolder and more specific, which means it makes more testable predictions — exactly what you want from serious reasoning about a hard problem.

These aren’t cranks on Reddit. These are people whose reasoning I respect on every other topic. And their arguments made me think hard: could they be right? Could the probability really be 10%, or higher?

I’m a physicist — or at least, a recovering one. I have a PhD from UT Austin, where George Sudarshan was my advisor and Steven Weinberg was on my committee. When Congress, in its infinite wisdom, canceled the Superconducting Supercollider — the pork wasn’t spread widely enough; only Texas and Illinois were getting the bulk of the spending, which was the whole point of keeping costs down and using existing expertise — the writing was on the wall, and I had to find something else to do. My first job interview (and only job interview — I have never technically had a job in my life, unless you count being a graduate student) was at a hedge fund that has since become very famous. They wanted someone to write code to “clean” data. I asked why they wanted a physics PhD to do that. There was a long, awkward pause, and they said: “Well, physics is boring, and cleaning data is boring, so we figured a physics PhD would like it.” I spent the next three decades in quantitative finance instead. The physics, it turns out, was not boring. It was just waiting. (BTW, I’m back to writing physics papers again. Check them out if interested: samirvarma.com/publications. Turns out you can take the physicist out of physics but you can’t take physics out of the physicist. But I digress. Back to the original story.)

So I gave Tyler and Robin’s arguments the standard physicist’s treatment: what does the physics actually say?

And I immediately got it wrong. Twice.

I’m not a debunker. I’m not saying these are camera tricks, parallax artifacts, or people who don’t understand how video works. Maybe some of them are, maybe some aren’t — that’s a different argument, and other people are having it. My argument is different: even if every one of these videos shows a real physical object doing exactly what it appears to be doing, the physics of interstellar travel makes the alien-probe interpretation extraordinarily implausible. The observations can be entirely real and still not be probes. That’s the option I want to put on the table.

What follows is the story of how I tried to find what was wrong with Tyler and Robin’s reasoning, kept failing, started to think they might be right — and then found a calculation that, as far as I can tell, hasn’t been part of this conversation. It resolves the version of this question people usually argue about. The answer is lim(ε→0) ε — zero for all practical purposes.

My First Mistake: Speed of Light

My first instinct was the speed-of-light argument. Probes have mass. They can’t exceed c. The nearest star systems are a few light-years away; interesting targets span anywhere from single digits to hundreds of light-years. The galaxy is 100,000 light-years across. Whoever sent them would be long dead before any signal returns. Surely that settles it?

It doesn’t.

The Milky Way has hosted star formation for over 10 billion years. A civilization arising even a few million years before us — a cosmic eyeblink, less than 0.1% of galactic history — could have sent probes at a modest 0.1c. At that speed, a single probe reaches the nearest stars in about 40 years. That’s within a human lifetime, let alone a civilizational one.

With self-replication — Von Neumann probes that arrive at a star system, mine local materials, build copies, and launch them toward the next system — you get exponential spread. In standard self-replicating-probe models (Tipler, Barlow, and others have worked through the math), the entire galaxy gets saturated in roughly 10⁶ to 10⁷ years, depending on replication time at each stop. That’s somewhere between 0.01% and 0.1% of the galaxy’s age. A rounding error on a rounding error. The question isn’t “can probes get here?” — it’s “why wouldn’t they already be everywhere?”

Anyone making the speed-of-light objection hasn’t done the arithmetic. I know, because I didn’t — initially.

And the reason I didn’t is itself revealing. A probe at 0.1c takes a million years to cross the full diameter of the galaxy — do the division, 100,000 light-years at 0.1 light-year per year. That feels like an eternity. But the relevant timescale isn’t the travel time. It’s the age of the galaxy. And ten billion is four orders of magnitude larger than one million. Your brain rounds both to the same emotional category — “incomprehensibly long” — but one is ten thousand times bigger. This is computational irreducibility in miniature: you cannot shortcut the arithmetic and arrive at the right answer, no matter how good your physical intuition is.

So my first objection was wrong. Score one for Tyler and Robin. Maybe they’re onto something. But I had one more idea.

My Second Mistake: Civilizational Coherence

Grant that probes can reach distant stars. Grant self-replication. The galaxy fills up in a million years. Fine. But by the time you get signals back, tens of millions of years will have passed. You need to maintain civilizational coherence — the institutional memory, the technological infrastructure, the interest — for timescales that dwarf anything in human experience. No civilization has ever maintained coherence at anything close. Even species don’t typically last that long. Homo sapiens has existed for roughly 300,000 years and we’ve had radio for about 130 of them.

This is a better objection — and it was my second instinct after the speed-of-light argument failed. It’s also wrong, but for an interesting reason.

Consider two models. First, the dandelion model: the probes were never designed to report home. Their purpose is simply to spread — replicate, observe, catalog, maybe monitor for emerging intelligence. The mission is the spreading. A dandelion doesn’t need to be alive when its seeds land. A civilization facing its own mortality might reasonably conclude: “We can’t live forever, but our probes can.” No return address needed.

Second, the distributed civilization model — this is the Hanson-style argument, and probably the strongest. Once you have self-replicating autonomous machines sophisticated enough to navigate interstellar space and build copies of themselves, what distinguishes that network from the civilization itself? Information is what matters, not substrate. If the probes carry the computational essence of their creators, asking “who’s listening back home?” is like asking “who’s listening to your neurons?” The home planet is just the initial condition. It can burn to a crisp and the project continues.

Tyler essentially acknowledges all of this when he frames drone probes as “discretionary surplus allocation” by civilizations operating on timescales where “concrete motives may not be the major consideration.” He’s thinking about this correctly. The coherence objection assumes NASA-style mission architecture scaled up, which is a failure of imagination, not physics.

Two for two. Both of my objections had failed. Tyler’s framework had survived everything I’d thrown at it, and I was starting to think he and Robin might actually be right. Maybe I needed to revise my priors upward. So I went deeper into the physics — and that’s when I found something.

The alien probe question is computationally irreducible — you can’t shortcut from “physics permits it” to “these sightings are aliens.” There are too many intermediate steps, too many conditional probabilities, too many physical constraints that need to be checked one by one. And the deceleration constraint — the one that resolves the version of this question people usually argue about — requires you to sit down and do the calculation. I have not seen it foregrounded in the UFO/UAP debate. So let me try.

The Deceleration Problem: You Can’t Quietly Park at 0.1c

The problem isn’t getting to 0.1c. It’s stopping.

Getting to 0.1c is hard but there are at least hand-wavy schemes: solar sails with giant laser arrays, fusion drives, pellet runways. But every acceleration method has a deceleration problem that’s equally bad or worse, because of the tyranny of the rocket equation.

If you carry your own fuel, the math is exponential. You need fuel to decelerate, but you had to accelerate that deceleration fuel, which means you needed more fuel to accelerate, which you also had to accelerate, and so on. For fusion drives (optimistic exhaust velocity ~0.05c), a total delta-v of 0.2c — accelerate to cruising speed, then brake to a stop — gives a mass ratio of e⁴ ≈ 55. That means 98% of your launch mass is fuel. (This is the non-relativistic Tsiolkovsky approximation; at 0.1c it’s still good as back-of-the-envelope.) For chemical propulsion, the number of zeros breaks the notation. Don’t bother computing it.

If you don’t carry your own fuel, you face the Forward problem: Robert Forward’s idea of a solar sail accelerated by a laser array back home, braking by detaching part of the sail as a reflector. This requires a laser of staggering power (whether or not Earth would see it depends on wavelength, beam geometry, and sidelobes), sail engineering of extraordinary precision, and someone back home actively pointing the laser at your destination for years.

But here’s the killer constraint. A 100-kilogram craft arriving at 0.1c carries about 4.5 × 10¹⁶ joules of kinetic energy — roughly 10 megatons of TNT. About 700 Hiroshimas. You do not get to hand-wave that away.

Could you bleed it off slowly? If you spread the braking over 10 years, the average power is only about 140 megawatts — a modest power plant, not a bomb. But the stopping distance is still enormous: at constant deceleration from 0.1c over a decade, you need roughly half a light-year of runway. That’s tens of thousands of AU. An entirely different class of object than “a thing hovering over an airport.”

Solar radiation pressure doesn’t rescue you either. At 1 AU from the Sun, even a perfect reflector feels only about 9 micronewtons per square meter. Decelerating a 100-kilogram payload on a decadal timescale pushes you toward sail areas around a square kilometer — and you’d have to begin slowing down absurdly far out, where the sunlight is weaker still. Magsails and electric sails change the mechanism, not the moral. The common feature of every serious deceleration scheme is a huge interaction cross-section, a long braking baseline, or both.

A 0.1c probe trying to “arrive quietly” is like a bullet trying to stop inside a room without hitting anything. You can do it — by firing it into a giant block of gel the size of the room. But then the gel is the story. And we don’t see any gel. (If we did see the gel, the New Jersey footage would finally be the least interesting part.)

The flyby vs. hover problem makes this airtight. For the UAP question, the options split into two broad families. Flyby: physically easy — you just keep going. Breakthrough Starshot is explicitly designed to do this: gram-scale chips pushed to 0.2c by ground-based lasers, streaking past Alpha Centauri in hours. But a flyby at 0.1c crosses the inner solar system in hours and is gone. It does not loiter over populated areas. It does not hover near airports. A flyby is a streak, not a sighting. Rendezvous: brake to a stop. This requires all the enormous apparatus we just described — massive fuel, huge sails, tens of thousands of AU of braking distance. None of which is consistent with a compact object showing up unaccompanied near Earth.

“Atmospheric loitering” is not a flyby signature. And it’s not a braking signature. It’s neither.

The only escape hatch for rendezvous is ultralight probes coupled to enormous braking structures. Heller and Hippke worked out what a Starshot-style rendezvous at Proxima Centauri would require: a roughly 76-gram sail, 10-gram payload, sail area around 10⁵ square meters — a sail hundreds of meters on a side, more stadium than spacecraft, attached to a chip — and even then the maximum injection speed for capture is about 4.6% of c, not 10%. The common feature is the same every time: ultralight payloads, huge apparatus, long baselines. Not things you’d film with an iPhone over New Jersey.

But what about slower probes? Everything above assumes relativistic cruise speeds — 0.1c, where the deceleration bill is enormous. But the self-replicating-probe literature doesn’t require 0.1c. Barlow works with 0.01c and still gets galaxy-wide exploration on ~10⁷-year timescales. At 0.01c the kinetic energy drops by a factor of 100 — from roughly 10 megatons to roughly 100 kilotons for a 100-kilogram craft. At 0.001c it falls to roughly 1 kiloton. The rocket equation also becomes much less cruel. So yes: slow Von Neumann probes could stop. They could be compact. They could be here.

But notice what changed. The problem is no longer “how did they brake?” The problem is “where is the archaeology?” If slow self-replicating probes have been expanding through the galaxy for millions or billions of years, the expected evidence is not ambiguous atmospheric footage. It is persistent structure: processed asteroids, factories, waste heat, communication nodes, orbital artifacts, spectral anomalies, or some repeatable astronomical signature. And in the actual slow-probe literature, a self-replicating probe is not a magic seed — it is an industrial system. Barlow’s model requires a collection of robots, assemblers, and factories. Slow probes make stopping easier, but they make the absence of persistent infrastructure harder to explain.

So the slow-probe objection doesn’t rescue the UAP inference. It moves the evidentiary burden. Fast probes should look like flybys or giant braking structures. Slow self-replicating probes should look like Solar System archaeology. Neither looks like fuzzy orbs over populated areas.

“But what if a probe arrived long ago, parked somewhere in the Solar System, and has been lurking?” Fair — that’s a genuinely different hypothesis, and the physics doesn’t rule it out. A probe that decelerated over a thousand years on a long interstellar baseline and has been sitting quietly in the asteroid belt since the Pliocene is allowed. But notice what that hypothesis commits you to: you should expect orbital anomalies, thermal glints, spectral signatures, persistent material traces — the signatures of a parked artifact. What you should not expect is ambiguous low-altitude videos. The lurker hypothesis and the atmospheric-sightings hypothesis are not the same hypothesis, and evidence for the former is not evidence for the latter.

“But what if probes arrive and then build larger structures from local materials?” Two problems. First: how? This requires a full autonomous industrial stack — mining, refining, fabrication — in an environment with no biosphere help. The first Fermi paradox is “where are the probes?” This escape hatch has a second Fermi paradox hiding inside it: if they built factories, where are the factories? Second: even if you grant the mechanism, large sustained construction from asteroid materials should plausibly leave signatures: thermal emission, spectral anomalies, unusual mass distributions, persistent orbital oddities. No such evidence appears in the UAP record people are actually arguing about.

Tyler seems to intuit something like this when he notes the probes look like “cheap crap.” He’s picking up on the right signal. At relativistic speeds, quiet stopping pushes you toward flyby chips or giant sails. At sub-relativistic speeds, stopping is feasible — but then you need to explain the absence of the persistent signatures that millions or billions of years of self-replication should have left. Either way, the physics pushes in the same direction Tyler’s instinct suggests: away from the interpretation that atmospheric UAP sightings are alien probes.

I’m not arguing against probes. I’m arguing against inference from these sightings to probes.

The deceleration calculation handles the fast case — that’s a mechanics argument. The infrastructure and behavioral arguments handle the slow case — that’s an archaeology argument. They’re different kinds of reasoning, but they do the same Bayesian work: at any speed, the observed UAP evidence doesn’t match the probe models being invoked in this debate — fast or slow, flyby or parked, nano or macro. But what if probes didn’t need to travel at sub-light speeds at all? What if there’s a shortcut?

But What About Wormholes?

No.

I should probably say more than that. The sophisticated reader’s objection at this point is: “What about wormholes? Alcubierre drives? Some exotic shortcut we haven’t thought of?” Every worked-out shortcut we know how to write down comes with the same bill: exotic stress-energy, negative energy, or violations of the usual energy conditions. The Einstein field equations — the same equations that predict gravitational lensing, gravitational waves (confirmed by LIGO in 2015), GPS corrections (confirmed daily by your phone), and black hole shadows (confirmed by the Event Horizon Telescope in 2019) — are among the most precisely tested in all of science.

A wormhole is a solution to these equations — a tunnel connecting two distant points in spacetime. Einstein and Rosen found such a solution in 1935, but their wormhole wasn’t traversable: it would pinch shut faster than anything could pass through. In 1988, Kip Thorne — who later won the Nobel for gravitational wave detection and served as the science consultant on Interstellar — and his student Michael Morris asked a different question: what would you need to hold a wormhole open?

They worked backward from the requirement of traversability and derived what the energy-momentum tensor would have to look like. The answer was stark: you need matter with negative energy density. Not low energy. Not zero. Negative. Physicists call it “exotic matter.” Normal matter curves spacetime in a particular way. To hold a wormhole open, you need matter that curves it in the opposite direction.

How much? The numbers vary wildly depending on the metric. Early Morris-Thorne-style constructions demanded planetary to stellar masses of the stuff. Later work by Visser, Kar, and Dadhich showed you can construct wormhole geometries with arbitrarily small total energy-condition violations — but subsequent analyses by Fewster and Roman argued that such minimal-violation wormholes are either submicroscopic, badly curved, or otherwise severely constrained. The robust claim isn’t any particular mass estimate. It’s that every known traversable wormhole solution requires violation of the null energy condition, and at engineering scales — the scales at which you could send a probe through — nobody, human or alien, has ever produced, detected, or demonstrated a mechanism for generating or stabilizing such stress-energy. This isn’t a technology gap. It’s a physics gap. The irony is worth savoring: the man who showed Hollywood what a traversable wormhole would look like is the same man whose equations show what it would actually take to build one.

“But the Casimir effect proves negative energy exists!” Yes — and this is where the scale problem bites. Casimir negative energies occur in gaps of nanometers between parallel plates, involve tiny total amounts of energy, and aren’t controllable or transportable in any useful sense. A traversable wormhole demands macroscopic negative energy — stable, concentrated, engineerable, held in a specific geometric configuration. We can measure Casimir effects in a tabletop experiment. We cannot scale them up, shape them, or make them persist. It’s the difference between observing a droplet of water and engineering a dam: both involve water, but nothing about the droplet tells you the dam is physically achievable. And no civilization, however advanced, gets to ignore the energy conditions unless the energy conditions are wrong — and so far, across every experimental regime we’ve ever tested, they’re not.

Alcubierre-style warp drives have the same problem. Later geometries can reduce the formal energy cost, but they do not remove the need for exotic stress-energy that we do not know how to create, confine, or stabilize. The direction is always the same: these solutions require violations of the energy conditions at scales demanding entirely new physics, not just better engineering.

So: sub-light travel works in principle, but quiet rendezvous constrains you to ultralight payloads coupled to huge braking structures, or to lurkers whose signatures should be in orbit. Superluminal shortcuts require stress-energy we do not know how to produce and that has never been observed anywhere in the universe, and require physics that is, with a good deal of confidence, ruled out by already-known physics. If rendezvous-capable probes are here at all, they’re gram-scale chips with stadium-sized sails, or they’re parked artifacts discoverable by telescope — not things people are filming with iPhones.

Tyler acknowledges you have to be “short usable wormholes.” Fair. I, too, am short usable wormholes. But the deceleration problem bites even without wormholes, and that’s the piece I think completes the picture.

What Are They Actually Doing?

Even if you ignore all the physics — even if you grant every generous assumption — the observed behavior of these objects is inconsistent with any coherent model of what advanced alien probes would do.

What we actually observe: things moving in ways we can’t readily explain. That’s it. Tyler has recently argued that debunkers who chalk everything up to camera tricks and parallax have been “proven wrong in general,” and that Mick West “is not your proper guide here.” I agree. That’s not the argument I’m making. The newly released government footage shows more of what we’ve always seen — glowing orbs, fast-moving objects, unpredictable trajectories. Grant that all of it is real. Grant that these are physical objects doing exactly what they appear to be doing. Now ask: is any of this behavior consistent with an interstellar probe? They’re not self-replicating. They’re not mining resources. They’re not building copies of themselves on the Moon. They’re not doing anything consistent with any version of the probe hypothesis — not the dandelion model, not the distributed civilization model, not even Tyler’s “seek out power sources and observe” model. They’re just... there. Doing stuff that looks weird on camera.

Two options, both terrible. Option A: they want to be seen. Why? If you can build interstellar probes, you can build sensors that dominate anything in our current atmosphere — so “hovering low and visible” is a strange choice unless the point is theatre. Robin’s “domestication” theory is a creative attempt to address this, and it’s exactly the kind of detailed model-building that forces careful thinking about what each assumption costs in probability. But each additional assumption narrows the parameter space considerably.

Option B: they don’t want to be seen but are failing. A system that solved autonomous interstellar navigation, self-replication, and million-year durability can’t manage basic stealth against a species that figured out powered flight 120 years ago? That’s like someone who built a functioning warp drive being unable to turn their headlights off.

Even the Navy’s Tic Tac footage — the strongest single case the UFO community has — remains officially unresolved. ODNI’s public position is that there isn’t sufficient information to attribute most incidents to specific explanations. The most recent AARO report found no evidence of extraterrestrial beings, activity, or technology. NASA’s public FAQ says the same. But “unresolved” isn’t evidence for any particular hypothesis. Nothing about the Tic Tac behavior points specifically toward alien probes: no mining, no replicating, no construction, just an oblong shape maneuvering near a military exercise. “We don’t know what it was” and “it was an interstellar probe” are not the same statement.

What the observations are consistent with: misidentified terrestrial objects, sensor artifacts, classified military programs, good old-fashioned observer error. Exactly the boring explanations. The observed behavior is exactly what you’d expect from mundane sources and exactly what you’d not expect from technology millions of years more advanced than ours.

Tyler asks “why don’t we see more of them?” and calls this a “perfectly valid (and embarrassing) question” for the hypothesis. He’s right. But the question isn’t just about frequency — it’s about behavior. The behavior is what’s diagnostic.

Bayes’ Theorem Finishes the Job

Time to formalize this. Three decades of quantitative finance teaches you one thing above all: the prior is not the posterior.

Prior: P(alien probes exist in our solar system) — non-negligible. We’ve granted this throughout. The physics of sub-light self-replicating probes plus galactic timescales makes this a legitimate nonzero number. Tyler’s framing here is essentially correct.

Likelihood: P(we observe exactly this UAP behavior | these are alien probes) — essentially zero. Rather than pull a single number from the air, let’s decompose this into roughly separable suppressions — weakly dependent but directionally reinforcing. The specific numbers below are illustrative, not measurements; the point is the structural form of the argument:

L(size): What’s the probability that an interstellar probe manifests as a compact atmospheric craft? At relativistic speeds, deceleration physics pushes toward ultralight payloads with huge sails — not compact craft. At slower speeds, compact probes are feasible — but after billions of years of self-replication, the signatures should be infrastructure, not isolated atmospheric objects.

L(loiter): What’s the probability it loiters visibly in the atmosphere rather than remote-sensing from orbit? A technology millions of years advanced choosing to hover over airports?

L(ambiguity): What’s the probability it produces ambiguous, fuzzy-video evidence rather than being either completely invisible or unambiguously present? The one thing a million-year-old technology wouldn’t produce is “weird stuff on camera that we can’t quite explain.”

Even if you’re generous and give each factor a 1% probability — which is far too high for L(size) given the physics — the product is 10⁻⁶. One in a million. Even if you think I’m double-counting, even if these factors aren’t fully independent, you’d have to increase each term by orders of magnitude to escape the conclusion.

The posterior collapses. Give the prior a generous 5%. Your posterior is 5% × 10⁻⁶ = 5 × 10⁻⁸. One in twenty million. That’s not “appreciably non-zero.” That’s noise. (This is a rough calculation, not a formal posterior — but including mundane alternatives in the denominator only makes it worse for the probe hypothesis, not better.)

Trading analogy: this is like saying “this company has great fundamentals” while ignoring that it’s priced at 500x earnings. The prior isn’t the posterior. Every trader who’s blown up learned this the hard way.

And this is where Tyler’s candor on the Sullivan podcast becomes genuinely illuminating. “It depends on my mood of the day... There’s nothing very scientific about any number.” Tyler is being transparent that his 10% is a prior — an informed intuition shaped by Fermi-paradox reasoning and genuine curiosity, not the output of a physics calculation. And he’s right to frame it that way. A prior is where you start. But it’s not where you finish. Tyler and I agree on the direction — he already thinks it’s probably not aliens, putting it at 10%, not 90%. The question is the magnitude. And the physics resolves it: the prior was the right place to start. The likelihood is where the story ends. When you combine Tyler’s honest prior with the likelihood the physics gives you, the posterior is crushed to the floor.

The Missing Calculation

The broader discourse on UAPs is almost entirely authority-based. The Free Press launched a four-part series on “What Should Smart People Think About UFOs?” — interviewing journalist Michael Shellenberger, physicist Avi Loeb, oceanographer Tim Gallaudet, columnist Ross Douthat, and religion professor Diana Pasulka. The series is well-produced and earnest, but not one of the guests, based on their public work, engages the deceleration problem. (Loeb does physics, but his published UFO-adjacent arguments focus on anomalous objects like ʻOumuamua, not on the rendezvous problem.) Obama said there’s “footage and records of objects in the skies that we don’t know exactly what they are.” Whistleblowers testified under oath. Senators seem convinced. The calculation that would actually resolve the question hasn’t been part of the conversation.

(The Free Press has since covered the Trump administration’s UFO file release — more ambiguous footage, more debate, same pattern.)

This is what I’ve called “conclusions without derivations” — accepting an endpoint because it sounds plausible, without checking whether the intermediate steps actually work. (I explore this failure mode in depth in The Science of Free Will.) The conclusion (”probes could theoretically be here”) is defensible. The derivation (”therefore these specific sightings are probes”) was never completed.

Tyler and Robin are doing something fundamentally different from that discourse — and it’s the reason this piece exists. Most UFO commentary is, to use Wolfgang Pauli’s famous dismissal, “not even wrong” — too vague or incoherent to engage productively. Tyler and Robin built careful, internally consistent frameworks grounded in real physics and economics. Tyler’s “discretionary surplus allocation” framing is genuinely insightful about how advanced civilizations might behave. Robin’s grabby aliens model is serious cosmological reasoning. Their frameworks are what made it possible for me to identify the missing step — because without a real argument, there’s nothing to build on. If I have seen further here, it is because I stood on their shoulders.

I would never have thought carefully about the deceleration problem, or worked through the momentum-transfer scaling, or connected the Heller-Hippke rendezvous constraints to the UAP question, if Tyler’s argument hadn’t been careful enough to resist easy dismissal. The cheap objections — speed of light! they’d be dead! — all fail, as we’ve seen. Only by engaging seriously with the strongest version of the argument do you find the constraint that actually resolves it.

The Real Mystery

Here’s what Tyler and Robin have gotten right, and it’s important: the Fermi paradox remains genuinely puzzling, for exactly the reasons they’ve explored. The physics is permissive. Sub-light self-replicating probes could fill the galaxy in a million years. One million years out of ten billion. The prior really is non-negligible.

So the question isn’t “are those sightings alien probes?” The physics says: almost certainly not, for multiple reinforcing physical and observational reasons.

The question is: where ARE all the probes that should be here?

That points toward genuinely deep problems: the Great Filter, the motivation problem, or something about the nature of intelligence we don’t yet understand. Perhaps intelligence turns inward rather than outward. Perhaps the conditions for it are far rarer than Drake equation optimists assume. Perhaps the answer involves physics we haven’t yet formulated.

Tyler’s instinct to take this seriously was exactly right. Robin’s willingness to commit to specific probabilities and build a detailed, testable model is exactly what serious reasoning about hard problems looks like. This is how intellectual progress works: someone builds a careful framework, someone else engages seriously enough to find the step that was missing, and in finding it you learn something neither of you knew before. I didn’t understand the deceleration problem before Tyler made me think about this. I hadn’t connected Heller-Hippke to the UAP question before Robin’s framework forced me to close every escape hatch.

Tyler’s framework was sound. The likelihood calculation completes it. And the answer it gives is: essentially zero.

That’s the question worth a real derivation. Not a conclusion.

Three things are true:

1. Interstellar probes are physically plausible.

2. Ambiguous atmospheric videos are not good evidence for them.

3. The real mystery is not why the UAP footage is blurry. The real mystery is why the Solar System is not full of unmistakable artifacts.

This post draws on ideas from my book, The Science of Free Will, which explores how computational irreducibility, physics, and first-principles reasoning intersect with questions about agency, consciousness, and meaning — including why “conclusions without derivations” is the most common failure mode in human reasoning, and why the universe doesn’t care how sophisticated your intuition is. Available on Amazon or anywhere else books are sold.

In April 2025, gunmen walked into the Baisaran Valley near Pahalgam—a meadow so cinematic that Bollywood has used it as a backdrop for love songs—and asked twenty-six tourists a simple question before shooting them: Are you Hindu?

India named its retaliatory military operation “Sindoor.”

Sindoor is the vermilion powder Hindu married women apply to their hairline as a sign of marriage. India named a bombing campaign after the symbol of married womanhood because the Pahalgam attack had left wives widowed.

Then the ceasefire came. Then Donald Trump, asked about the crisis, said the two countries had been fighting “a thousand years”—an observation that is not only historically wrong by approximately 953 years but also somehow captures the exact quality of intellectual engagement the Kashmir dispute has attracted since 1947.

This is where we are in 2026. This is where we have always been.

And if you want to understand why we are here—why two nuclear-armed democracies (one of them nominally so) have been unable to resolve a territorial dispute about seventeen million people for nearly eight decades—you have to go back. Not a thousand years. Seventy-eight. And before that: one week.

The Week Britain Owned a Country

Here is the founding document of the Kashmir problem, and it is, at its core, a livestock invoice.

On March 9, 1846, the Sikh Empire—having lost the First Anglo-Sikh War—signed the Treaty of Lahore. Because the defeated Sikhs could not pay the war indemnity imposed upon them, they ceded to the British East India Company “the hill countries, including the Provinces of Kashmir and Hazara” as equivalent payment. One crore rupees’ worth of mountains, rivers, valleys, and several million people who were not consulted.

Seven days later—literally seven days—on March 16, the British signed the Treaty of Amritsar, which transferred those same mountains, rivers, valleys, and uncontested millions to Gulab Singh, the Raja of Jammu. Price: seventy-five lakh rupees.

The annual tribute was as follows: one horse, twelve shawl goats of approved breed (six male, six female), and three pairs of Kashmiri shawls.

No Kashmiri signed. No Kashmiri was asked. Britain had held legal sovereignty over Kashmir for seven days before selling it, and one is entitled to wonder whether they had the legal right to sell what they’d seized from people who’d seized it from the Afghans who’d seized it from the Mughals who’d been contesting it since forever. But nobody asked that question in 1846, because nobody was asking that kind of question in 1846.

Gulab Singh paid the money, received a receipt, and became the Maharaja of a Muslim-majority state ruled by a Hindu dynasty. His great-grandson, Hari Singh, would inherit this arrangement—and then have to decide what to do with it when the British finally left a century later.

He was not well prepared for the responsibility. As we shall see.

Mr. A

In 1919, the young Crown Prince of Kashmir visited Europe for the first time, traveling with approximately four million dollars’ worth of cash and jewelry and accompanied by his aide-de-camp, an Irishman named Captain C.W.A. Arthur.

Captain Arthur introduced the prince to a charming London widow named Maudie Robinson, described in contemporary accounts as “bewitching.” The prince and Mrs. Robinson went to Paris for Christmas. On Boxing Day, a man burst into their hotel room, introduced himself as her husband, and demanded an enormous settlement to avoid citing the prince in divorce proceedings.

It was a classic honeytrap. Captain Arthur, it turned out, was in on it. The prince wrote two cheques totaling £300,000—equivalent to roughly £14.5 million today—and paid.

The scandal eventually ended up in London’s King’s Bench Division in 1924. At this point, the India Office intervened and asked the court to keep the young ruler’s identity secret, because the optics of an Indian prince paying what amounted to the GDP of a small county to emerge from a Parisian honey-trap were not considered ideal for British India’s image. The prince was referred to throughout the proceedings as “Mr. A.”

Lord Birkenhead, the Secretary of State for India, had specifically requested this discretion. The story was so explosive that the India Office classified its files for one hundred years—not the standard thirty—”because espionage was involved.”

In the end, other Indian princes complained that the secrecy was generating damaging rumors about them. On December 3, 1924, the India Office finally permitted the disclosure of his identity. It was first announced on London radio station 2LO, which interrupted its music programming for the news. Time magazine ran the full story. The Daily Express, the Daily Chronicle, and papers in Australia, Jamaica, and beyond all published.

When the prince returned to India, his uncle—still the reigning Maharaja—sent him to a remote jungle estate for six months of religious penance. The terms of his humiliation included the shaving of his moustache.

This man—Maharaja Hari Singh, the former Mr. A, the shaved-moustache penitent, the man whose full official title was Shriman Inder Mahinder Rajrajeswar Maharajadhiraj Shri Hari Singhji Jammu & Kashmir—was the person who, twenty-three years later, would sign the document upon which India’s entire legal claim to Kashmir rests.

History, as a rule, is not run by the best-prepared people.

The Signature That Launched Seventy-Eight Years of War

By August 1947, Hari Singh had a problem. Actually, he had several problems, but the main one was this: the British were leaving, and every princely state had to decide which of the two new countries to join. Hari Singh didn’t want to join either.

He had signed a standstill agreement with Pakistan. He had tried to get India to sign one too; India declined. He kept hoping for a third option—independence, with friendly relations with both sides. He wrote to Mountbatten explaining his position: Kashmir had ties to both dominions, shared borders with Soviet and Chinese territory, and he needed time.

Mountbatten told him to make up his mind. Hari Singh did not make up his mind.

On October 22, 1947, Pakistan launched “Operation Gulmarg.” Pashtun tribal fighters from the Northwest Frontier Province poured across the border toward Srinagar, traveling in, as Hari Singh later described, “motor trucks.” They were armed, organized, and moving fast. Baramullah fell. Eleven thousand people were killed there. The tribesmen were within hours of Srinagar.

Hari Singh panicked and fled his summer capital for Jammu by car.

He sent an urgent message to India requesting military assistance. India’s response, delivered by V.P. Menon, was essentially: Sign first. Troops later.

On October 26, 1947—according to the official account, at some point after Menon flew to Jammu carrying the Instrument of Accession—Hari Singh signed. Menon flew back to Delhi. The next day, Indian troops were airlifted to Srinagar airport.

This is the document. This is the signature. This is India’s legal title to Kashmir.

Now observe the structure of the paradox:

Pakistan sent in the tribal raiders to prevent Kashmir from going to India. The invasion panicked Hari Singh into requesting Indian military help. India made accession the condition of help. Pakistan’s invasion created the document that India uses to justify holding the territory that Pakistan wanted.

Pakistan called the document fraudulent. Pakistan is the reason the document exists.

Even Mountbatten, accepting the accession on October 27, added a caveat: once law and order is restored and the invader cleared, “the question of the State’s accession should be settled by a reference to the people.” India’s own first acceptance of Kashmir contained a promise of a vote.

Hari Singh was subsequently banished from the state he’d just signed away. He departed Jammu in 1949 and never returned. He died in Bombay in 1961, reportedly farming polo ponies.

Shriman Inder Mahinder Rajrajeswar Maharajadhiraj Shri Hari Singhji ended his days raising horses in Bombay.

The Lawyer and the Theologian

Here is the core of the Kashmir dispute, and it is almost too neat to be real—but it is real, and it has been real since 1947, and almost nobody states it this plainly.

India’s claim is legal. Kashmir is Indian because Hari Singh, a ruling Maharaja with recognized legal authority, executed the Instrument of Accession under the established procedures of the Indian Independence Act 1947. The process was followed. The paperwork is valid. The border is settled.

Pakistan’s claim is demographic-religious. Kashmir is Pakistani because Kashmir’s population is overwhelmingly Muslim, and the entire logic of Partition—the two-nation theory, the founding principle of Pakistan—is that the Muslims of the subcontinent constitute a separate nation and deserve their own state. Muslim-majority areas belong to Pakistan. This is literally why Pakistan exists.

India argues like a lawyer. Pakistan argues like a theologian.

Neither is entirely wrong. That’s the actual problem.

Pakistan’s claim to Kashmir is the direct logical extension of its own founding philosophy. You cannot say “Muslim-majority territories belong to Pakistan” as your reason for existing and then refuse to apply that logic to the largest Muslim-majority state that didn’t go to Pakistan. Pakistan is being consistent. Annoyingly, infuriatingly consistent.

India’s claim requires you to accept that the legal acts of Hindu hereditary monarchs determine the nationality of Muslim-majority populations without their consent—which is... also a defensible legal position, and also somewhat uncomfortable.

But here is where it gets genuinely dizzying. Stick with me.

India’s legal position depends on the principle that religion is irrelevant to Indian identity. You are Indian because you are Indian—because you live in India, because a legal instrument incorporated your territory. It doesn’t matter what religion you are, what language you speak, what ethnicity you belong to. India is a secular state. That’s the whole point. That’s Nehru’s entire project. That’s the constitutional architecture.

And India’s current government—the BJP, whose ideological roots run through the RSS and whose project is explicitly Hindu nationalist—has spent the past decade suggesting, through its actions, rhetoric, laws, and bulldozers, that actually, religion is quite relevant to Indian identity. That there are “infiltrators.” That certain people are “termites.” That the Citizenship Amendment Act should fast-track everyone’s citizenship except Muslims from neighboring countries. That temple-mosque disputes have a correct answer that happens to align with Hindu claims. I explored this domestic rewriting in Post 18A, The Third Rail—the slow, grinding conversion of a secular republic into something its founders would not recognize.

Which means India is simultaneously arguing:

“Religion doesn’t determine nationality—that’s why Kashmir is India” (the secular-legal Kashmir argument)

and

“Religion does determine who really belongs here” (the domestic Hindutva argument)

You cannot hold both positions. They are mutually exclusive. The party currently in power in New Delhi is using the logic of Pakistan’s founding philosophy domestically while using the logic of India’s secular founding philosophy internationally to justify holding Kashmir.

The intellectual whiplash is audible from space.

The Algorithm That Cannot Terminate

On January 1, 1948, India took the Kashmir dispute to the United Nations Security Council. This seemed reasonable. The UNSC passed Resolution 39. Then Resolution 47. The resolutions were quite clear: Pakistan would withdraw its forces and tribal fighters; India would reduce its forces to minimum necessary levels; a free and impartial plebiscite would be held.

Both India and Pakistan said yes, great, wonderful, very much in favor of this.

Then they negotiated about the sequence.

Pakistan said India needed to go first. India said Pakistan needed to go first. Both parties claimed the other had not fulfilled the preconditions for the next step. The UN appointed mediators. Five of them, in fairly rapid succession. The mediators reported back that the situation was complex.

It is now 2026. The United Nations Military Observer Group in India and Pakistan—UNMOGIP—still exists. Its headquarters rotates between Islamabad in winter and Srinagar in summer. A UN audit document summarizes the current state of play: India maintains UNMOGIP’s mandate has lapsed following the 1972 Simla Agreement. Pakistan maintains it remains essential. The UN Secretary-General’s position is that only a Security Council resolution can terminate it. India disagrees. Pakistan disagrees with India’s disagreement.

This is the geopolitical equivalent of two software teams in a perpetual deployment freeze because each insists the other must merge their branch first. Except instead of sprint reviews, there are wars. Three full ones. Multiple partial ones. A four-day conflict in May 2025 that ended in a ceasefire—except India says the ceasefire was not brokered by anyone, it was entirely bilateral, and Pakistan says it was brokered by Donald Trump, and Pakistan nominated Donald Trump for the Nobel Peace Prize.

Pause here.

India says nobody brokered it. Pakistan says Trump brokered it and deserves a Nobel. Trump, who had said the two countries had been fighting “a thousand years,” apparently solved in four days what seventy-nine years of diplomacy could not—or didn’t, depending on which of the two parties you ask.

Then the plot thickened in ways a novelist would be embarrassed to submit to an editor.

The relationship between Modi and Trump had been, until approximately May 2025, one of the great bromances of authoritarian-adjacent governance. “Is baar, Trump ki Sarkar”—”This time, it’s Trump’s government”—Indian BJP supporters had chanted when Trump won in 2024, with a warmth that suggested they viewed his victory as somehow also theirs. The two men had held rallies together. They had exchanged compliments of extraordinary sincerity. Trade negotiations had been proceeding with the ambient goodwill of men who genuinely liked each other’s energy.

And then came Kashmir. And then came the ceasefire India says didn’t happen. And then—in the coup de grâce, the chef’s kiss, the final absurdity that completes the collection—General Asim Munir, the Pakistan Army Chief who had given that speech about Hindus and Muslims being fundamentally different peoples, the man whose theological worldview makes the Taliban look nuanced, the architect of Pakistan’s “moral and diplomatic support” for the groups India had just bombed—that man was received at the White House with great acclaim.

The man Modi had implicitly accused of sponsoring the killing of twenty-six Hindu tourists was a guest of honor in Washington within months.

You genuinely cannot make this up. India’s closest great-power friendship, shattered. Pakistan’s General, feted. The Nobel nomination, pending. The ceasefire, simultaneously real and fictional depending on your nationality.

The Kashmir dispute does not merely resist resolution. It actively destroys the attempts.

And yes, the plebiscite was promised. By Mountbatten in his acceptance letter. By Nehru at the UN. By multiple Security Council resolutions. The people most enthusiastically promising the plebiscite were the same people most enthusiastically not holding it.

“Geopolitics,” as a concept, should probably come with a disclaimer about implied promises.

“Thence North to the Glaciers”

Not all of Kashmir’s absurdities are philosophical. Some are literally ink on paper.

The 1949 ceasefire agreement described the line between Indian and Pakistani positions in meticulous detail—named points, coordinates, measured distances. And then, at a certain point near coordinate NJ9842, the text continued: “thence north to the glaciers.”

“Thence north to the glaciers.”

That phrase—that lazy, elegant, catastrophically incomplete phrase—is why the Siachen Glacier became a battlefield. The ceasefire line stopped at a named coordinate, and beyond that, the treaty offered only vague directional poetry. When strategic incentives changed in the 1980s, both India and Pakistan decided “north to the glaciers” meant something favorable to their position. India moved first, in 1984. Pakistan protested. Both countries have maintained forces at altitudes above 6,000 meters, where temperature regularly drops to minus fifty degrees, ever since.

The strategic analyst Stephen P. Cohen once described the Siachen dispute as “a struggle of two bald men over a comb.” The glacier has no strategic value, no resources, no population, no particular beauty that couldn’t be found somewhere warmer. It is an uninhabited refrigerator at the edge of the atmosphere, and it has been militarized because someone in 1949 couldn’t be bothered to specify a coordinate north of NJ9842.

That’s not a metaphor. That’s the actual explanation.

The Missing Protagonist

There is a population of approximately thirteen million people in the Indian-administered portion of Jammu and Kashmir. There are roughly four million more in Pakistani-administered Azad Kashmir—”Azad” being the Urdu word for “free” or “independent,” which is either the most optimistic place name in human history or the most ironic, depending on your familiarity with how Pakistan administers it. These people have been argued about, administered, militarized, censused, democratically excluded, and periodically shot at since 1947. Their preferences are rarely the first item on any agenda.

When they have been asked what they want, the answers are inconvenient for everyone.

A Chatham House poll found that support for independence in the Kashmir Valley runs between 74 and 95 percent. Support for joining India in the Valley: approximately 2 percent. Support for joining Pakistan: approximately 6 percent. A Hindustan Times survey found that 66 percent of Valley residents wanted “complete freedom to entire [sic] Jammu and Kashmir as a new country.” Another poll, conducted in Srinagar specifically, found that 90 percent wanted independence, 7 percent wanted India, and 3 percent wanted Pakistan.

The option that both India and Pakistan are disputing—join us—is, in the area most people think of as “Kashmir,” the least popular option on offer.

This is the definitional absurdity at the center of the dispute. Two countries have been fighting for seventy-nine years over who gets to govern a population that, to a remarkable degree of consistency across decades of polling, wants neither of them.

The Kashmiris want to be left alone. This is the one option neither country is willing to offer.

After the Pahalgam massacre in April 2025—after the gunmen asked tourists their religion before pulling the trigger, after Operation Sindoor, after the Indian Army deployed hundreds of thousands of troops and arrested thousands of Kashmiris in the aftermath—ordinary Kashmiris in Srinagar came out to mourn the dead and condemn the killings. The Muslim majority of the Valley, with striking consistency, refused to become the anti-Indian mob that Pakistani state narrative requires them to be. They condemned the attack. They lit candles.

They do this every time. And every time, they are ignored—because they are not useful to either national narrative as complex human beings. They are useful as a population to be governed, a statistic to be cited, a grievance to be deployed.

The Thing Nobody in Delhi Can Say

Let me return to the philosophical contradiction, because it matters more now than it did in 1947.

In August 2019, the Modi government revoked Article 370—the constitutional provision that had given Jammu and Kashmir special autonomous status since 1949. The legal mechanism was exquisite. Since the state’s Constituent Assembly (which had to give consent to constitutional changes) had dissolved in 1957, the government issued a Presidential Order that simply redefined the phrase “Constituent Assembly” to mean “Legislative Assembly.” Since Kashmir was then under President’s Rule, the “state government” that gave “concurrence” was the Governor—appointed by the Union government.

India asked itself permission. India said yes.

The Supreme Court upheld this in 2023, on the grounds that Article 370 was always intended as a “temporary provision” and that Parliament’s action was constitutional. The Chief Justice noted it was “an interim arrangement due to war conditions in the state.”

Fine. Perhaps it was temporary. Perhaps the revocation was legally valid. Let’s assume it was.

But then observe the coincidence: the argument India uses externally (Kashmir is Indian because of a secular legal document) was simultaneously being strained by the argument India was making internally (the reason Kashmir was special is being eliminated; the state that signed the accession in exchange for specific terms is having those terms unilaterally revised).

Former Chief Minister Omar Abdullah called it “a total betrayal of the trust that the people of Jammu and Kashmir had reposed in India when the state acceded to it in 1947.” That is a fairly precise description. The accession came with conditions. The conditions were modified without consent. The legal instrument that grounded the accession explicitly contemplated a reference to the people. The reference to the people was never made.

You cannot hold that you have an ironclad legal title based on a specific document and then revise the terms of that document without consulting anyone. That’s not how contracts work. That’s not how sovereignty arguments work. That’s not how trust works.

And if you add to this the broader BJP domestic project—the Citizenship Amendment Act that explicitly excludes Muslims, the “anti-infiltrator” rhetoric, the demolition of mosques, the temple politics, the casual redefinition of who is “really” Indian—then the secular-legal foundation of India’s Kashmir argument is being quietly eroded by the same government that insists upon it.

You cannot argue Pakistan’s logic at home and India’s logic in Srinagar.

Pakistan’s Army Chief Asim Munir, just weeks before the Pahalgam attack, gave a speech describing Kashmir as “our jugular vein” and explicitly arguing that Hindus and Muslims are fundamentally different peoples who cannot coexist in a single nation—the two-nation theory, stated plainly in 2025, as if the subsequent seventy-nine years had produced no evidence worth considering.

What Pakistan’s Army Chief said openly, India’s ruling party is implementing quietly.

Both are wrong. Both know it and proceed anyway.

The Quantum State

I’ve argued throughout this series that India exists in quantum superposition—that any statement you make about India is true, as is its opposite. It’s the operating principle of the whole project, stated plainest in Post 1. The Kashmir dispute is quantum superposition weaponized. It is simultaneously:

A legal dispute (but the legal document was signed under duress)

A religious dispute (but both states officially deny this)

A democratic dispute (but the plebiscite was promised in 1947 and not held)

An anti-colonial dispute (but both successor states are using colonial-era legal frameworks)

A human rights dispute (but both sides are implicated in human rights violations)

In 2019, India shut down the internet in Jammu and Kashmir for 175 days—one of the longest internet shutdowns anywhere in the world that year. The people India was suppressing to maintain sovereignty were the people India’s sovereignty was nominally there to protect.

Pakistan’s Army—whose official position is that Kashmir deserves self-determination—has spent decades supporting militant groups that have made normal life in the Valley a continuous emergency, which has been India’s justification for refusing to hold the self-determination vote that Pakistan demands.

Both countries have found ingenious ways to make the thing they claim to want impossible.

Coda: The Polo Ponies and the Paperwork

Hari Singh—Mr. A, the reluctant signatory, the man with the shaved moustache and the spectacular bad judgment in European travel companions—died quietly in Bombay on April 26, 1961.

He had once wanted Kashmir to be independent. He had signed it away at three in the morning (some accounts say he was woken from sleep by V.P. Menon) while Pakistan’s tribals raced toward his capital. The document he signed has been the basis of three wars, dozens of insurgencies, one nuclear standoff, and one brief four-day conflict in 2025 named after the symbol of Hindu marriage.

Kashmir has never been independent. It has never had a plebiscite. It has been administered by India (two-thirds of the territory) and Pakistan (one-third), with China occupying a piece of Ladakh since 1962 because—as long as we’re cataloguing absurdities—why not add a third nuclear power?

The British owned it for seven days before selling it for goats.

The man who ultimately decided its fate was known, for three years, only as Mr. A.

The people who actually live there want neither country.

And the two countries arguing over them are, with increasing symmetry, borrowing each other’s worst ideas: Pakistan borrowing India’s nuclear program, India borrowing Pakistan’s approach to Muslim minorities, both countries borrowing Britain’s talent for treating a mountain valley full of real human beings as inventory.

As of early 2026, UNMOGIP still rotates its headquarters between Islamabad and Srinagar by season. The plebiscite has not been held. The ceasefire line still says “thence north to the glaciers.”

The Siachen Glacier is still militarized.

The shawl goats were presumably long ago discontinued.

Everything else continues.

If you like this, you will enjoy my book, The Science of Free Will. Among other things, I explore why we can’t trade with ants (they can detect cancer), what that has to do with the future of AI, and why we need a Supreme Court if the world is deterministic. Available on Amazon or anywhere else books are sold.

The India Paradox is a series exploring how the world’s most diverse democracy somehow functions despite—or perhaps because of—its beautiful contradictions. Previous posts available at samirvarma.substack.com.

I recently did a two-part podcast series (Part 1, Part 2) with Simon M. of , which inspired me to write this post.

The Byrds had a song: “So You Want to Be a Rock ‘n’ Roll Star.” The advice was straightforward. Get a guitar. Learn to play. Find some people who’ll tell you you’re great. Then get ready to lose everything that made you interesting in the first place.

The advice for aspiring quants is similar. Except for the guitar.

If you want to trade quantitatively — if you want to build strategies, manage risk, understand markets at the level where you can actually make money — you need to learn to program.

Not because you’ll write much code. In the age of AI, the code increasingly writes itself. I can describe a strategy in English and have working Python in minutes. That part of the job is rapidly going to zero.

No, you need to learn to program because programming is the only reliable way to learn computational thinking. And computational thinking is the actual skill.

What do I mean by computational thinking? Stephen Wolfram has written extensively about this, but the short version is: it’s the ability to think in terms of processes, not just states. To ask not “what is the answer?” but “what is the procedure that generates the answer, and what are its failure modes?” To understand that a system can follow perfectly deterministic rules and still produce behavior you cannot predict without running it. (If that last bit sounds familiar, it’s because it’s computational irreducibility — the central concept in my book and arguably one of the most important ideas in modern science.)

Wolfram would add — and he’s right — that you should learn a high-level language like Mathematica or the Wolfram Language, where the abstractions are close enough to thought that you’re expressing what to compute, not wrestling with how to allocate memory. The less distance between the idea and the code, the more the programming teaches you to think. But the specific language matters less than the main point: learn to program. Period.

A physicist thinks about what’s true. A mathematician thinks about what must be true. An economist thinks about what’s optimal. A computational thinker asks: what happens when you actually run it?

That question — what happens when you actually run it? — is the question that separates people who understand markets from people who have theories about markets.

Theories are beautiful. Backtests are seductive. Models are comforting. And then you run it. With real money. In real time. Against other agents who are adapting to you as you adapt to them. And you discover that the map is not the territory. That it was never the territory. That the entire skill of trading is navigating the gap between the two.

Hacker Wisdom

The hacker culture at MIT figured this out decades ago. Not about trading — about computing itself. They encoded their hard-won insights in a peculiar literary form: koans.

If you haven’t read the koans from the MIT AI Lab, you should (be patient, the page takes time to load!). They were composed by Danny Hillis (who would later build the Connection Machine) and concern legendary figures like Marvin Minsky, Gerald Sussman, and Tom Knight. They’re short. They’re funny. And they contain real technical insight wrapped in Zen Buddhist packaging.

Here’s my favorite. A novice was trying to fix a broken Lisp machine by turning the power off and on. Tom Knight, seeing what the student was doing, spoke sternly: “You cannot fix a machine by just power-cycling it with no understanding of what is going wrong.” Knight turned the machine off and on. The machine worked.

This is not a story about hypocrisy. It’s a story about the difference between doing something without understanding and doing the same thing with understanding. Knight power-cycled the machine because he understood what was going wrong. The novice power-cycled it instead of understanding what was going wrong. Same action. Completely different epistemology.

Trading is full of exactly this distinction. Here’s one.

The Master Trader and the Losing Position

A novice was adding to a losing position in a biotech stock that had dropped 20% on an FDA delay.

The Master Trader, seeing what the novice was doing, spoke sternly: “You cannot fix a losing trade by making it bigger with no understanding of why it went against you.”

The Master Trader added to the position.

The position recovered.

Jim Morrison opened “The Soft Parade” by screaming: “You cannot petition the Lord with prayer!”

You cannot petition the market with backtests. But you must do them — just as the devout must pray. The key is to understand what you are doing when you backtest. A backtest is not a petition. It is not “I ran the numbers, therefore the market owes me this Sharpe ratio.” A backtest is a diagnostic — it tells you something about your assumptions, not about the future. Confuse the two, and you are praying for profits.

Some Trading Koans

In that spirit, I offer more koans about the quantitative trading arts. The Master Trader is a figure of longstanding reputation. The Quant is a recent PhD. The Risk Manager is the person who works either for the regulators, in which case he is worse than useless, or for the fund, in which case he is merely useless.

The Quant and the Backtest

A novice came to the Master Trader and said: “I have backtested my strategy over twenty years of data. It has a Sharpe ratio of 3.2 and never had a losing year.”

The Master Trader said: “How many parameters does it have?”

“Only seven,” said the novice. “Each one is necessary.”

“And how many did you try before settling on these seven?”

The novice was silent.

“With enough parameters,” said the Master Trader, “I can fit an elephant. With one more, I can make it wiggle its trunk.”

The novice said: “But this is not a curve fit. I have an economic rationale for each parameter.”

“Ah,” said the Master Trader. “Then you found the rationale before or after you found the parameter?”

The novice was enlightened.

The Risk Manager’s Koan

A student asked the Risk Manager: “What is the worst loss our portfolio can sustain?”

The Risk Manager said: “Our Value at Risk is two percent.”

The student said: “Then we cannot lose more than two percent?”

The Risk Manager said: “No. I said that is the Value at Risk.”

“What is the difference?”

“Value at Risk is the amount you will lose on an ordinary day when you happen to lose money. It tells you nothing about extraordinary days.”

“Then what protects us on extraordinary days?”

“Nothing that can be expressed as a single number.”

The student said: “Then what good is the number?”

The Risk Manager said: “It satisfies the regulators.”

The student was enlightened.

Drescher and the Toaster (for a Trading Desk)

A disciple of another sect — a wealth advisor — once came to the Master Trader as he was reviewing his morning positions.

“I would like you to fill out this risk tolerance questionnaire,” said the outsider, “because I want you to manage your portfolio appropriately.”

The Master Trader took the questionnaire and fed it into the shredder, saying: “I wish the shredder to have the appropriate risk tolerance, too.”

Debug the Strategy

A Quant came to the Master Trader in great distress. “My strategy worked perfectly for three years and has now lost money for six consecutive months. I have checked the code line by line and there are no bugs.”

The Master Trader said: “The code is not the strategy.”

“What do you mean? The code implements the strategy.”

“The strategy,” said the Master Trader, “includes all the other funds running the same code.”

The Quant was enlightened.

The Nature of Alpha

A novice asked the Master Trader: “Where does alpha come from?”

The Master Trader said: “From other traders’ mistakes.”

“Then what happens when the other traders stop making that mistake?”

“Then your alpha was never yours. It was merely on loan.”

“How do I find alpha that is truly mine?”

“You are asking the wrong question,” said the Master Trader. “Ask instead: whose mistake am I?”

On Simplicity

A young Quant, freshly arrived from a machine learning program, presented his model to the Master Trader. It used an ensemble of gradient-boosted trees, a transformer for sequence modeling, and a graph neural network for cross-asset dependencies.

The Master Trader examined it carefully and said: “What does the model do when the Fed raises rates by 75 basis points for the first time in 28 years?”

“It has never seen such an event in training,” admitted the Quant.

“And what does your model do when it encounters something it has never seen?”

“It extrapolates from the nearest—”

“It guesses,” said the Master Trader. “Your model is a very expensive way to guess.”

“Then what should I use?”

The Master Trader showed him a chart with a single moving average on it.

“That’s it?” said the Quant. “A child could do that.”

“Yes,” said the Master Trader. “And it will work on the day nothing else does, precisely because a child could do it.”

Moon Instructions (for a Trading Desk)

A junior trader came to the senior trader and said: “I have found an arbitrage. The ETF is trading at a discount to its net asset value. I will buy the ETF and short the underlying basket.”

The senior trader said: “What is the carrying cost of the short?”

“Negligible.”

“What is the redemption mechanism?”

“I am not an authorized participant.”

“Then who will close the discount?”

“Market forces.”

“You have described a trade that requires someone else to do the work and give you the profit,” said the senior trader. “This is not arbitrage. This is hope.”

The Empty Order Book

A novice was studying the order book, watching bids and offers flicker and vanish.

“The book is full of lies,” he said to the Master Trader. “Ninety percent of these orders will be cancelled before they execute.”

“Yes,” said the Master Trader.

“Then how can I make decisions based on the order book?”

“You cannot,” said the Master Trader. “But neither can anyone else. And that is the information.”

Sussman Attains Enlightenment (Hedge Fund Remix)

A Quant was training a reinforcement learning agent to trade equities. The Master Trader came to him as he sat before six GPUs.

“What are you doing?” asked the Master Trader.

“I am training an agent to maximize Sharpe ratio. It has already learned to avoid drawdowns, size positions, and hedge tail risk.”

“Interesting. What happens if you reset the weights to random?”

“It would have to learn everything again from scratch.”

“And how would you know when to reset the weights?”

“I wouldn’t reset them. The agent has already converged.”

“But the market has not,” said the Master Trader.

The Quant was enlightened.

The Meta-Koan

Here’s what all ten koans have in common: they’re about the gap between the map and the territory.

The backtest is a map. The live market is the territory. The VaR number is a map. The tail event is the territory. The risk tolerance questionnaire is a map. The decades of scar tissue are the territory. The code is a map. The strategy-plus-competitors is the territory. The order book is a map. The intentions behind the orders are the territory. The converged model is a map. The non-stationary market is the territory. Averaging down is a map. Whether the thesis survived the drawdown is the territory.

And this is exactly what computational thinking teaches you.

When you learn to program — really learn, not just memorize syntax — you develop an instinct for the ways that formal systems diverge from the reality they’re supposed to model. You learn that code compiles but doesn’t work. That tests pass but the software fails. That the specification is not the system. You learn, at a visceral level, that running it is different from reasoning about it.

That instinct is worth more in trading than any amount of mathematics. Not because mathematics is wrong — the math is essential. But because mathematics tells you what should happen in the model, and computational thinking tells you all the ways the model is not the market.

The market is a computation that cannot be shortcut. You cannot derive what it will do from first principles. You cannot backtest your way to certainty. You cannot model your way past regime changes. The only way to know what happens is to run it — with real money, in real time, against real adversaries.

If that sounds like computational irreducibility, it’s because it is.

The Master Trader is not smarter than the Quant. The Master Trader has merely been enlightened more expensively.

For more on computational irreducibility and why even deterministic systems resist prediction, see my book The Science of Free Will. For the original AI Koans that inspired this post, see the Jargon File. For the rigorous version of why your drawdown rules are probably hurting you, see my paper “The Stop-Loss That Stops Gains.” And for a longer conversation about these ideas, see my two-part interview with Simon M. of Algo Advantage: Part 1 and Part 2.