Put the Petrov back in

This one was dictated into a reply box. August 2026; a senator I admire posted about corporate greed, a corrupt political system, and the unrestricted growth of AI data centers, and I found myself agreeing with every word and unable to use any of it. In this house that’s not trivia, it’s a citation. The arguments here carry where they were made, and this one was made losing an argument I was on the right side of.

So here it is said plain, and it starts with a concession: right and wrong. Right that the harm is real. Wrong about where to aim.

Greed is a motive, and a motive can’t be legislated

Corporate greed is the most popular sentence in American politics and one of the weakest, because it is a claim about an appetite. Here is the test that settles it, and it takes one move: replace every executive in the country with a saint tomorrow morning. Does the incentive change?

It does not. The quarterly obligation is unchanged. The fiduciary duty is unchanged. The board meets on the same day and asks the same question, and the saint either answers it the same way or is replaced by someone who will. If swapping the people fixed it, the motive frame would win and this page would be wrong — which is exactly why the test is worth running out loud. An argument that can’t lose isn’t doing any work.

And notice what the greed frame quietly implies: that the fix is better people. That’s a fix nobody has ever shipped. It also hands the other side an easy honest answer — I’m not greedy, I employ four hundred people — and they’re often telling the truth, and now the conversation is about their heart instead of the structure. Unfalsifiable claims are doors that only swing one way. This house doesn’t keep them.

The right actor, the wrong footing

And here is the part that matters most, because the phrase is closer to right than its critics think. Corporate greed names the right actor and stands it on the wrong footing.

Right actor. It is the corporation. Not “elites,” not “the system,” not some fog of bad vibes over the economy; the thing that made the decision is a specific, chartered, filed entity with a name and an address. Naming the actor correctly is most of the work, and that sentence does it.

Wrong footing. Greed is a moral property, and a moral property needs a moral subject. You can only be greedy if you can want something. A corporation cannot want. It is a filing. A charter, a set of obligations, a structure that persists after everyone who signed it is dead. Accusing it of an appetite is like accusing a filing cabinet of ambition — and the mistake isn’t a small stylistic one, because it is the exact same mistake the other side makes.

Sit with that, because it is the whole argument. “Corporations are greedy” and “corporations have speech” are one category error pointed in opposite directions. Both personify the entity; they only disagree about whether the person is nice. And the moment you grant it a character, you have granted it an interior — and a thing with an interior has beliefs to hold, a conscience to follow, a voice to protect. You paid the personhood tab to score one rhetorical point. That was the most expensive sentence in American politics and the left has been buying rounds with it for fifty years.

You cannot accuse a filing cabinet of wanting anything.

The house’s own rule already covers it, and it extends one word further than it usually gets to. If the entity cannot suffer, the entity cannot speak — and it cannot sin either. Suffering and vice run on the same wiring; both require somebody home. The corporation has neither, which is not an insult to it and is not a defence of it. It is the reason the thing needs a constraint rather than a conscience.

Which is why the footing decides the remedy, and this is where the frame does its real damage. A moral diagnosis licenses only moral medicine: shame it, boycott it, ask it to do better, wait for a generation of nicer executives. Every one of those is an appeal to an interior that is not there. Greed does not merely misdescribe the actor; it prescribes the one treatment guaranteed not to work on it. A structural diagnosis licenses structural medicine — the charter, the liability, the lever, the vote somebody actually holds — and those work on a filing, because a filing is exactly the kind of thing that can be rewritten.

You do not accuse a thermostat of cruelty. You change the number.

The injury is exemption

The structural version costs nothing and survives everyone: the entity deciding is not standing where the decision lands. That is the whole complaint, and it has the great advantage of being visible rather than inferred. If the entity cannot suffer, the entity cannot be answered; you can vote out a senator, and you cannot vote out a data center. The town carrying the load is the only party at that table with no lever, and no amount of goodwill in anyone’s heart changes the seating.

So the size of the thing was never the injury. Aim a law at unrestricted growth and you have started an argument about whether the number is right, which is an argument won by whoever can afford the most lobbyists to argue it. Aim it at exemption and there is no number to haggle over.

26 September 1983

A Soviet early-warning system reported American missiles inbound. It reported this with total confidence, which is what such systems do. The officer on duty, Stanislav Petrov, judged it a malfunction and reported a false alarm. It was sunlight on high-altitude clouds.

Sit with the shape of that, because it is this whole museum in one night: the machine was confident and wrong, and the only thing between that confidence and the end of everything was one human being who could be wrong and knew it. A system that requires a man to overrule it correctly on the worst night of his life is not a solved system; it is a coin that keeps landing on its edge. He did not accept the word hero. He was right about that too, and it doesn’t matter, because the job he did that night is the job this argument is about.

Now look at what is being built. Every one of those buildings is a place where that man’s job is being removed on purpose, and the removal is the product. Not a side effect, not an accident of scale — the pitch. Faster than review. No hesitation. No one to slow it down.

Kubrick got there in 1964

The object at the center of Dr. Strangelove is a doomsday machine, and the film is careful about why it is automated: so that no human can call it off. That is stated as its entire virtue. Remove the person who might hesitate and the deterrent becomes credible. Then it triggers by accident and nobody can stop it, precisely because they built it so that nobody could.

Two directions, one point. Strangelove’s nightmare was a system with no Petrov in it — and the safety was never in the hardware. It was in the one part everybody keeps proposing to remove.

What the government should actually be aimed at

Three asks, and every one of them is bill-shaped — a thing with a number, a text, a vote and a date, rather than a mood:

1. Consent at the site. The community carrying the load gets a vote on the load. Not a comment period after the permit; a lever before it. The people who will live beside a thing are the ones currently holding nothing.
2. The full cost billed back. Whatever it draws — power, water, grid capacity, road wear — the entity carries the whole cost instead of the difference landing quietly on the people already there. Not a cap. A bill. A thing that pays its own way can grow all it likes.
3. A reachable human name on every decision. Every approval, and every consequential output, carries a person who signed it and can still be found when the outcome arrives. Not to punish anybody — to end the condition where a decision exists that no one made.

Do those three and the cap is unnecessary, because the growth either pays for itself with the neighbours’ consent or it stops on its own, and nobody had to guess the right number.

The steelman, at full strength

Automation is frequently safer than the human it replaced. Autopilot, anaesthesia monitors, anti-lock brakes; the record is not close, and the honest version of this page has to say so. Humans are noisy scorers — tired, biased, inconsistent — and a rubric is frozen prior human judgment precisely because a fresh judgment at 4am is worse. And “a name on every decision” can decay into theatre: a signature block, a compliance officer, a person whose entire job is to be blamed.

All granted. Here is the distinction the objection misses, and it is the whole hinge: steady and exempt are two different properties, and only the second one is the problem. You want the person deciding to be calm. You do not want them safe from the answer. The ask was never no automation; it is no unreviewable automation — a reachable human, not a slow one. Petrov was not detached and he was not panicking. He was exposed and steady at once, which is the only combination that has ever produced a good decision about anything that matters, and the one combination that cannot be built. A model cannot be exposed. A corporation cannot be exposed. A room cannot be exposed. Only a person can stand in the blast radius of their own judgment.

Which is, finally, the whole house: machines generate; only humans score. (His coin, dictated 2026-08-07 into a reply to a public post about data centers; the machine held the pen close to his words — refine to his. Values page, labeled opinion. The Petrov account and the Strangelove mechanism are given from the public record as commonly told; the argument does not rest on either detail being exact, and no figures for power, water or cost appear here on purpose — a claim about utility economics deserves to be answered on the argument rather than on a remembered number.)

If the sentence ever feels like it wants to end — in the US, call or text 988 (Suicide & Crisis Lifeline), any hour; or find help anywhere at findahelpline.com. You are not alone, and the story isn’t over.

⌂ The full Ethos — all the arguments on one page