The Long Game for All: Counting What Frontlines Know

For four years at Lansing Grand River we were taught the Toyota Production System, and we lived it. It worked. The plant took J.D. Power’s gold, and Cadillac came back.

I thought we would be its prophets. I thought General Motors would send us out across the company to teach what we had learned.

Others went out. I was told to stick to my knitting in my own trim shop. My boss said, you’re good at this, we need you here. I was told I was still too soft on the floor.

And then I left. What I knew left with me. Nobody at General Motors wrote it down, and for a long time neither did I.

I am writing it down now, with a machine, and I can hear what someone will say. All Venki has is a Claude Max subscription.

He would be right about the subscription. Here is the rest.

On a Sunday morning in October 2026, a chart came down my LinkedIn feed. A brokerage had ranked countries by what they pay factory workers: Germany at $33,000, India at $2,500. I thought: slop. It turned rupees into dollars at the bank’s rate, as if a painter in Chennai bought his rice in Ohio.

So I asked the tool to redo the chart at the prices workers pay. It found the surveys and the filings, built a table of six wages in three currencies, and wrote a draft. The draft had nobody in it. I asked why anyone would read the thing. It said the draft would be dismissed, and fairly.

So I told it about Selvam.

The tool found every number in that essay. It could not find Selvam. What I publish is testimony, not reporting, and I should be held to that. The machine can lower the cost of the sentences. It cannot supply the witness, and it cannot make it safe for the witness to speak. That’s the argument.

What the tool did

It did the work of a research desk, and it did it in a day.

The lens was mine. I have run plants in both countries, and I knew a wage is not what the exchange rate says. My work on Bloom said something more. A firm sells in one currency and its people live in another, and the job is to thread the needle between the two. The tool put the two currencies into one clean sentence before I did, and I recognised it as the thing I had been circling. I asked for every figure in rupees, euros and dollars so that a reader anywhere could find himself in the table. The arithmetic was the tool’s. I could not have done it in a month.

It also told me no. I wanted the essay to say that paying for recognised skill funds a firm’s growth. The numbers showed something smaller: the firms that paid did not fall behind. The essay says the smaller thing.

I told it no as well. It suggested I blur the details of Selvam’s life. I kept them as they were. It offered an ending. I wrote my own.

It checked its figures against the original documents only at the end, after I asked why it hadn’t. The ones it checked held. That it needed asking is the part I would remember.

And it wrote most of the sentences. It wrote most of these. In September 2026 Jason Gay wrote in the Wall Street Journal about Brandon McNulty winning the world road race, the first American man to do it in thirty-three years. It read like poetry to me. Nobody has ever confused my sentences with his, and the machine has not changed that.

What it could not do

It could not find what we paid Selvam. It searched, and came back with a range from the phone factories on the other side of Chennai.

It had no bench. Selvam hand-sprayed our fuel tanks at Tiruvottiyur in 2010. I went out and found him on the bench by the door, getting his breath. His hands and feet were coated in paint. He told me what he knew. The tanks came to him coated in oil to keep them from rusting, and the cleaning stage on the line worked some days and not others. Then he told me about the booth. It was clogged with excess paint. The air moved so poorly that on some shifts he walked out every ten minutes to breathe.

Then he looked back at me and asked, “And you’re worried about paint blisters?”

Every painting robot follows a path that some painter’s wrist knew first. Robots later painted those tanks, and they never knew why the paint had peeled. Selvam knew. He told me in two minutes.

He was a temporary worker. A temporary worker has every reason not to volunteer what is wrong with the plant. He told me because I had walked out to the bench.

That knowledge was in one man, and nobody had written him down. A tool built from everything ever published has nothing on Selvam, because nobody ever published him. Nor could it have known that lunch in the cafeteria was the bright moment of his day, and of mine.

You can watch what the witness was worth. The tool scored its first draft against my own rubric, a hundred points with ninety to publish, and gave it about eighty. Rebuilt around Selvam, the draft got ninety-three from the tool that wrote it, and a cold reading later put it at ninety.

The reporter with the same tool

Nick Lichtenberg is a business editor at Fortune who once wrote about my work. In the spring of 2026 the Wall Street Journal reported that he had produced more stories in six months than any of his colleagues managed in a year, with AI doing much of the drafting. The reaction was loud. One website’s headline called editors like him sloplords.

So Nick and I have the same tool and the same accusation. We do not have the same job. By the Journal’s count he had filed more than six hundred stories. I spent a long day arguing with the tool over one. He told an interviewer that the tool had pushed him toward more original reporting, because it can only look backward. That is my finding from the other side of the desk.

A reporter would have done things I did not do. He would have looked for Selvam. He would have called the plant, and the company in Coimbatore, and asked each for its side. He would have doubted my memory, which nobody did. He would have had an editor, whose job is to disbelieve him. I had a tool that scored its own draft. My essay has one witness, and the witness has something to sell.

That is testimony. It is worth having. It is not reporting, and good arithmetic should not dress it up as reporting.

The rules already draw that line. The Wall Street Journal’s newsroom lets its reporters create no text with the tool, while its opinion pages let an outside contributor use it if the position is truly his. The Morning Context, where I also publish, takes nothing written with it, so what I send there I write by hand. I live under two rules myself. A newspaper vouches for every sentence its reporters file. A contributor is asked for what he knows and where he stands.

Bloom has a rule for machines on a factory floor: augment the person, don’t replace him. The tool did that for me. It took the part of the work that was never the point and left me the bench. Whether it does the same for a newsroom is not settled. The gain from a tool is either banked as fewer people or spent on more of what only people can do.

Why the witness stays quiet

People who know a floor rarely write about it. They have no time. Nobody pays them for it. And it carries risk, because an employer seldom thanks you for describing what you saw. The tool helps with the first of those. It does nothing about the other two.

What happened to me at Lansing Grand River was Selvam’s situation in a better office. I was not temporary, and the message was still the same. Know your work, and stay in your shop. The boss who told me so had come in on a high-potential rotation, a short assignment. The ones over her did not seem to care. It happened in other jobs too. Each time I got closer to why margins, or growth, or relations with the union were stuck, I found that knowing it and being free to say it were different things.

The slop charge says the machine made writing cheap. The typing was never the expensive part of testimony. Thirty-six years on the floor was, and so is saying what you saw.

You can see the risk in my own pages. The pipefitter who kept our glass cells fed appears under a changed name. So does the painter. The people who know the most can least afford to say it, so testimony arrives late, from those who have left. Selvam’s bench is from 2010. It reached print sixteen years later.

What the books never held

Now look at the ledger. The subscription is on my card statement every month. Anyone can buy the same one and get the same research desk.

What General Motors spent teaching us at Lansing Grand River was expensed in the year it was spent. What we learned was on nobody’s books. So when I left, the company lost nothing it could count. It is why silencing a witness is so easily done. It costs the firm nothing it can see.

That is what I say about every plant I write about. The machine is the expense everyone can see. The capability is the asset nobody booked. I did not expect to be my own example.

What it would take

I think slop is writing that nobody answers for. Nobody checked it, nobody would be embarrassed if it were wrong, and nothing in the making of it could tell the author no. So before I draft anything now, I ask three questions. What is in this that only I could write? What would an informed reader not already know? And what did the evidence make me give up? If the answers are nothing, nothing and nothing, the subscription wrote it.

Put the brokerage’s chart to those questions. There was nothing in it that only its maker could know. Its figure for America was about half the true one. It gave up nothing. That was slop, whatever made it.

So my numbers should be checked. My memory should be doubted. Someone who reports for a living could go and find the Selvams I can’t reach, and tell me what I got wrong.

Here is what I am trying to solve. Business has to change for the good of everyone it touches, and not only its investors. Practitioners have to make that case, because they are the ones who know. But a practitioner is muzzled by his paycheck. I do not know how, in God’s name, his voice and his knowledge can be brought forth.

The tool is not the answer. It made my sentences cheaper. It did not make anyone safer.

If you have stood on a floor, you have a Selvam. What would it take for you to write him down?

The Long Game.

The painter’s and the pipefitter’s names are changed. Nothing else is.

Venki Padmanabhan is a co-founder of the Capability Capital Institute and the author of the forthcoming Built to Extract and Already Paid For (Capability Capital Press). He writes at thelonggameforall.substack.com.

Written with AI assistance. The argument, the judgments, and the floor testimony are the author’s own.

Capability Capital Institute — the case for counting what working people know: capabilitycapital.org

Years ago I ran improvement work for a large automaker. I taught lean — the Toyota tools, waste and flow, the whole grammar of doing more with less. I believed in it. I still do.

One afternoon I got pulled into a room. The company was looking hard at productivity — hours per vehicle, the number every plant lives and dies by. Someone asked me a plain question. These lean methods you’re teaching — can we use them to get the reductions we need?

I felt the color go out of my face. I stuttered. And then I told the truth, because there wasn’t another answer in the room. I said: if there is waste in the system — if the jobs are designed and run with motion that adds nothing — then yes. Take the waste out, and you can do the same work with fewer people.

It was the most honest sentence I’d said all year, and the most expensive.

I spent the months after that going plant to plant, floor to floor, sitting with the people who did the work, learning from them where the waste actually was. We found it. We took it out. The one mercy was that the people we reduced didn’t hit the street — they went to a jobs bank. Some of those banks emptied out anyway in the years that followed. I knew that then. I know it better now.

I have spent a long time trying to name what I was asked to do in that room. What follows is the best I have. It is not finished, and I’ll tell you exactly where it runs out. But I can give you the shape in one line: I’ve come to believe there are only two operations that ever reach the worker — and they run in opposite directions. Everything else is surface. That’s the argument.

Here is how I get there.

There are, in the end, only a handful of things you can do to a business. Each is an operation, like arithmetic.

You can add. Pour leads into the top of the funnel. A whole industry sells this now, and its meter measures volume in.

You can subtract. Strip cost and waste. This is the trade I came up in — lean, constraints, Six Sigma, the operational-excellence field. The slogan is more with less, and the less is the tell.

You can divide. Split what already exists. This is the bargain — and here I have to say something both sides dislike: collective bargaining and management are not opponents at the level of operation. They are co-authors of the same one. The wage fight is a division. The strike is a division. The contract is a signed division. Both parties walk in having already agreed the pie is fixed, then fight over the slices. I have stood in the brutal American version and in the civilized German one. The German version is better. It is still division — even the best split only governs who gets which piece.

Three operations. Three industries built to sell them. And here is the thing it took me thirty years and that room to see.

None of them touch the worker. Add fills a funnel. Subtract trims a cost line. Divide carves a pie. The person doing the work ends exactly where they started — same skill, same standing, one quarter older.

The operations everyone sells you act on quantities. Not on people.

But there are two more operations. And they are the only two that reach the person — and they run in opposite directions.

The first is the one I performed in that room. I called it subtraction. That’s not quite right. Subtracting waste is honest work — motion that adds nothing should go. What I did went one notch further. The waste came out, and so did people whose capability the company had already paid years to build and would never get back. That is not subtracting waste. That is subtracting value — same motion, aimed one notch over, off the waste and onto the worker. It has a truer name. Extraction.

And it is not a villain’s act — that’s the part I most want an owner to hear. The machine is built to extract: the default architecture every one of us inherited, which treats a trained worker as a cost to recover rather than a capability to grow. You don’t have to be cruel to run it. You only have to answer a productivity question honestly on an ordinary afternoon, the way I did.

Years later I watched the same operation from the other side of the table. At Lansing Delta Township, one of the trim shop team leaders — I’ll call him John; that isn’t his name — looked like a rebel, dreadlocks and all. He was a spectacular team leader. He knew every job on his line, took care of his people, and the quality off his team was superb. He found a simple, creative way to improve one part and how it went onto the car, did the due diligence himself, and turned it in as a suggestion.

For the two years I was on that floor, at least once a month, we sat down with the higher-ups about the answers he was getting. He honestly believed the company had already made his change and given him no credit for it. I escalated it several times and never got it resolved. It stood between us like a bone: me as management, unable to give him the credit he wanted.

I used to think the opposite of extraction was a fair split — that if the worker got a bigger slice, the wound would close. I was wrong. A fairer division is a better floor, not the opposite operation. You can divide the pie perfectly and still never touch whether the work forms or deforms the person doing it.

The opposite of extraction is compounding. Form the worker, and their capability compounds — the setup man trains two more; the tacit knowledge gets written down before the man who held it retires; this year’s skill is the base next year multiplies. It is the only operation that grows the underlying stock instead of moving it around, and the only one with a clock. Add, subtract, and divide all settle this quarter. Compounding needs years to show its power — which is exactly why no one sells it, and exactly why it is the long game.

So the five collapse into two. Everything sold to manufacturers — more leads, less cost, a fairer split — settles this quarter and leaves the worker where they stood. Only extraction and compounding act on the person across time. One depletes the human. One compounds them. Every business is running one of them, whether the owner chose it or not.

And this is not a blue-collar story. The same operation runs in the office, faster there every month. The analyst whose judgment the firm spent a decade building, then hands to a model. The engineer whose feel for the work becomes a training set. Extraction does not care about your collar — only whether your capability can be banked and the rest of you let go.

So why does it feel like progress? Not because anyone doubts the man — everyone on that floor knows the setup man’s thirty years of judgment are worth more than the press he stands at. Seeing the value was never the problem. Measuring it is. The press is a half-million-dollar asset, depreciating on a schedule the accountant can recite. The judgment beside it is just as real and nowhere on the balance sheet, because we never built the instrument to weigh it — a habit that runs back to Frederick Taylor, who taught us to measure what is easy and ignore the rest. So the machine became an asset and the capability an expense. On the income statement, subtracting the worker and subtracting the waste look identical. They are not. One removes motion that was never worth anything. The other removes the only asset you own that compounds — and records the loss as a saving.

That is the whole trick — an accounting trick built on a measurement no one ever undertook, because it was hard and Taylor told us not to bother. Here is the work I think matters most, and I don’t yet know it can be done: give capability an account that could survive an audit. Not a slogan — a real measure. Build that, let it accrue, and the arithmetic changes. You stop reading a formed worker as a cost to recover and start reading them as the one line on the page that grows.

Now here is where I run out, and I’d rather show you the edge than pretend there isn’t one.

The people we sent to the jobs bank were spared the street. They were not spared the message — you are surplus. And they had already paid in: years of their lives crystallized into that company’s capability, banked, then found redundant anyway. I do not know — and I want to be honest that I don’t — whether compounding prevents that message or only delays it a cycle, until the next downturn finds those people anyway. I have one shop where I’m testing it: a small manufacturer whose owner let me run an audit on his people instead of his costs. It is early. What I’m watching for is whether the operator who ran one line last year is training new hires this year, whether the knowledge in one man’s hands gets written down before those hands retire. One floor. Not proof — a single data point and a man’s conviction, which is exactly what you should be suspicious of.

So I’m not handing you a system. I’m handing you a direction and a magnitude: compound the worker, at the base, on a horizon of years. I think it’s the right way to point. I don’t yet know how far it goes, or what’s between here and there, and I am looking for people who’ll build the rest of it with me — on real floors, where it will either hold or it won’t.

That afternoon, I gave the only honest answer to the wrong question. The right question was never how to get the hours per vehicle down with fewer people. It was how to make the people worth more than the hours. I didn’t have that question then. I’m not certain I have the whole of it now.

But I know which direction it’s in. And I know it doesn’t run through that room.

Subtract the waste. Never the worker. The difference between the two is a single question, asked on an ordinary afternoon — and the rest of a working life spent making sure you answer it better than I did.

That is the long game. Not the quarter you cut. The decades you compound.

Every business is running one of the two. Which one is yours running? Tell me.

Venki Padmanabhan is a co-founder of the Capability Capital Institute and the author of the forthcoming Built to Extract and Already Paid For (Capability Capital Press). He writes at thelonggameforall.substack.com.

Written with AI assistance. The argument, the judgments, and the floor testimony are the author’s own.

I learned the Mahabharata from my grandmother, in the years I lived in her house in Chennai for school. She told it from the verandah ledge of Ganesh Sadan on Venkataraman Street, eyes closed, her thumbs circling one another in a slow chase that never caught.

She had favorites, and they were never the winners.

She always came back to Karna. The greatest archer of the age, son of a god, set adrift as an infant and raised by a charioteer — and therefore, his whole life, refused. Refused at the tournament, when Kripa asked for his lineage in front of everyone. Refused by Drona, who would not teach a suta. He learned in the end by lying about his birth, was cursed for the lie, and the curse came due on the last day: his chariot wheel sank in the mud, and he stepped down unarmed to lift it, and was killed while he lifted.

And the part she told with the most feeling, every time, was that they all knew. Kunti knew — she had borne him and set him on the river. Krishna knew. Neither said a word until the eve of the war, when they came to him separately and offered what had been his by birth the whole time: his name, his brothers, an empire. They named him at the precise moment naming him became useful to them. He refused it, and went out to die under the name they had let him carry instead.

Then Ekalavya came to Drona. A tribal boy who wanted to be taught, turned away for his birth by a man who had himself been turned away for his poverty. So the boy built a clay image of Drona in the forest and taught himself in front of it, and became better than Arjuna. When Drona found out, he demanded the traditional teacher’s fee for instruction he had never given: the boy’s right thumb. Ekalavya cut it off and handed it over.

And she would tell me how the epic itself was made — Vyasa composing aloud, Ganesha writing it down on the condition that Vyasa never pause, and on the counter-condition that Ganesha understand every verse before he set it down. When the pen failed, Ganesha broke off his own tusk and kept writing.

I have been living inside that one great epic my whole life. Watching Christopher Nolan’s Odyssey with my son Dakshin, I finally met Homer’s.

David French wrote in the Times that Nolan has rebuked the spirit of the age — that his Odysseus carries a moral injury, and that the film insists moral transgressions always carry a consequence. That reading is right. But one line put me somewhere French didn’t go, and my grandmother had put me there fifty years earlier. Odysseus says it of the gods, who walk among men disguised as strangers because that is the only way to learn what a man is actually worth: the clearest view of men is from below.

Not Odysseus, then. The men who rowed for him.

Here is the argument. The suppression of frontline judgment and the extraction of frontline capability are not an efficiency problem, not a culture problem, not an HR problem. They are two things at once, and I have never been willing to say both in one breath. They are the violation of a moral code — and the code says what codes always say: the transgression carries a consequence, the consequence arrives late, and it lands on people who were never in the room. And they are an act of self-blinding, because the clearest view of any operation is from below, and we built an entire profession on the opposite assumption. The moral claim and the operating claim are the same claim. That is what I could not see for most of a career.

That’s the argument.

Ekalavya is the one that has never let go of me, because Ekalavya is not a war story. It is an operating decision.

Look at what happens. A system declines to invest in forming a person. The person forms himself anyway, at his own cost, out of clay and repetition and no instruction whatsoever. He becomes the most capable practitioner in the country. And at the moment his capability becomes visible, the institution that refused to form him arrives to claim a fee — and takes it in the one form that guarantees he can never use what he built. Not the arrows. Not the bow. The thumb.

That is not a fable about cruelty. It is a precise description of extraction: capability the system refused to develop, harvested at maturity, in a manner that protects the credentialed incumbent standing behind the teacher. Arjuna complains, and the thumb comes off. It happens every time a company declines to train and then poaches, declines to promote and then consults, declines to form and then buys the formed.

And do not miss who takes it. Drona had himself been refused once, humiliated for his poverty by a childhood friend who told him equals only befriend equals, and then spent the rest of his life administering the hierarchy that broke him. That is not the profile of a villain. It is the profile of a manager.

So let me say the plain thing once, because I have spent thirty-six years in this business and do not recall hearing anyone say it. What we do to the people on our frontlines is not inefficient. It is not suboptimal. It is not a missed opportunity for engagement. It is wrong. Suppressing a person’s judgment to protect your own authority is wrong. Taking capability you declined to pay to build is wrong. We have a hundred words for the cost of it and no word for the wrong of it, and that absence is not an oversight. It is the design.

And the ledger is where the thumb goes.

Look at what a public company must name. Named executive officers, to the dollar, equity broken out by grant date. Goodwill, by acquisition. Intangibles, by class, with amortization schedules. Each gets a name, a line, a footnote.

Now look at the people. Under ASC 805, when one company buys another, the assembled workforce — the actual trained, formed, functioning capability of the people who do the work — is not recognized as an identifiable asset. It is subsumed into goodwill. It vanishes into a residual. The most valuable thing you just bought is the one thing the standard forbids you to name.

What a ledger names, it protects. What it totals, it forgets.

Homer does the same. He does not look at the crew. Odysseus’s men are eaten, drowned, turned to swine, killed by their own hunger on Thrinacia, and they are barely named. Twenty years they followed him, and they get no epic. The poem is titled after the man who came home.

Vyasa looks. Eighteen akshauhinis march to Kurukshetra and the text gives the arithmetic down to the last man — and then, when it is over, it stops. Book Eleven is the Stri Parva, the Book of the Women. The widows walk onto the field. They find the bodies. They are made to speak, the king is made to listen, and the reader is not permitted to leave until they have finished. That is the difference between an epic that counts and an epic that names, and it is exactly the difference between two ways of keeping books. We built ours on Homer’s model.

In 2000 I was a shift leader in the trim shop at Lansing Grand River, where a painted body arrives naked and gets its interior before it goes to chassis for an engine and wheels. In 2022 I took the same job again, at the larger and thriving Lansing Delta Township, after twenty years in Germany and India spent mostly as an MD or a CEO running two-hundred-million-dollar companies. I wanted to see what two decades of lean had actually done to the work in my own town. I wanted to stand on my favorite frontline again.

Nobody on that line needed to know any of that, and it would not have helped them if they had. What I could see from the floor had been invisible from every seat I ever held above it — not more detail, which is what people mean when they say get closer to the work, but a different order of fact. From above you learn what the plant produced. From below you learn what it cost, who paid it, and what got burned to make the number. Bloom came out of those years. Both books came out of those years. The institute came out of those years. None of it came out of the years when I held the most authority.

Jamie was my team leader on the Trim 2 line. Before the plant, he was a Marine sharpshooter.

We ran fifty-five jobs an hour — a car a minute, near enough. When a defect surfaced a hundred cars down the line, Jamie was already moving before most people had finished reading the alert. He knew what the fault was. He knew which tools to carry. He walked it back through the noise with the line still running and the cars still coming, and he was almost never wrong. The nervous system that reads wind and range at a mile reads a process drifting out of tolerance through a wall of noise. It is not a related skill. It is the same skill, carried off a ridgeline onto a trim line.

He earned about seventy-five thousand dollars a year, standing watch over a fifteen-billion-dollar plant — keeping a drift from becoming a defect, a defect from becoming a recall, a recall from becoming a morning on the news.

He carried that plant on his back at night. Silently, and like a loaded weapon.

One night Jamie told me how the Corps had made him. Don’t tell anyone, he said, but I’m a Marine. Then how you learn to read the wind and put a round on a target a mile out — practice, practice, practice, and tons of discipline. That was the whole curriculum in his telling: a repetition and a discipline. It is also a more complete account of formation than anything my plant had to offer anyone in it.

I told people. Of course I did — with visitors, with my staff, in rooms where he was standing right there, often enough that I am fairly sure I embarrassed him. He never said so. He wouldn’t have.

The Marine Corps formed him deliberately, over years, graded him by name, and wrote it down. He walked onto my floor carrying an exact record of what he had been made into, and there was no field in any system I had to receive it. So I carried it around as an anecdote instead. An anecdote is what you get where there is no account, and it costs the man something an account would not: it makes him a story rather than an asset, and a story belongs to whoever is telling it.

He asked me not to tell anyone, and I have told everyone, and I am telling you now.

My grandmother’s favorite was a charioteer’s son. So is the only man in the epic who can see.

Sanjaya is a suta — Karna’s class, the class the poem spends eighteen books refusing. Not a warrior, not a counsellor. He is given divine sight so he can see the whole battlefield, every movement and every death, and he sits beside the blind king Dhritarashtra and reports it. All eighteen days. Everything that happens, as it happens, in full detail, to the one man with the authority to stop it.

Dhritarashtra hears all of it and does nothing.

That is the most accurate description of an operating review I have ever encountered. The intelligence is not missing. It is never missing. Sanjaya is standing on the floor of every plant in this country, and the report is accurate, and the king is blind by choice and asks about the schedule.

Notice what the poem says about sight. The throne is the highest seat in the kingdom and it sees nothing. The charioteer’s son sees everything. That is not a moral bolted onto an operating fact — it is the operating fact, and it runs against the whole vocabulary we built the profession on: dashboards, line of sight, visibility, seeing the system. Every one of those words assumes the clearest view is from above. From above you get aggregates. You get the total, never the names.

So when a system suppresses the judgment of the people nearest the work, it is not only committing the transgression. It is putting out its own eyes. The wound and the blindness are one wound.

And this is no longer a factory argument. The paralegal who knows which precedent the model invented. The tech who knows the machine is drifting. We are automating the exact layer where judgment gets formed, calling it productivity, and booking the savings this quarter. The savings are real. So was the Trojan Horse.

So: what of the soldiers?

The Pandavas win. Then the Mausala Parva comes and the clan destroys itself in a drunken brawl on a beach, and the brothers walk into the Himalayas and fall one by one on the road. Odysseus reaches Ithaca with no ship and no man alive, and he is home and he is not restored. That is the consequence both epics promise, and it is not a satisfying one. It gives no one back their life. It does not grow the thumb back.

But there is one thing Vyasa does that Homer does not, and it is the only instruction I can find in either poem. He makes the king stand in the field and listen to the women count their dead. Not as penance. As accounting. The whole of Book Eleven exists so that the number becomes names.

We can do that, and it is not mystical. It is a standard, a ledger line, a verification authority, a capability account owned by the worker and not the firm. It is the Stri Parva written as accounting policy. I have a version of how that should work, I have been wrong about parts of it in public before, and I will be again.

What I am no longer willing to do is call it an efficiency problem.

And I keep returning to how my grandmother said the thing was made. Not by one man. Vyasa spoke it and Ganesha wrote it, and Ganesha’s condition was not speed — it was that he would not set down a line he did not understand. The epic exists because the scribe refused to be a stenographer. She recited it with her eyes closed, from the ledge of a house named for that scribe. For years I thought it was only the address. It was the instruction.

The clearest view of men is from below. It always was. The question is who is standing down there, and whether anyone above them has the nerve to ask what they can see.

Tell me what you would put in Book Eleven. I will not set it down until I understand it. That’s the long game.

Venki Padmanabhan is a co-founder of the Capability Capital Institute and the author of the forthcoming Built to Extract and Already Paid For (Capability Capital Press). He writes at thelonggameforall.substack.com.

Written with AI assistance. The argument, the judgments, and the floor testimony are the author’s own.

In 2010 the complaints came in a rash. Paint was peeling off the fuel tanks of motorcycles we had built at Royal Enfield’s plant in Tiruvottiyur, in the north of Chennai. It was my line. I pulled the team into a huddle beside it to find the cause.

One man was missing.

Selvam hand-sprayed our frames, tanks and nacelles. Somebody said he was outside. I went out and found him on the bench by the door, getting his breath. His hands and feet were coated in paint.

He told me what he knew. The tanks came to him coated in oil to keep them from rusting, and the cleaning stage on the line worked some days and not others. Then he told me about the booth. It was clogged with excess paint. The air moved so poorly that on some shifts he walked out every ten minutes to breathe.

Then he looked back at me and asked, “And you’re worried about paint blisters?”

Here is what I knew about Selvam. He walked to work, thirty minutes from a house deep inside Tiruvottiyur. He lived with his brother’s family. He wanted to marry and rent a room of his own. He was trying to save, and he could not.

He was a temporary worker.

I thought of him in October 2026, when a chart came down my LinkedIn feed. A brokerage had ranked countries by what they pay factory workers. Germany was first at $33,000. India was thirteenth at $2,500.

The chart asks what a worker costs. Selvam was asking what his wage could buy: a room, a wedding, air.

A wage is two numbers. One is a cost, read at the exchange rate by the firm that pays it. The other is a life, bought at local prices by the man who earns it. I managed the first number. Selvam lived in the second. The firms that manage both, with secure years and a wage that rises on proven skill, buy their people a far better life and do not fall behind for it. That’s the argument.

Two rates for one wage

The exchange rate tells you what a rupee fetches at a bank. As I write, a dollar is ₹96.3, or €0.89. That is the firm’s rate, because a firm counts its wage bill in the money it sells in.

Purchasing power tells you what a rupee fetches at a market stall. It takes about ₹19.9 in India, or €0.72 in Germany, to buy what a dollar buys in America. That is the worker’s rate. Selvam’s rice and the room he wanted were priced in Tiruvottiyur, not in Ohio.

Take India’s average factory wage today, ₹2.28 lakh a year. At the bank it is $2,400, the number on the chart. At the stall it buys what $11,500 buys in Ohio, or what €8,200 buys in Bavaria. It is a thin life. It is also nearly five times the one the chart drew.

So every wage below is shown three ways: rupees at Indian prices, euros at German prices, dollars at American prices. Find your own currency. That is the life.

Six wages

I looked for firms that already give secure years and a rising wage, one on each of three continents, and set each beside its local base.

The Ohio and Lincoln rows are from 2021. The Bavarian annual figures are my estimates from the monthly tariff.

Read down the rupee column. A permanent worker at ELGi, the compressor maker in Coimbatore, lives close to the life of the average Ohio factory worker, on four times his own country’s base. Lincoln’s production worker earns one and a half times Ohio’s. In Bavaria the skilled grade is 1.3 times the entry grade, and the step between them is public: three years as an apprentice, then an examination set by the chamber of commerce and not by the employer.

Now read across. The base runs from $11,500 to $57,900, a fivefold gap. The recognised worker runs from $46,100 to $73,700, a gap of 1.6. Countries set the bottom of each pair. Firms set the top.

It did not cost those firms their growth. In their latest full year Lincoln outgrew ESAB, its rival in welding, and ELGi outgrew Kirloskar Pneumatic, at margins level with their rivals’ or behind them. One year and two pairs prove little. They do answer the man who says the premium can’t be afforded.

Why Selvam was temporary

None of those rows was open to Selvam.

I don’t remember what we paid him. I remember what we fed him. There was free coffee on the floor, and breakfast and lunch in the cafeteria. I looked forward to it every day, along with Selvam and everyone else. It was the bright moment of the day. We fed them.

The record tells me roughly what he took home. In 2009-10 the average Indian factory worker earned ₹75,277 a year, about ₹6,300 a month. A study of the phone factories west of Chennai at that time put contract workers and trainees at ₹4,100 to ₹5,500 a month, beside permanent workers on ₹6,000 to ₹11,700. I would guess a temporary painter in the north of the city was near the bottom of that range. At that year’s prices, ₹5,000 a month bought about what $4,100 a year bought in America. We fed him, and he still could not save for a room.

Selvam stood on the short side of that gap, and there was a reason. Under Indian law a permanent job was very hard for an employer to end. So plants like mine kept the permanent roll small and filled the line with temporary men. It was rational, and it was common.

But the statute did not clog the booth. That was my plant. He was right, and I was worried about paint blisters.

In time the hand-spray booth gave way to an automated paint shop built by Dürr. The company went from about fifty thousand motorcycles a year to more than a million. Since 2010 the average Indian factory wage has tripled in rupees. Measured as a life it has hardly moved: one study found the real wages of directly employed factory workers grew 0.4 percent a year over two decades. I can’t tell you whether Selvam’s life got better. The averages say not to assume it.

The table we already keep

Firms know how to read a wage as a life. I know, because mine was read that way. From 2006 to 2009 I was an American expatriate, first in Stuttgart and then in Chennai. My employer paid for an international school for our three children in both cities, a company car in both, and part of the housing in both. There were R&R trips every year. Nobody read my pay at the exchange rate. For India there was a hardship allowance: I was paid extra to live in the city where Selvam lived with his brother’s family. We own the lens. We point it at the top of the org chart.

The ledger says the lens is cheap to turn. Across India’s registered factories, wages to workers are about 2.3 percent of output. On the firm’s books the wage is small change. In the worker’s kitchen it is everything.

The books make it worse in one more way. What a plant spends forming a painter is expensed in the year it is spent, so a temporary man is all cost and no asset. Nobody defends a cost.

A middle rung

Selvam had two possible lives at that plant: temporary for as long as it lasted, or permanent for life. We offered him the first.

In Bloom we call the rung between them Water. It is three secure years, with a wage that steps up each time someone qualified confirms the skill, and a table on the wall that says what each step pays. The worker gets a runway long enough to rent the room. The firm gets an end date, and three years to earn back what it spends forming him. India’s new labour codes now allow a fixed term on the same pay and benefits as a permanent job. That is the opening.

I have not found a firm that puts all of it together. ELGi shows the destination exists. Lincoln has guaranteed the job from a worker’s third anniversary since 1958. Bavaria shows the step can be public. Bloom has a first draft of the ladder, and it is a draft. I don’t know the right steps for a foundry in Coimbatore or a moulding shop in Ohio, and I would rather be corrected by the people who run them.

If you run a plant, work out one number this week: what your newest temporary worker’s wage buys at the prices in your own town. Could he rent the room?

The Long Game.

Then work out the other number. What percentage of your P&L is wages? What movement in that number can you correlate with growth? Try Bloom!

The painter’s name is changed. Nothing else is.

Venki Padmanabhan is a co-founder of the Capability Capital Institute and the author of the forthcoming Built to Extract and Already Paid For (Capability Capital Press). He writes at thelonggameforall.substack.com.

Written with AI assistance. The argument, the judgments, and the floor testimony are the author’s own.

Capability Capital Institute — the case for counting what working people know: capabilitycapital.org

On Tuesday, GM’s board approved $20 billion to buy back its own stock. Let me tell you about a man who helped make that money.

Jim, an operator on my line at Buick City in Flint, kept a good book beside him and read between jobs. I once gave him my statement of purpose in life. He gave it back with careful notes. I still have them. His job card asked for parts installed on time. It never asked for what he gave me in those margins.

Jim did not buy either story he was told. Management told one to investors. The union told another to the rank and file. He heard both for decades, and he kept asking the same question: who knows the real story on the floor?

Over thirty years, on a line building fifty-five cars an hour, Jim grew his judgment, his knowledge and his ideas many times over. On the day the plant closed, the last car went by, and Jim stayed on his stool, looking down the line as if another one was coming.

The buildings and the tooling had lines on the balance sheet, and somebody wrote them down. Jim had no line. There was nothing to write.

Jim was one man in a long line. In the five years after 2002, GM cut its U.S. workforce by about 40 percent. Then came the 2009 bankruptcy. By 2019, GM had roughly 49,000 UAW workers left in the United States. Every one of those exits was booked as a saving. Not one was booked as value going overboard.

That $20 billion is now a fight between GM’s CFO and the union president. Neither asked what made it. Jims made it, and no ledger counts them. Until a board can count what formed workers make, it will keep handing the cash back. That’s the argument.

Tuesday, two men

On Tuesday morning, General Motors reported record third-quarter revenue of $50 billion and net income of $3.3 billion. It raised its free cash flow forecast for the year to as much as $12.5 billion. It had finished January’s $6 billion buyback, and its board approved a new $20 billion authorization with no end date.

Before most plants finished first shift, chief financial officer Paul Jacobson was on television explaining it. The buyback, he said, was driven by “stellar free cash flow generation.” The stock was undervalued, near the bottom of the S&P 500 on earnings multiple. GM has averaged about $7.5 billion a year in buybacks for four years and cut its diluted share count by 37% since 2023.

Within hours, UAW President Shawn Fain answered. He called buybacks “legalized mass looting,” said GM had sold its soul to Wall Street, and promised the union would remember it in 2028.

Two men, one morning, one subject: the cash. Jacobson says it belongs with the owners. Fain says it belongs with the workers. Neither asked what made it. Jim would have recognized both stories. He believed neither.

The middle

For thirty-six years my jobs sat in the middle: supervisor, shift leader, plant manager, chief operating officer. Every one of them stood between the finance office and the floor.

The middle is not where the power is supposed to be. But it is the only place you can see the asset both men are fighting over.

The CFO sees cash. The union president sees wages. The plant manager sees Jim.

From the CFO’s chair, a formed workforce is a cost line. From the union hall, it is a membership and a contract. Only from the floor do you see what it is: thirty years of judgment, built on company time, paid for in wages, owned by no one, recorded nowhere.

That is the power in the middle: sight, not authority. Jim had it from his stool. I came to it from the other side of the line. Neither of us bought the story told up to investors or down to the rank and file.

The hurdle Jim can’t clear

GM’s capital rule is plain: reinvest where the money earns a 20% return, keep the investment-grade rating, return the rest. It is a disciplined rule, and a blind one.

A return needs a denominator. Every press and robot has one, because it sits on the balance sheet. Jim does not. When one company buys another, U.S. accounting folds the assembled workforce into goodwill. Inside the company it is not recorded at all. Training shows up as an expense. Formation shows up nowhere.

So formation never clears the 20% hurdle. Not because it fails the test, but because it can never be entered. The cash goes back to shareholders by default. Jacobson even told analysts one reason cash came in higher this year: capital spending ran below plan.

Nobody in this story is a villain. The system is extractive by design, not by malice. I know, because I ran it for years.

What each man misses

Jacobson’s case rests on a stock the market undervalues. Maybe it does. Or maybe the market sees what the books cannot show: an asset that walks out the door every time a Jim retires, with no write-down, no disclosure, and no line for an analyst to question. A buyback priced off a ledger that cannot see the workforce is a bet made with one eye shut. On Tuesday the market heard the $20 billion and sold anyway: the stock fell about 8% during the day, on doubts that this year’s cash would repeat in 2027.

Fain’s case rests on fair share. He is right that the people who build the profits share less of them than their parents did. But a fair-share argument sends boards to the bargaining calendar. Every contract becomes a fight over a fixed pie, and the pie stays smaller than it could be, because half of what baked it was never measured.

One says the cash belongs to the owners. The other says it belongs to the workers. From the middle, the question is what the cash was made of.

Counting it

We have fixed a blind spot like this before. Until 2019, a public company could lease billions of dollars of buildings and equipment and keep most of it off its balance sheet. ASC 842 brought those leases on as right-of-use assets. The profession decided that an asset a company depends on, but does not own, belongs on the books. A formed workforce fits that description better than a leased forklift.

I am not asking for a GAAP rewrite. I am asking for management accounting. The Bloom is one way to build it: a capability account kept beside the capital-return schedule, a three-year runway, pay that rises with capability, and a statement with each worker’s name on it recording what they actually moved.

That takes nothing from Jacobson’s shareholders and writes no clause into Fain’s contract. It gives the board the denominator.

And the denominator pays. Invest in formation, wire it into the P&L, and the same people throw off more cash, year after year. Today that capability sits on the floor unused, languishing and wasting away, while the board buys back shares. A buyback returns cash. Formation makes it.

Before 2028

Between now and the next contract, every automaker board will approve another buyback, and every one will hear about it from the union hall. Some of them will want a third voice: not the CFO, not the union, but someone who stood where the cash was made.

I own no GM stock and sit at no bargaining table. I have two books coming that make this argument at length.

Jim stayed on his stool, waiting for a car that was not coming, still asking who knows the real story on the floor. If GM’s board could see that line from where he sat, what would it still choose to buy back?

—

Venki Padmanabhan is a co-founder of the Capability Capital Institute and the author of the forthcoming Built to Extract and Already Paid For (Capability Capital Press). He writes at thelonggameforall.substack.com.

Written with AI assistance. The argument, the judgments, and the floor testimony are the author’s own.

Capability Capital Institute — the case for counting what working people know: capabilitycapital.org

A man walks into a bank and asks for a 5-year loan. The officer asks for his financials. He slides across a single sheet, handwritten, that he prepared himself: a list of the assets he says he owns, the income he says he earns, the debts he says he carries. No audited statements. No credit history. No collateral anyone has verified. Just his own attestation that he is good for it. The officer would not reach the second line before declining: you do not commit capital for 5 years against a number the borrower wrote about himself.

Now change 2 words. The sheet is a resume. The loan is a hire. And the institution that would never accept a self-declared balance sheet from a borrower accepts exactly that from every candidate it brings through the door.

The size of the bet

Start with what the hire actually is, because “hire” is the wrong word. When a firm brings someone on, it is not closing a transaction; it is opening an investment. The first 5 years are the Water phase of formation — the deliberate watering of a capability at Year 1, 3, and 5, wage and training and exposure compounding into judgment that did not exist on day 1. Capital is crystallized labor; labor is capital in formation. The decision at the door is whether to fund the forming.

A commitment of that size has a name everywhere else in the building: a capital-allocation decision. The CFO does not approve a 5-year expenditure against a vendor’s self-written quote; he demands audited numbers and a trial. Yet the largest multi-year commitment most firms make — 5 years of formation poured into a human being — is approved against a document the asset wrote about itself plus an hour of conversation. The rigor is exactly inverted: we audit the photocopier lease and take the human being on faith.

Just muscle

Here is the comedy of it. The extraction model does not see selection as a capital decision at all — it sees a slot and a body to fill it. Muscle. So it does the consistent thing: it does not select with care, because anyone will do; it does not train; it does not form; and it discards the worker at the first infraction, because a body is replaceable. Then — the funny part — the same firm is genuinely aggrieved by its turnover, baffled that performance keeps walking out the door, certain the labor market failed it. It manufactured the shortage it laments, and calls the bill an act of God.

The cost is not only the firm’s. It falls on the person treated as interchangeable — selected carelessly, formed not at all, discarded over a lateness, never given the 5 years that would have made them the very worker the firm now swears it cannot find. The resume farce survives because the model beneath it never believed the decision mattered.

Two broken instruments

Selection today runs on 2 instruments, and both measure the wrong thing.

The resume is self-attestation with no evidence trail. Every line is a claim the candidate makes about the candidate; the reader has no auditor, no standard, no way to check. It is the handwritten balance sheet, and we have collectively agreed to pretend it is audited.

The interview is worse in a subtler way. It measures performance under interview conditions — composure, narrative, the charm to talk well about past work — none of which is the capability you are buying. The candidate who tells the best story and the one who can do the work are different people, and the interview rewards the first. Together the instruments give you a self-declared asset validated by a charisma test. No bank on earth would lend against that; we invest 5 years against it daily.

The deeper error

But suppose you could verify every line. It would still measure the wrong variable, because a resume — even a true one — records what a person has been, and the investment is in what they will become. You are not buying the finished article; you are funding a climb. The question selection has to answer is not “who is most formed today?” It is “who is most capable and worthy of ascending through the Water if we invest?”

Those are different questions, and they select different people. The first rewards the candidate who arrived already polished — often because earlier formations were generous to them. The second rewards the one most able to convert investment into capability — a property of trajectory, not position. An investor backs not the company with the most revenue today but the one most able to grow it. Selecting on credentials selects the value when you should select the derivative.

I have watched this gap walk around a plant floor. Todd Pearson came to Lansing Grand River as a supply chain man from Saturn, and he became my door line supervisor. He is now a chief operating officer. The resume that would have screened him at the door would not have predicted the executive he became. What made him worth the investment was never on the page — it was his capacity to ascend, which a self-declared sheet cannot show and an interview cannot test. Select for the page and you miss the Todd Pearsons; select for the climb and you find them.

What exacting selection actually looks like

If the stakes are an investment, selection has to be conducted like one — with the 3 instruments any serious investor uses before committing capital. Diagnosis is half the work; here is the proposal.

First, an audited statement: the Capability Account. Replace the self-declared sheet with an external, evidence-backed, accredited record of what a person can actually do — verified by someone other than the person, portable across employers the way a Social Security number is portable across a life. The candidate stops being the author of their own balance sheet. The audit is performed outside, and the firm reads a statement it can trust.

The world already built this — just not for capability. Before the credit bureau, every lender took your word for it; the bureau replaced self-attestation with a record verified by a third party, standardized so a number means the same thing everywhere, and cumulative across a lifetime, owned by no single lender. We trust that to decide whether you will repay a car loan — and built nothing equivalent for what you can actually do. So a firm weighing a 5-year formation investment has less verified information than a dealership financing a used Civic. The Capability Account is not utopian; it is the overdue extension of infrastructure we already rely on for smaller stakes. With one difference that matters: a credit score is static history — whether you paid. A Capability Account must be dynamic — not only what you can do but how fast you form, the slope of the climb. Credit history tells a lender you are safe; a Capability Account tells a formation investor you will grow.

Second, due diligence: a simulated work environment. No investor wires the money on a document; they run the asset first. Before committing the capital, put the candidate into conditions that resemble the real work and watch capability under load — not narrative under questioning. The simulated environment does to the interview what the audited statement does to the resume: it replaces a performance with evidence.

Third, the decision: a 6-month gate. Give the investment a due-diligence window — roughly 6 months of the Capability Account plus observed performance in real or simulated conditions — then make the call a CFO would recognize: commit the multi-year formation, or do not. 6 months of evidence, then a deliberate yes. Only after that yes does the firm open the full Water phase — the Year 1, 3, and 5 deposits that carry a person to the standard a formed worker reaches by 30.

And the decision belongs to the right officer — not an HR screen run for cost and speed, but the Chief Formation Officer, the second CFO, steward of the labor strand, making a capital-allocation call with the rigor the title implies. The firm’s CFO will not fund a vendor on a self-written quote. The Chief Formation Officer should not fund a 5-year formation on a self-written resume.

The objection, answered

Make selection this exacting and a fair reader flinches: you are building a higher wall. Will fewer people get in — does rigor at the gate become exclusion?

It would — if the gate were the whole system. But the gate sits on top of 3 earlier formations: the Home, the School, the College or vocation that should bring a person to the door already worthy of ascending. So many cannot clear a rigorous bar today not because the bar is unfair but because those formations failed them upstream. The answer is never to lower the bar — a lowered bar funds formation that will not take, wasting the very capital meant to lift people. It is to fix what feeds the gate, so more arrive formed enough to be invested in.

Exacting selection and a society that actually forms its people are not opposites; they are the same commitment seen from two ends. A real bar without formation beneath it is cruelty. Formation beneath it without a real bar is waste. When all flourish requires both: the rigor to invest seriously, and the upstream formation that lets more and more people meet it. The wall is not there to keep people out, but to ensure the investment lands on ground that can grow.

The difference that changes everything

Return to the bank. There is one place the analogy breaks, and it is the most important place. A borrower’s assets are static — the bank audits them once because they will not change. A worker’s capability is not static. It forms. So the audited statement at the door is not the end of the analysis; it is the baseline you compound from.

What a Capability Account should measure in your industry — which proofs of skill, which early signals of formability — is not yet settled, and the firms that work it out first will own an edge competitors cannot quickly copy. That is the open question worth getting in front of.

The discipline itself is not complicated. Select with care. Form with care. Then enjoy what careful selection and real formation return: capability that compounds, reliability you can build a schedule on, people who stay long enough to become your best — and, on the scoreboard the CFO actually reads, growing revenue and profit. No one would lend 5 years against a sheet a stranger wrote about himself. Stop investing 5 years against one — and let the asset you chose pay you back.

sarve janāḥ sukhino bhavantu — when all flourish, the owners flourish too.

Venki Padmanabhan is a co-founder of the Capability Capital Institute and the author of the forthcoming Built to Extract and Already Paid For (Capability Capital Press). He writes at thelonggameforall.substack.com.

Written with AI assistance. The argument, the judgments, and the floor testimony are the author’s own.

In 1974, a management professor named Jerry Harvey published a story about his in-laws.

It was a July afternoon in Coleman, Texas, population 5,607. One hundred and four degrees, according to the thermometer at the Rexall. West Texas topsoil blowing fine through the house. But the afternoon was tolerable — pleasant, even. A fan on the back porch. Cold lemonade. And dominoes, a game that asks nothing of a man in that heat but to mumble shuffle ‘em and move his arm.

Then Harvey’s father-in-law said: let’s get in the car and drive to Abilene and have dinner at the cafeteria.

Abilene was fifty-three miles away. The car was a 1958 Buick with no air conditioning.

Harvey thought this was madness. Before he could say so, his wife said it sounded wonderful — and asked him what he thought. Finding himself apparently alone, he said it sounded good to him too. He added that he hoped his mother-in-law wanted to go. Of course she wanted to go, she said. She hadn’t been to Abilene in a long time.

So they went. The dust cemented itself to their skin with their own sweat. The cafeteria food was the kind you make antacid commercials about. Four hours. A hundred and six miles. They arrived home finished and sat back down on the porch under the fan.

To be polite, somebody said it had been a great trip.

The mother-in-law said no. Not really. She’d have rather stayed home. She only went because the other three were so worked up about going.

Harvey said he hadn’t wanted to go either. He’d gone to keep everyone else happy.

His wife said the same. She’d only gone because it was so hot and the rest of them seemed to want it.

And then the father-in-law — the man who proposed the trip, the author of the whole afternoon — said he had never wanted to go to Abilene in his life. He’d only suggested it because he thought the other three looked bored.

Four people sat on a porch in Coleman, Texas, having just driven a hundred and six miles through a dust storm to eat food none of them wanted, in a direction none of them had chosen, at nobody’s request.

Harvey called it the Abilene Paradox, and the argument he built on it is the reason it lasted. Organizations, he said, do not fail because people disagree. They fail because people agree — silently, separately, everyone in the car reading everyone else’s silence as enthusiasm. The problem is not the management of conflict. It is the management of agreement.

He was clear-eyed about the stakes, too. Some organizations take a side trip, he wrote. Others take terminal journeys to Abilene — when Dallas or Houston or Tokyo was where they actually wanted to go.

For thirty years it was standard curriculum. I was taught it at General Motors. Every executive I came up with was taught it. It has the shape of a joke and the sting of a confession, and it works because everyone in the room recognizes the drive they are on right now.

I want to tell you what I never noticed for thirty years.

The road had houses on it

Harvey drove fifty-three miles across West Texas. There were people living along that road.

He never mentions them. He had no reason to — it wasn’t his subject. But the curriculum never mentioned them either. Nobody in thirty years of teaching this thing ever stopped to ask what the trip cost the people the car drove past. And I never asked. I sat in those rooms and I nodded at the joke and I went back to work.

The Abilene Paradox has four people in a car and a moral architecture built entirely out of their unspoken preferences. It is a masterpiece of self-criticism — and it trained two generations of executives to interrogate the silence of the passengers and never once the silence of the fenceline. Our most honest management parable still only counts the people with seats.

The dateline

On July 9, 2026, WIRED published an investigation by Evan Simon of Floodlight. It concerns Omaira Garcia, an Air Force veteran, who bought a small ranch with her husband after more than a year of looking. Their piece of heaven in the country. Their kids loved it.

She learned that OpenAI was building its flagship Stargate data center directly beside her property only after the construction had started — when the dust began to engulf her house. Today a natural-gas power plant sits roughly five hundred yards from her home. She can see the exhaust stacks from her kitchen window. At night, she says, the light is like being inside a stadium.

She tried to sell. She did not get a single offer. A realtor suggested she turn the house into an Airbnb for the data center’s workers.

“We’re trapped here,” she said.

Her neighbor Paul Daniel is eighty-one. His family has lived in that area for more than two hundred years; he has lived in his own house for forty-four. Nobody told him either. He made a sign out of a plank — one word and an arrow, pointing toward the entrance — and put it on the road leading into the site.

The word is HELL. The arrow points to Stargate.

The ranch is in Abilene, Texas.

Fifty-three miles from Coleman.

The permit is the paradox, written into law

Normally, if you intend to build a major new source of air pollution, you need a major air permit: environmental review, public notice, a comment period, a chance for the people who breathe the air to say something about it before it exists.

Stargate’s developers didn’t use it. In 2024 they secured their on-site power through minor permits — “permits by rule” and standard permits — the instruments ordinarily used for dry cleaners, gas stations, auto body shops. Small things that happen a lot, as one former EPA air enforcement chief put it. No environmental study. No public notice. No comment period. Granted fast and quietly.

Stop on that list, because this is where I stop being a commentator.

I have run paint shops — in the United States, in Germany, and in India. I have sat on the wrong side of the table from air regulators on three continents. A paint shop is the largest emissions source in an assembly plant and usually its single most expensive building. You are permitting volatile organic compounds and hazardous air pollutants — solvents, isocyanates. You install a thermal oxidizer that burns fuel around the clock destroying what you emit. You monitor continuously. You stack test. You file deviation reports. You are inspected. It takes years, and then it never ends, because a permit is not a document you obtain — it is a condition you maintain, and every year somebody comes to check. The answer in all three countries was the same: prove it, and keep proving it.

So I know exactly where the line sits between an auto body shop and a paint shop, because I have spent my career on the far side of it. The permit by rule is for the corner body shop with a booth and a fan. It was never for me.

I never once got a paint shop through on a permit by rule.

Stargate got ten gas turbines and sixty-two diesel generators through on one.

Under those dry-cleaner permits, Stargate is currently allowed to run ten gas turbines and sixty-two backup diesel generators, and to emit more than 1.6 million tons of greenhouse gases and a thousand tons of harmful air pollutants a year.

Sixty-two. The rule was written for an operator who wanted a backup generator. Maybe three.

The year after, having built the thing, they filed their first major permit application — for forty-one more turbines and eighteen more generators. If it goes through, Stargate becomes one of the largest fossil fuel power plants in Texas, emitting more greenhouse gas annually than roughly two million cars.

Watchdogs have a name for this shape: small first, big later. The EPA’s own handbooks have a harsher one. Minor permits that precede major ones are called sham permits.

Garcia found out about the expansion when a reporter told her.

And Stargate is not unusual. Since 2024, at least thirty-eight Texas data centers have taken the same route, and the state has quietly sanctioned more than 2,100 backup diesel generators. More than half of these facilities filed nitrogen oxide estimates that landed just barely under the threshold that would have triggered public input. Outside San Antonio, one came in at 99.8 tons a year against a 100-ton trigger.

Two tenths of a ton. That is not an engineering number. That is a number that was chosen.

Understand what the permit by rule actually is. It is not a loophole in the sense of an oversight. It is an instrument that defines who counts as being in the vehicle. It is the four-person car, drafted by lawyers and signed by a state agency. Everything outside it — Garcia’s kitchen window, Daniel’s two hundred years — is road.

The agency grading its own exam

Asked about all this, the Texas Commission on Environmental Quality said its low number of enforcement actions reflects high compliance rather than an absence of enforcement.

Read that again. We are not writing many tickets, which proves nobody is speeding.

Its backlog runs past fourteen hundred unresolved cases. It cleared thirty-nine last year. And the people saying so are not activists — they are the agency’s own alumni and a Republican-administration EPA enforcement chief. Every one of them is saying it now.

Nobody said it in the driveway.

What Stargate does not need

Now the part that is mine to say, because I was in the car.

The Abilene campus is eleven hundred acres. Roughly four million square feet of building. A gas plant that could power every home in Abilene twenty times over.

It employs almost nobody.

Buick City needed twenty-eight thousand people. It needed them the way a lung needs air. And because it needed them, it formed them — not out of generosity, not out of any theory of human development, but because it could not run otherwise. It made millwrights and diemakers and general foremen because a plant that size cannot exist without them. The formation was real, it was expensive, it was permanent, and it was almost entirely accidental.

Flint was destroyed by capital leaving. Abilene is being destroyed by capital arriving — same wound, opposite directions. And Flint drove itself there. I knew. The most-taught parable in American management is a story about four people ruining an afternoon in a Buick, and we taught it inside the company that built the Buick, and it did not save us.

And this dateline says it for me:

Formation was never a gift. It was a byproduct of needing people.

Remove the need and the byproduct stops. A hyperscale data center is the first industrial asset in history that does not require a workforce in order to run. It requires land, water, gas, and silicon, and it requires a construction wave that leaves. It does not require a town. It does not require anyone to become anything.

This is why the extraction in Abilene feels so naked. There is no exchange to point at. The factory at least had to make somebody in order to make something. Stargate does not — not because Crusoe is wicked, but because the asset has no mechanism for it.

The fuse is longer than the tenure

Capital has run without the labor strand before. It works. That is the whole problem — it works, and it keeps working for about as long as one man’s tenure.

Pullman, 1894.

George Pullman built a town for his workers south of Chicago, and by the standards of its day it was magnificent. Indoor plumbing. Gas, water, sewers. A library, a church, parks, a theater. Visitors came from Europe to admire it. He funded every bit of it — and the word that matters in that sentence is he. Every brick was an asset on the Pullman Palace Car Company’s balance sheet. None of it was in anybody’s name.

Then the depression of 1893 arrived, and Pullman cut wages by roughly a quarter and did not cut the rents. The rents were revenue on an asset. The wages were a cost.

That is the whole lesson, and it took fourteen years to arrive. The men had been given a library and not a ledger. They could not carry it, sell it, borrow against it, or stop the man who owned it from raising their rent on the day he cut their pay. The strike became a national boycott. Federal troops came to Chicago. Men died in the yards. Pullman died in 1897, and in 1898 the Illinois Supreme Court ordered the company to sell the town off.

Crusoe’s fire trucks are Pullman’s library.

Ford, 1914.

The same lesson from the other side, and the reason I am not asking anybody to be generous.

By 1913 Henry Ford had a turnover problem that ought to be taught in every business school on earth and is taught in almost none. The moving line had deskilled the work so completely that the work had become unbearable, and annual turnover at Highland Park ran around 370 percent. Ford was hiring on the order of fifty thousand men a year to hold a workforce of about fourteen thousand. The machine had severed the labor strand exactly as designed, and the asset was choking on it.

So in January 1914 he doubled the wage to five dollars a day.

It has been remembered ever since as philanthropy, or as marketing — the story that he was paying his men enough to buy the cars. It was neither. It was a man buying back the strand his own machine had cut, because the machine could not run without it. Turnover collapsed. Quality rose. Training cost fell off a cliff. It paid for itself inside a year.

Ford is the proof that depositing is not charity. Ford is the proof that depositing is arithmetic.

Boeing, 2001 to now.

In 2001 Boeing moved its headquarters to Chicago, deliberately away from the people who build airplanes. Over the next decade it pushed the 787’s design and fabrication across a global supply chain on the theory that formation could be somebody else’s cost. The engineering judgment nearest the work was routed around — not once, but as policy.

Two 737 MAX aircraft went down. Three hundred and forty-six people. Then a door plug left an Alaska Airlines fuselage in flight in January 2024.

Now watch the fuse. Boeing admitted a conspiracy to defraud the United States in a deferred prosecution agreement in 2021. In 2024 the Justice Department found it had breached that agreement. A guilty plea was negotiated; the court rejected it. The Department then reached a non-prosecution agreement, and in November 2025 the charge was dismissed. The families are still in the Fifth Circuit.

Every executive who made the decision was gone. Not one of them was in the building when the bill came.

That is the pattern, and it is the only sentence in this essay I would defend to the last:

Severing the labor strand always looks good for exactly one tenure.

The deposit you skip is booked this quarter. The depletion lands after you have left. So the trip to Abilene is not irrational at all. It is perfectly rational, for the driver.

The trip is fifty-three miles. The bonus is paid at mile twenty.

That is why nobody says Coleman. It isn’t cowardice. It’s arithmetic.

So run the fuse on this one. The silicon in a data center depreciates in five or six years. The turbine runs thirty. Omaira Garcia’s lungs have to run forty. There is not one person in that car who will be present for the back half of any of those numbers, and every one of them knows it.

The two ledgers

Before you get in a car you settle two things. Why are we going. What are we going for.

Harvey’s four people never said either one out loud. That is the entire parable. Nobody in Abilene said them either — and the proof of that is not an argument I have to make. It is a pair of columns.

Here is everything the record knows about the capital strand at Abilene:

Eleven hundred acres.

Four million square feet.

Ten gas turbines. Forty-one more applied for.

Sixty-two diesel generators. Eighteen more applied for.

Three hundred and sixty megawatts, headed for twelve hundred.

One-point-six million tons of greenhouse gas a year.

Over a billion dollars in phase one.

An eighty-five percent property tax abatement.

Fifteen hundred people building it, around the clock.

Eighty gigawatts of new gas in the Texas queue — second in the world only to China.

Here is everything the record knows about the labor strand at Abilene:

Fifty-seven.

That is the contractual floor for full-time jobs. It is the entire second column.

Every number in the first list existed before a shovel moved. Each one was engineered, financed, insured, permitted, disclosed. Somebody was paid to be certain of it. Capital counts what it owns, to two decimal places, in advance.

The second list has one number in it, it is a legal minimum, and there is no line anywhere — not in the permit, not in the impact study, not in the 10-K, not in the governor’s press release — for the only question that matters.

How many people did this thing form?

Not employ. Form. Nobody knows. Nobody is required to know. Nobody has been asked.

What is the net formation?

I am not telling anyone to stop building. Abilene got fire trucks. It got road improvements and school expansions, and those are not nothing, and any argument that pretends they are nothing deserves to lose. The eighty gigawatts will get built by someone. The turbines will be installed and piped and wired and commissioned and then maintained for thirty years by human beings who have to be made first, because nobody is born knowing how.

Ask it of the construction site. Ask it of the room where the model gets used.

The developer says the fifty-seven will really be more than a hundred. The mayor says two hundred and fifty. Take the mayor’s number. Take double it. It doesn’t matter — those are jobs, and a job is a rental agreement. It tells you a seat was filled. It tells you nothing about whether anybody became anything.

And the construction number is worse than it looks. Go read the hiring notices for that site. Certified crane operators. Surveying journeymen. Heavy equipment diesel mechanics. Read those words again — certified, journeyman. Every one of them describes a man who was already formed, somewhere else, by somebody else, years ago, in a hall this project did not build and does not fund.

The build-out is not making tradesmen. It is bidding for them.

That is a withdrawal. It draws down a stock a previous generation paid for, deposits nothing back, and when the campus is done those men leave and take it with them. Two years, around the clock — and the honest net formation number for that site may be close to zero.

Now the other site, the one nobody is looking at at all. Go into any office where a model now does the first draft, the first analysis, the first pass at the code. Ask what that work was ever actually for.

It was never for the output. The first draft was always bad. The junior’s spreadsheet always had to be rebuilt. That work existed to make the junior. It was the rung. It was expensive and inefficient and it was the most wasteful thing in the whole enterprise, and it was how you got a senior fifteen years later.

Automate the rung and you keep the output and you lose the senior.

I am not the first to notice. Erik Brynjolfsson’s team at Stanford tracks it in payroll data: a sixteen percent relative decline in employment for workers aged twenty-two to twenty-five in the occupations most exposed to AI, while experienced workers in those same jobs hold steady. Their term for it is task substitution at the apprentice margin. And serious economists dispute the finding.

But look at what that fight is about. Headcount. Both sides are counting jobs — which means both sides are holding the gauge I have spent this essay telling you is the wrong one. A junior can be employed and unformed. If the skeptics win outright, if entry-level employment never drops another point, my question is still unanswered. Nobody has asked it.

And look at the fuse on it. You book the saving this quarter. The absence arrives in fifteen years, when there is nobody in the building formed well enough to know the machine is wrong. That is one tenure. That is Pullman and Ford and Boeing, in a room with no exhaust stack in it.

So here is the ask, and it is small.

One formed person per megawatt.

Abilene is headed for twelve hundred megawatts. Twelve hundred people.

Megawatts, because that number is already in the permit — filed, reviewed by ERCOT, impossible to bury in a footnote. Nobody can hide a turbine. And because it is the honest index: the megawatts are the gas plant. The thing that fouls Omaira Garcia’s air is the same thing that sets the price.

If you would rather picture it than audit it — the campus is about eleven hundred acres. Call it one formed person per acre. Same answer.

And if you want to argue the number down, take any other unit on the developer’s own ledger. You land in the same place:

One per acre: eleven hundred.

One per four thousand square feet: a thousand.

One per million dollars of phase-one infrastructure: eleven hundred.

Or just the gap — fifteen hundred men came to build it, fifty-seven stay: fourteen hundred and forty-three.

Every unit the capital strand chose for itself prices the labor strand between a thousand people and fifteen hundred.

I did not pick twelve hundred. The developer’s own filings picked it, four different ways. It is the asset, weighed in the one currency it refuses to count.

And not twelve hundred jobs. Twelve hundred formed people, carried all the way through to journeyman, employable in Abilene and employable anywhere. That last clause is the whole thing. Company training makes you employable here. Formation makes you employable anywhere.

The difference between them is the difference between a library and a ledger.

Then scale it. Eighty gigawatts in the Texas construction queue is eighty thousand people. And publish the count, so the second column has something in it besides a legal minimum.

Now notice when it happens. Not in a classroom. Not afterward. During the build. Fifteen hundred men are already on that site, around the clock, for two and a half years. Indenture them there. Year one and year three land on Abilene’s campus. Year five lands on the next campus, because there is always a next campus — and the record travels with the man and not the employer, which is the entire reason this has to be an account and not a program.

The build-out forms the workforce for the build-out.

Call it a sin tax if you like, but be precise about the sin. The sin is not the data center. The sin is building a thing this size and forming nobody — which makes it a levy a developer can zero out completely by making people. That is the only kind of tax I have ever respected: the kind whose entire purpose is to be avoided by doing the right thing.

And it is not even new money. Abilene handed the developer an eighty-five percent property tax abatement to build there. The public is already paying. I am not asking Texas for a dollar it hasn’t already committed — only that the dollars it committed be conditioned on a number.

Britain tried this. The Apprenticeship Levy has run since 2017 — half a percent of payroll — and the way it failed is the most useful thing I can give you. Apprenticeships for under-twenty-fives fell by a third. Four hundred million pounds a year went to people who already held a degree. The single most popular programme funded by the apprenticeship levy was Senior Leader.

Britain built the levy without the gauge, and the money walked straight to the people who were already formed.

So: the number, then the levy, then the ledger, in that order — and the ledger verified by somebody who did not do the training. Harvey’s remedy was to speak up in the car. Garcia was never in the car. So widen it. But you cannot widen what you refuse to count.

And before anyone calls this a cost — somebody has to build eighty gigawatts. There are not eighty thousand spare journeymen sitting in Texas waiting for the call. That is exactly why the hiring notices say certified. The industry’s scarcest input is the thing it refuses to make, and it is bidding against itself for a stock that is not being replaced.

Ford had this identical problem in 1913. His turnover was three hundred and seventy percent. He did not solve it with a training initiative. He bought back the strand his own machine had cut, and it made him the richest manufacturer on earth.

The deposit was not the price of doing the right thing. The deposit was the asset.

And source them from inside a hundred miles.

The Abilene education region is forty-two school districts, thirteen counties, a hundred and sixty campuses, fifty thousand kids. Three graduating classes is about ten thousand young people. Take all ten thousand. Select twelve hundred. One in eight — an achievement, not a consolation prize, which is exactly why they will come. And draw the radius by the harm, not by an administrative line. Whoever breathes it has first claim on the formation.

Fifty-three miles down that road, inside the radius, there is a high school in Coleman, Texas.

I have to ask it of myself

I wrote this with a machine. It is disclosed below. So the question I am putting to Crusoe and to Abilene and to every company in America is one I have to put to myself, out loud, in front of you.

The machine drafted and retrieved and arranged. The judgment is thirty years of paint shops on three continents, and no machine supplied that, and none could.

But I already have the thirty years. Somebody paid for them. GM paid. Chrysler paid. Chennai paid. And they paid in the only currency formation takes, which is time on a floor with a man watching me who had done it before. That is the deposit I am spending here, and I did not make it.

The question is not whether the machine augments me. It does. The question is where the next man gets his thirty years, if the rung I climbed is the exact rung we are now automating.

I don’t have an answer to that one. It is the only question in this essay I cannot put a number on.

Omaira Garcia’s kitchen window is the gauge. Somebody go read it.

Harvey’s family got home. Garcia cannot. That is what he meant by a terminal journey.

The trip is on. The heat is brutal. Everyone in the car is nodding.

Somebody say Coleman.

Venki Padmanabhan is a co-founder of the Capability Capital Institute and the author of the forthcoming Built to Extract and Already Paid For (Capability Capital Press). He writes at thelonggameforall.substack.com.

Written with AI assistance. The argument, the judgments, and the floor testimony are the author’s own.

Dear Charles, Kumar, and Jim,

I went back to GM’s Lansing Grand River plant after twenty years. The first person I ran into still remembered my team-leader coaching classes. Then I went looking for Ramon Hernandez.

Ramon had been my team leader on instrument-panel install. There was no better one. I found him working as a material handler, moving parts from one end of the plant to the other. He was cordial. Then he drove away on his tugger, slowly, and I watched him go. He was waiting out his last year before an early retirement.

I thought of him in June, Charles, when you said the thing most executives spend an entire career not saying:

“Mistakenly we thought that by just introducing artificial intelligence and ingesting the design requirements that we had, that that would produce a high-quality product.”

Mistakenly we thought. That is a confession, on the record, from the executive whose job was to be right about this.

Over roughly three years you hired, rehired and promoted about 350 veteran engineers — some former employees, some pulled off supplier benches — the people you call, without irony, your “gray beards.” Their assignment: hunt for failure points before a part ever reaches the plant floor, run the quality reviews, train the younger engineers, and reprogram the very AI tools that had failed them. Kumar, you said the company had been “relying more and more on automated quality systems” and not getting what it paid for.

Jim, you called the result a tailwind: lower warranty and recall costs, “hundreds and hundreds of millions of dollars of a tailwind for Ford on cost.” In June you took the top mainstream spot in the J.D. Power Initial Quality Study — above Toyota, above Honda.

I have spent thirty-six years on factory floors on three continents arguing one idea three ways. Your own numbers now make the case for all three: Augment people, don’t replace them. Capability is the only profit that compounds. Form people to their full capability. Together they say that quality was never Job One. It is the dividend of forming people. That is the argument. The rest of this letter is the proof — and the one move you still haven’t made.

* * *

Augment people, don’t replace them.

Charles, you said the truer thing in the same breath: artificial intelligence “is only as good as the information you use to train it.” You poured your design requirements into the machine and waited for quality to fall out the other end. You left out the one ingredient that was never in the requirements document — the judgment of people who had lived through enough product cycles to know exactly where a part will lie to you.

I have watched one person do exactly what those 350 are now doing. It was Ramon. During the CTS launch at Lansing Grand River, we had a chronic defect — the instrument panel rubbing the door opening as it went in. The easy move was to blame the operator. Ramon did the opposite. He took the blame onto himself, gathered the data shift after shift, and traced the fault back to a tool mechanism on his own team. He found the failure point before it became a warranty claim, and pulled the repair cost off the plant’s P&L. No system handed him that. He had been formed to look.

Ramon was an hourly team leader on a trim line. Your 350 are salaried engineers. The asset is the same.

You kept the AI and put formed humans back on top of it. The gray beards now reprogram the tools. They are the layer that tells the model which failures to look for, because they have met those failures in person. You have the warranty data now.

* * *

Capability is the only profit that compounds.

For a generation, the experience walking out of a plant at retirement has been booked as a cost coming offthe payroll — a salary saved. That is how the ledger sees a gray beard leaving: a line item going to zero, a small win. Your number says the ledger is lying. Those people were not a cost. They were the asset. And the proof is that you had to go and find that judgment again — some of it on supplier benches — to get the quality back.

Divide a “hundreds of millions” tailwind by roughly 350 people and you get — by my arithmetic, not yours, and crediting them with all of it — something on the order of a million dollars a head. It shows up as lower warranty reserves, fewer recalls, a J.D. Power ranking that moves metal off lots. That tailwind is a dividend showing up on the P&L. The asset that pays it is the people, and it appears on no balance sheet. My cousin and co-author Sridhar Ramamoorti has spent his career arguing that our accounting standards can already hold this asset — that the categories exist and we refuse to aim them at the right thing. You just aimed them, involuntarily, and the number came back in nine figures.

You have now written that number down, in public. It compounds whether or not the accounting is brave enough to record it.

* * *

Form people to their full capability.

Kumar, you said something in June more important than the rehiring itself. You said Ford is “moving from a find-and-fix mentality to preventing issues before they occur.” You meant parts — catch the defect upstream, before it reaches the floor, instead of chasing it through a warranty claim.

It is the correct principle. And you applied it to everything except your own people.

You let your most experienced engineers walk out the door. Then, a great many warranty claims later, you went looking for that judgment again. That is find-and-fix applied to human capital — the most expensive version of it that exists. You built a prevention culture for steel and a find-and-fix culture for talent, and never noticed the contradiction.

There is a preventive version. It is the heart of what we are building, and it fits in one sentence: take your most senior people in their final three years and use them to form the next cohort — so that the master’s last three years become the apprentice’s first three.

I mean a structured apprenticeship with a real curriculum, in which the veteran’s judgment is transferred deliberately, and each transfer is verified and recorded. We are building the record for that, and we call it a Capability Account: a lifetime ledger the worker owns, an outside authority accredits, and the employer pays into. A deposit is made when the apprentice can demonstrate the capability under the master’s sign-off, and prove it to someone who didn’t train them.

It is a design, and it is not yet proven. What a deposit is worth, and what becomes of it when a person leaves, are still open. Our first pilots are meant to answer them.

One thing I would like to learn from you. The last three years are the opportunity, and tenure alone qualifies no one. You rehired technical specialists, people whose experience had compounded into transferable judgment. Plenty of veterans can do the work. Few can form another person. How did you tell which was which? You have just had to do it 350 times.

Use the last three years that way, and you may not have to buy back your own experience again.

You once built an identity on three words: Quality Is Job 1. It ran seventeen years, and in its best years the cars bore it out — because the quality came from people. That was the Deming era, when Ford humbled itself, retrained its workforce, and let the floor catch its own defects. I owe that era a personal debt: in 1994 Don Petersen, the man who made Quality Is Job 1 Ford’s creed, put the Engineering Society of Detroit’s Outstanding Young Engineer award in my hands.

So I do not say this lightly. Quality was never Job One. It was the dividend, and formation was the job.

* * *

Ford let that judgment leave, watched the quality fall, and spent years buying it back. Ford printed the receipt. It says the people were the asset all along.

General Motors never printed a receipt for Ramon. From that trim shop he rose to the pilot team that launched every car LGR built. He is the kind of asset you have just spent years recovering. He belonged in a room with the engineers, working out how to build the next great car. He spent his final working years on the tugger, and he retired last year.

That failure is not his — a man does not owe his employer the wisdom to use him well. The waste is ours: a system with no account to hold what he had become, and no way to make his last three years the first three of the ten Ramons who should have followed. We let the most formed judgment on that floor depreciate to zero on a loading dock and called it a pension. I wrote this letter for him.

The cheaper way has been on the floor the entire time, in the last three years of every master we are about to lose.

That is the Long Game.

So, Charles, Kumar, Jim: somewhere at Ford today there is a gray beard in the last three years of a career. Who is learning at that elbow, and where is it being written down? If you were building the account for it, what would you put in it that I have left out? I would be glad to build it with you. And to anyone reading over their shoulders: the same questions are yours.

Respectfully,

— V

This letter is the second of a pair. The first is Ford Has Read This Book Before.

Venki Padmanabhan is a co-founder of the Capability Capital Institute and the author of the forthcoming Built to Extract and Already Paid For (Capability Capital Press). He writes at thelonggameforall.substack.com.

Written with AI assistance. The argument, the judgments, and the floor testimony are the author’s own.

Capability Capital Institute — the case for counting what working people know: capabilitycapital.org

It takes some nerve for a GM man to recommend a book to Ford. In my defense, GM did not want it.

In 1943, General Motors invited a young Viennese writer named Peter Drucker inside. He spent the better part of two years in its meetings and its plants. In 1946 he published what he saw there as Concept of the Corporation.

The book admired GM. It also said two things GM did not care to hear. A corporation is a social institution, and it answers to more than its owners. And the worker is a resource to be developed, not a cost to be controlled.

GM’s president, Charles Wilson, was curious enough to ask. In 1947 the company ran an essay contest for its employees called “My Job and Why I Like It.” About 175,000 of them wrote in. Drucker thought it the richest record of what workers want that anyone had ever gathered. Then the findings went into a drawer, and the idea went with them.

Alfred Sloan’s memoir never mentions the book. Drucker wrote years later that it had shaped business everywhere “and none on General Motors.”

Across town in Dearborn, a twenty-eight-year-old named Henry Ford II had just inherited a company in disarray. He hired managers out of GM. By Drucker’s account, and by the historians who followed, that team rebuilt Ford with his book open on the desk.

Ford read what GM shelved. That is the first thing I would like Jim Farley to know about the debate he has started. His company has done this before.

Farley is counting the workers America is missing. Nobody is counting what the workers it already has can do. That capability is the largest reserve of skilled labor in the country. Most of it stands idle, and it will stay idle until someone forms it on the floor and writes it down where the business can see it.

That’s the argument.

What Farley got right

For two years Farley has been telling the country it has neglected what he calls the Essential Economy: everything that gets built, moved, or fixed. That is ninety-five million people.

On September 30, at his second summit at Michigan Central, he gave the number. America has to fill 1.7 million skilled trades openings a year for the next ten years. Training programs graduate 55 people for every 100 needed. Nearly a quarter of the skilled workforce is 55 or older. He called it “a total killer for productivity.”

He is right. He is also doing more than talking: scholarships, an alliance of employers, a national report, an Ad Council campaign. I have no quarrel with a word of it.

I have an addition. Every one of those numbers counts people who are not there yet. None of them counts the people who are.

His word for all this is pipeline, and it is a fair one. If you want more water out the far end of a long pipe, you start by putting more in. But this pipe leaks. The report released at his own summit found that only one in two people who enroll in trades training finish. Inside the plant the loss is larger. My own estimate, worked out in a forthcoming book, is that a typical frontline operation uses about 30 percent of the capability it is already paying for. Treat that as an estimate. Halve the stranded share if you like. It is still a great deal of skilled labor, and no training program has to graduate it.

The man they flew in

I came to General Motors in 1989 still writing a doctorate on diagnostic expert systems. At Flint Metal Fab we had huge transfer presses with complicated Unico controllers overhead. When a press went down and the electricians were flummoxed, they flew in a man named Jim from Chicago to troubleshoot it. The expert system I called Trades tried to emulate what Jim knew.

Trades may have cut the diagnosis time for a novice electrician. The crew still lost the hour hunting through the basement for spare boards. And Jim still flew in. I did not dent that expense much. What he knew was too vast, and most of it had never been said out loud.

Three decades later Ford ran the same experiment with better machines. It installed some 900 AI inspection cameras and leaned on automated quality systems. They missed what experienced people catch. So over three years Ford hired, rehired and promoted 350 veterans. It calls them gray beards. These are salaried engineers and technical specialists, people with degrees and titles. Charles Poon, who runs vehicle hardware engineering, said plainly what had gone wrong. Ford had let its experienced engineers go before their knowledge could teach the machines.

In June, Ford topped J.D. Power’s initial quality ranking among mainstream brands for the first time in sixteen years. Farley counts the savings in the hundreds of millions. Ford still leads the country in recalls, and its chief operating officer says that number trails. We will see.

The quality came from capability that already existed, some of it formed on Ford’s own payroll, put back where it could decide things. I recognized the story. I had built the 1989 version.

Grace

I have also seen it done the other way.

On night shift at GM’s Lansing Delta Township plant, there was no one upstairs in engineering to call. I was a shift leader. When the line went down, and we had eliminated all the Neanderthal mechanical culprits, and what was left was a controls fault, I got on the radio.

The clock on that call is short. When a trim shop stops at fifty-five jobs an hour, a hundred and fifty people stop with it. Fifteen cars sit in the buffer to chassis. In about fifteen minutes they are gone, chassis starves, and two hundred and fifty more people come to a stop. The plant loses at least five thousand dollars of gross margin for every minute it stands.

Within minutes of the radio call, Grace Major would roll down with her apprentice in tow. As she wheeled into the area she gave me the long stare. It meant hold your horses, and don’t bug me. So I hovered, looking for small signs of progress. She gave me an update only when the progress was real and she was sure of the fix, down to when to send people off early on break to save the losses. Now and then, to keep me busy, she would enlist me: watch that light, and let me know when it changes. I was delighted to be of use.

The worst nights were up in the mezzanine, at the transfer from paint to trim, where the robots had to pull open the hood to inscribe the vehicle number. Hairy. She always found a way to get it running.

Grace retired while I was there. After she left we were comfortable, because her apprentice had learned the work under her, right in front of our eyes. His name is Malachi Smith. GM published a profile of him this year. It does not name her.

Nobody put Grace in a computer. She put what she knew into Malachi, on nights, with the line down, and no ledger at General Motors shows that it happened.

The half still on the shelf

Here is the part of the history I left out. What Ford took from Drucker in 1946 was the half about managers: divisions, professional management, controls. The half about the worker stayed on the shelf in both towns.

Farley’s summit is the nearest an American automaker has come to opening it, and he and one of his guests were already reaching for the page. Larry Fink said we should invest in people’s futures the way we invest for their retirement. Farley himself told Fortune in Detroit that week that the line between engineers and hourly skilled-trades workers “is completely blurred out.”

Take them both at their word. Retirement has an account. It has a statement, a balance and an employer’s contribution, and it goes with you when you leave. Capability has none of that. Ford can tell you to the dollar what 900 cameras cost. I doubt anyone could have said what the gray beards knew on the day they walked out. That is why they were allowed to.

So here is what I would bring to the table: a capability account. It would show what a person can actually do, confirmed on the floor, visible to the worker like a paycheck, and portable. It would have no column for collar. A line that is blurred on the floor has no business staying sharp in the books. The engineer, the electrician and the operator would be entries in the same book.

The account is the easy part to describe. Capability has to be formed before it can be recorded. That takes a place safe enough to learn in, a wage that rises with what a person can do, and a floor leader who teaches the way Grace did. We call our version of that system Bloom. It is a working design and it is not finished. I do not yet know how a deposit in that account should be valued, or how the books should treat it when a person leaves. Our first pilots are meant to find out.

There will be a third summit, and it will have a report. The report will have a column for the workers America is missing. Mr. Farley, what would it say if it had one for the workers America already has?

Ford built a company on the first half of that book. The second half is still on the shelf. Counting what working people know, and forming more of it, is the Long Game.

I would be glad to help you count. And to everyone else who runs a floor: who was your Grace, where is it written down, and what belongs in that account that I have left out?

Venki Padmanabhan is a co-founder of the Capability Capital Institute and the author of the forthcoming Built to Extract and Already Paid For (Capability Capital Press). He writes at thelonggameforall.substack.com.

Written with AI assistance. The argument, the judgments, and the floor testimony are the author’s own.

Capability Capital Institute — the case for counting what working people know: capabilitycapital.org

Trim 2 was the bottleneck at Lansing Delta Township. Night shift, nine hours, and most nights it gave up fifty or sixty minutes of downtime. Supervisors had tried. Everyone above them had tried. I had tried.

Then Abdelhak Radjai came to run Trim 1 and Trim 2 on nights.

He had been one of the best team leaders on that floor, and he never stopped acting like one. He would jump on any job on his line so someone could take a bathroom break, which is the highest-value currency on a factory floor. And he watched. He saw a new team leader drowning in repairs before she had mastered the six jobs on her own team, so he had another team leader cover them and sent her back to build her footing. He saw team leaders who never left their zone to chase a repair, and he parked them off the line, inspecting and fixing the defects coming out of his team and the four others in Trim 2.

Within a month the shift was closing under ten minutes.

Look at what he did. He observed the system, changed it, tested the change, and kept what worked. And the thing he improved was the people who improve the line. The improver improved the improvers.

Then Abdelhak moved on. One of the team leaders who had pulled for him went out on sick leave. Another left for the pilot team. The people who came in behind them held the line anyway, because Abdelhak had set the standard.

Now look for that standard in the plant’s books. The downtime number moved. The capability that moved it, and kept it moved after he was gone, appears nowhere.

That loop now has a name and a funding round. On August 5, Jeff Dean, Sanjay Ghemawat, Oriol Vinyals, and Quoc Le founded Discovery Loop to automate the experimental loop of research, including using AI to build more powerful AI. The field calls it recursive self-improvement. Another firm, Recursive Superintelligence, raised some $650 million within months of its founding.

We are ready to afford that loop to machines, in both senses of the word. We will pay for it and we will permit it. The people who already run it, in the frontlines of six industries across the United States and India, get neither. That’s the argument.

What the machine is given

Look at what a self-improving system requires, and what its builders gladly provide.

It gets its own performance data, in full, in real time. It gets permission to change itself; the whole design assumes it will propose changes and some will be adopted. It gets a protected budget for experiments, and the failed runs are understood as the cost of learning. It gets a memory that keeps the gains, in the weights and the next version. And it gets an evaluation it can trust, a test that tells it honestly whether it got better.

Data, permission, budget, memory, evaluation. We build all five for a model before we ask it to improve.

What the operator is given

Now check the same five against a machine operator, a maintenance tech, a line lead.

Her performance data goes up the chain. The Pareto charts and downtime codes are analyzed in a conference room she is not in.

Her permission to change the work is nil. Standard work belongs to engineering. She learns early to run it the way it’s written and keep the better way to herself.

Her experiment budget is zero. Every hour she spends learning lands as labor variance, and the Monday call asks why the number is red.

Her memory isn’t kept by the enterprise at all. Her capability appears on no ledger. When she leaves, it walks out the gate with her, and the accounting never registers the loss, because it never registered the asset. At best it is folded silently into goodwill on the day someone buys the company.

Her evaluation is a review written by someone who watched her work a few hours all year.

Deny a machine those five conditions and it would not improve either. We would call that bad system design. When the same outcome shows up in people, we call it a skills gap.

Three reasons we give, none of them true

The case against extending the loop to people is rarely stated out loud. It rests on three quiet assumptions.

The first is capacity: that frontline workers can’t do this kind of thinking. Abdelhak answers that by himself. Every plant I have run had operators who knew the machine better than the drawings did: which fixture drifted in the afternoon heat, which supplier’s lot ran soft, which alarm could be cleared and which meant stop now. That is a model of the process, built from observation and revised under pressure. In Already Paid For I estimate roughly 30 percent of frontline capability is deployed and 70 percent is stranded. We are not short of intelligence on the floor. We are wasting most of what we have already paid for.

The second is flight: why invest in someone who may walk out the door? A model’s gains stay in the weights, and the weights belong to the company. A person’s gains belong to the person. True. But it gets the causation backwards. People walk out because nothing holds the capability in place: no runway, no wage that rises with what they can do, no record that it exists. We build the exit and then cite it as the reason not to invest. The Germans settled this long ago. A Meister certificate is fully portable, and German firms form tradespeople anyway, because a system that forms everyone is worth more than a firm that hoards a few.

The third is credentials: the degree as a stand-in for cognitive capacity. The worker who could run the loop often lacks the paper that would let anyone believe it, usually because she couldn’t afford the time or the tuition. So the proxy screens out the people with the most floor intelligence and the least money. And the failure runs the full height of the org chart. An engineer’s improvement loop gets written down, certified, carried to the next job, and priced. An operator running the identical loop gets none of that. Notice we don’t hold machines to this standard. Nobody asks a model where it went to school. We judge it on what it can demonstrably do, today. We extend to software the performance-based judgment we deny to people.

The pantheon

Abdelhak is not rare. I watched the same loop run in Wally Vinton. Ramon Hernandez. Jamie. Jason Blackie. And more. Their stories are told across more than 110 essays at The Long Game for All(thelonggameforall.substack.com).

Each of them did, on a factory floor, what Discovery Loop is raising money to do in a data center. And those loops went nowhere. No ledger recorded them. No wage rose to meet them. No credential certified them. No one was assigned to learn from them. The improvement lived in one head, served one line, and ended when the person moved, retired, or simply stopped offering what nobody was asking for.

Malachi Smith is the exception, and he proves the rule. GM built him a ladder, co-op to controls engineer, and the loop in him compounded. Same capacity. Different system.

Those are names from U.S. plants alone. I could write a second list from India. The capacity was present, proven, and allowed to evaporate.

The Germans built a human recursion

Consider Lukas: a fully formed German tradesperson with roughly 25,000 hours of formation by age 30, through Ausbildung, then Geselle, then Meister, now deployed at Siemens Amberg.

Read that pathway in this season’s vocabulary and it is a recursive self-improvement architecture. The apprentice gets data, supervised permission to change, a protected learning budget, a credential that keeps the gains, and a Meister who evaluates honestly. Then the recursion closes: the Meister forms the next apprentice. The loop runs across generations, not training epochs.

What Bloom does

The Bloom System is my attempt to install that loop on American and Indian floors, where the guild tradition never took root. It supplies the five conditions a machine gets by default.

A guaranteed three-year employment runway is the protected experiment budget. You cannot ask a person to take learning risks while her job sits on a quarterly contract with the labor variance line.

Continuous Meister confirmation is the honest evaluation: a standing judgment, from someone who has done the work, of what the worker can do now. It is also a credential, earned on the floor rather than bought in a classroom. It replaces the proxy with the thing the proxy was always pretending to measure.

Wage rising on capability is the reward signal. The system pays for improvement, so improvement is rational.

Capability accounts are the memory. Formation gets recorded as an asset that accumulates, not an expense that vanishes, which is the core of the accounting work Sridhar Ramamoorti and I have been building. They also answer the flight objection. Capability that is recorded, rewarded, and given a runway has reasons to stay. Capability that is invisible has none.

And the Frontline Leader closes the loop: the Meister-in-formation whose job is to turn the improved worker into the improver of the next one. Which is exactly what Abdelhak did on nights, with no system behind him.

The question we skipped

There is a vigorous public debate about whether it is safe to let machines improve themselves. Good. That debate should happen.

Nobody convened a panel on whether it is safe to keep denying that loop to the millions of people who already run it. The costs don’t make headlines. They show up as stranded capability, flat wages, brittle plants, and a workforce that has learned not to tell anyone what it knows.

If recursive self-improvement is worth $650 million in a few months for a machine that might one day do it, it is worth a three-year runway for a person who already can. That’s the long game.

Here is what I don’t know yet. I don’t know whether a capability account changes who stays. I don’t know how capability should travel when a person leaves, or whether three years is the right runway for every floor. The pilots will tell us some of that. The people who run floors will tell us more.

So I’ll ask the operators reading this: who is your Abdelhak, and what would it take to write down what he knows before he walks out the gate?

Venki Padmanabhan is a co-founder of the Capability Capital Institute and the author of the forthcoming Built to Extract and Already Paid For (Capability Capital Press). He writes at thelonggameforall.substack.com.

Written with AI assistance. The argument, the judgments, and the floor testimony are the author’s own.

Capability Capital Institute — the case for counting what working people know: capabilitycapital.org