The West’s monster takes from you. India’s monster knows things. In 1870 an Englishman collapsed the second into the first, and we have been using his word ever since.

Start in a cremation ground, at night, in a story that is at least a thousand years old.

A king has been sent to fetch something. It hangs upside down from a tree, and it is not alive and not dead. It is a vetala — a spirit that enters a corpse and animates it without letting it decay, that leaves whenever it likes, and that knows what has happened and what is going to happen.

The king takes it down and puts it on his back and starts walking.

And the thing on his back begins to talk. It tells him a story, a long one, with a complication in it — a father and a son, a queen and a princess, a man who gave up his life for another man’s wife. Then it stops and asks the king a question. Who was right. Who was the true husband. Which of them owed what to whom.

The rules are strict and they are cruel. If the king answers correctly, the vetala is released and flies back to its tree, and the whole walk begins again. If the king knows the answer and refuses to say it, his head bursts into a thousand pieces.

He knows the answer every time. So he answers, and loses it, and goes back for it. Twenty-four times.

I want to sit inside that image for a minute, because it is the most accurate picture of industrial knowledge I have ever encountered, and it was drawn nine hundred years before anyone wrote a balance sheet.

Here is where this goes. The West built a monster defined by what it takes from you. India built one defined by what it knows. In 1870 an Englishman collapsed the second into the first, and the word we inherited — the one I used on Tuesday and will keep using — carries his mistake inside it. We named a category of people after a creature that drains, when the older and truer story was about a creature you consult and cannot keep. That’s the argument.

What the creature is for

The oldest surviving version is Book Twelve of Somadeva’s Kathasaritsagara, assembled in eleventh-century Kashmir out of older material that no longer exists. There are other recensions — Kshemendra’s, Shivadasa’s, Jambhaladatta’s. The cycle passed into Hindi as the Baital Pachisi, the twenty-five tales of the Baital, and from there into more or less every Indian childhood with a television in it.

Here is the part that matters.

Sorcerers in these stories want to bind a vetala. Not to feed it, not to be fed on by it. To consult it. A being that stands outside time makes an unmatched informant, and the whole reason anyone goes to the cremation ground in the first place is that the thing in the tree knows things nobody else can know.

The Indian monster is not defined by what it takes. It is defined by what it knows, and by the fact that everybody wants it and nobody can keep it.

Read the riddle again with that in mind. The knowledge is real. The knowledge is exactly what the king needs. It is available only while the creature is on his back and the walking is hard. And the instant the knowledge is voiced, it is gone and he has to go back and carry it again.

That is not a ghost story. That is a description of what happens to what a workforce knows.

What Burton did

In 1870 Richard Burton published Vikram and the Vampire, or Tales of Hindu Devilry.

He kept eleven of the twenty-five. He rewrote them freely. And he hung the word vampire on a creature that does not drink blood.

The label stuck so hard that it is now the first thing said about the vetala in most languages, including in India. A being defined by knowing became, in English, a being defined by draining — and the substitution was performed by a man who was not from there, in a book with a subtitle about devilry, and it outlived the thing it described.

I am not interested in this as a colonial grievance. I am interested in it as a mechanism, because it is the same one I watched operate on factory floors for thirty-six years.

Take a category of being defined by what it knows. Relabel it as a category defined by what it costs. Do it in print, with authority. Wait.

The monster we imported instead

The vampire Burton reached for came from somewhere specific. Not Transylvania in any real sense — the Balkans and the Slavic world, where the word upir shows up in a Kievan document as early as 1047.

And between roughly 1725 and 1755 it stopped being folklore and became administration.

In 1725, in the Serbian village of Kisiljevo, a man named Petar Blagojević died and nine villagers followed him within days. The grave was opened. The body had not decayed as expected; there was blood at the mouth. He was staked and burned, and an Austrian official wrote it up. A year later, the Arnold Paole case went further still: the Habsburg administration sent a military surgeon, Johannes Flückinger, who conducted the examinations, wrote a report, titled it Visum et Repertum — Seen and Discovered — and filed it with the imperial court in 1732. It was printed in a Vienna newspaper and translated across Germany, France and England.

Sit with what that means. A state dispatched a qualified medical officer, opened graves, took evidence, produced a formal document titled Seen and Discovered, and lodged it with the imperial court.

For a creature that does not exist.

The bureaucracy of eighteenth-century Europe was fully capable of generating an official record of a thing that was not there. What the bureaucracy could not do — then or now — is generate a record of the things that are.

Why did they see it? Because they were looking at a loss they could not explain. They did not understand decomposition. Later, in New England, the same wave ran on tuberculosis, which by the 1800s accounted for something near a quarter of all deaths in the Northeast: a family would sicken one after another, and the survivors decided the first to die was draining the rest.

A loss nobody can account for, and a body assigned the blame.

You have seen this. A plant fails and the workforce is the explanation. Too expensive, too rigid, too slow. Nobody understood decomposition then. Nobody understands capability accounting now.

The arithmetic of feeding

There is a footnote to this that I find genuinely funny.

In 2006 two physicists ran the numbers on the Western vampire. Assume the first one appears on the first of January, 1600, when there are something over five hundred and thirty-six million people alive. Assume it feeds once a month and each victim turns. The vampire population doubles monthly; the human population falls to meet it. Humanity is finished by June of 1602. Two and a half years. Their conclusion was that vampires cannot exist, because their existence contradicts the existence of human beings.

Unlimited extraction is self-terminating. It kills the host faster than the host can replace itself, and then it starves.

A later critic objected that the model was unfair — that it treated vampires as purely greedy consumers and ignored what he called a rational strategy of managing their human resources.

He meant it as a joke about population dynamics. I have been in the meetings. It is not a joke.

The only reason extraction from working people has run for forty years without collapsing is that it is slow. Nobody drains a plant in a month. You draw down formation a quarter at a time, book it as expense, and the host keeps regenerating just fast enough that nothing ever reads as a crisis. The Balkan vampire got caught because he was greedy. Ours is patient, and patience is why there has never been a Visum et Repertum for what happened to the American shop floor.

The man at the edge of the village

There is one more thing about the vetala that I cannot stop thinking about.

Along the Konkan coast, the same name belongs to a guardian. Vetoba, who patrols the village boundary at night. The villages treat him as both — a danger to be kept outside the settlement, and a protector to be kept fed at its edge.

Kept at the boundary. Fed. Depended on absolutely. Never brought inside.

At Lansing Delta Township we installed Yazaki harnesses. The main body harness is the nervous system of the car and heavy enough that you feel it going in — a modern vehicle carries close to forty separate harnesses, seven hundred connectors, three thousand wires, two and a half miles of wire weighing something near a hundred and thirty pounds. Then a run of smaller ones down the length of trim.

The supplier sent a representative. I walked out to meet him and it was a man who had been a repair man at Lansing Grand River twenty years earlier, on my floor, in my other life. He had retired. He came back because he could not stay off a floor.

For the next stretch, every harness quality problem we had, we solved with him. He brought decades of repair knowledge and something I have no better word for than floor EQ — the read on a room, on a job, on what a person is actually telling you when they say the connector seats wrong. Yazaki respected him. I trusted him. Those are two different things and he had both.

And he was never once ours. Different badge, different payroll, another company’s name on the paperwork when the quality issue closed out. He stood at the boundary of my plant for years and fixed things inside it. I used to wonder whether Yazaki knew what he was worth to them — whether anyone had ever put a number on a retired repair man who came back for the love of it and made a multi-billion-dollar assembly plant run cleaner.

I want to give you his name and I cannot. I have turned it over for days. I can see him coming up the aisle. I can hear how he talked about a connector that seated wrong. I remember what it felt like to see him again after twenty years and to know the problem was going to get solved.

The name is gone.

That is the essay, and I did not plan it. He was on another company’s payroll, so he is in no record of mine. He was retired, so he is in none of theirs. The last ledger he existed in was my memory — which is not a ledger, and it has failed.

The riddle nobody has answered

So there are two monsters, and we chose one.

The European one is about being taken from. It is a story told by people who could see a loss and needed a body to pin it on. The Indian one is about knowledge you can consult but cannot keep — available only while you are carrying it, gone the moment it is spoken, and the carrying is the whole job.

We built our entire vocabulary of work out of the first. Cost. Burden. Headcount. Drain. Every one of those words assumes the worker is the thing taking something from you.

And the second story — the accurate one — got eleven of its twenty-five tales translated by a man in a hurry, under a title about devilry, and the rest was left in the tree.

What the Bloom system is, underneath all the instrumentation, is an attempt to answer the vetala’s riddle honestly: to write down what the creature on your back actually knows, before the answer escapes and you have to walk back to the cremation ground for it again. The accounting is the modern Burton. It takes a category defined by knowing and files it under costing, and it does so with authority, in print, in a document that goes to the board.

I do not have the full answer. I have four unsolved questions about certification and custody and who holds the record, and I will not pretend otherwise.

But I will leave you the way the Baital leaves the king.

A man on your floor knows why the line goes down. He has known for eleven years. Nobody has asked him, and he has stopped offering. The knowledge is not lost — it is in the building, on the clock, drawing a wage.

Whose asset is it?

Answer that one honestly and the thing flies back to the tree, and you have to go and get it again. Stay silent knowing the answer, and something bursts.

It is never the king’s head.

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

The best thing I ever watched a man do, I did not watch.

His name was Dale Ketterman and he was the pipefitter on my glass leg of trim at Lansing Delta Township.

The glass leg was not a line. It was a set of robot cells — front glass, backlite, left and right quarter glass, moonroof — and each cell had its own robots, its own urethane feed, and its own bead pattern. The bead was not fixed, either. It changed with the build option of the glass going in, so the pattern the robot laid down had to match the particular piece arriving on that particular body.

Dale fed all of them.

Urethane is the adhesive that bonds the glass into the body. It is not trim. In a rollover, a properly bonded windshield is part of what keeps the roof off your head. If the heaters drift, or a feed packs off, or the bead lays down wrong for the option that showed up, you are not building a car with a cosmetic defect. You are building a car that lies to the person who buys it.

Dale was in the plant before the shift. He purged the lines. He checked the heat, cell by cell. He watched pressures that had not moved in a year and watched them anyway. Then the cells ran, and the glass went in, and the bodies moved down, and nothing happened.

Nothing happened for years.

I ran that leg, and I could not tell you the last time I thought about Dale Ketterman before I sat down to write this. He was so deliberate, so disciplined, and so fast on the rare occasion something did go down that we simply assumed him. He became a property of the building, like the air handling.

Here is what took me two decades to understand, and it is the argument of everything I write now.

Dale was invisible because he was good. Not in spite of it. The only thing that makes work like his legible to the people above it is the failure of that work, and he did not fail. Excellence in frontline work erases the worker, because the ledger has no way to record the thing that did not go wrong. That’s the argument.

The mirror

On Tuesday I argued for a word. Frontlines — not frontline workers, but Frontlines, the way you would name a category of being, the people who stand where the enterprise touches reality.

Here is the part I left out, because it needed its own essay.

The word works the way vampires works. Not as an insult. As a name for a class of creature the daylight world has agreed not to look at directly, and has told a story about instead.

The parallel holds further than it has any right to. A vampire casts no reflection: stand him at a mirror and the glass reports an empty room. Stand Dale Ketterman in front of a balance sheet and it reports the same thing. The presses have a value. The conveyors have a value. Every robot on his glass leg is a capital asset with a depreciation schedule and a residual. The man who kept the adhesive flowing into all of them is an operating expense, which is the accounting word for something you consume and do not own and cannot lose.

And nobody believes in vampires until there is a body. Nobody sees the pipefitter until there is downtime.

That is the whole mechanism. Frontlines enter the record only as exceptions. The ledger logs the four hours the line was down, never the eleven thousand hours it wasn’t. The grievance is filed; the shift that went clean is not. Every instrument we point at these people is calibrated to detect failure, which means the better they are, the less of them exists.

The days it did let go

There were days.

The bead goes down, the robot picks the glass, and somewhere between the cell and the body the urethane lets go and the glass slides off the edge. Now the line is stopped and people are cutting glass back out of cars that already have it in them, which is as bad as it sounds.

And nobody knows whose problem it is. Is it the urethane — the heat, the cure, the lot that came in this week? Is it the glass — the frit, the primer, the supplier? Is it the robot — the path, the pick, the tip? Is it the operator?

Four suspects. One of them is Dale’s.

Here is what he did. He appeared. Nobody sent for him. He came into the middle of it quietly and worked the problem as though his own feed were as likely a culprit as anything else in the room. It took me years to understand how rare that is. The ordinary response to finding your system on a suspect list is to prove it off the list. Dale walked toward it.

Then it would get solved, and the line would come up, and the cars would move, and I could not have told you afterward which part of the answer had been his.

That is what the vampire story gets right and gets backward at the same time. He was not invisible only in the quiet years. He was invisible in the crisis too — standing in the middle of a stopped line, in front of everyone, still unseen, because he never once made the solving about the solver.

The other half of the story

Then there is what the daylight world says about them.

They suck your blood. That is the accusation, and you do not have to look hard for it. Labor is the line item that gets attacked first in every downturn. Headcount. Burden. The cost that has to come out. When a plant fails, the story is almost never that capital misjudged the market; it is that the workforce was too expensive, too rigid, too slow to change. The parasite framing is not subtext. It is in the transcripts.

So look at the actual direction of flow.

Buick City ran a just-in-time system the Japanese flew in to study. It closed in 1999 anyway. When the accountants totaled the assets, the presses had a value and a buyer, the tooling had a value and a buyer, and the most valuable thing in that plant had no line on the sheet, could not be sold, could not be transferred, and therefore could not be argued for in the room where the decision was made.

Thirty years of formation walked out of the gate and was booked at zero. Not written down — never written up. The company had spent decades drawing capability out of those people and expensing every dollar of it as it went, so that on the day it mattered the asset it had built did not exist.

Somebody was feeding on somebody. It was not the pipefitters.

What the life is actually like

I want to be careful here, because the romance about factory work is nearly as useless as the contempt.

The life is early. It is loud in ways you stop hearing and your ears remember anyway. It is physical in a manner that arrives in your forties as a bill. The line sets the pace, the pace is not a suggestion, and eleven seconds behind is a real event with consequences.

It is also — and this never survives the trip to the boardroom — intellectually serious. Dale held a model in his head of every feed on that leg — how each one failed, how each behaved as the ambient temperature moved through a shift, the particular temperament of each robot, and which bead pattern belonged to which option code on which piece of glass. That model was never written down. It could not have been. He built it the only way it can be built, which is by standing next to the thing for years and paying attention on the days nothing was wrong.

And the precarity is a different species from the white-collar kind. When a plant closes the engineer carries his credential to another company. The Frontline’s competence was plant-specific, unrecorded and uncertified, so it does not travel. He does not lose a job. He loses the only ledger that ever held him — and it was not a ledger, it was the memory of a supervisor who is also now out of work.

Two countries, one result

I have run floors in both, and the difference is real.

America hides its Frontlines architecturally. There is a service entrance and there is a main entrance. There is a third shift and there is a business day. The loading dock faces away from the road. The hotel room is cleaned by nobody, the trash leaves by itself, the package arrives at 5 a.m., and the app has managed the remarkable trick of converting a man in a car into a notification on a screen. The invisibility is engineered into the plan of the building and the schedule of the day.

India hides nobody. The Frontline is three feet from you, in your kitchen, in your lobby, on your stairwell, present continuously and in numbers. The separation is not spatial — it is social, and it is achieved by looking past a person who is standing directly in front of you. In America you are not given the chance to see them. In India you are given the chance constantly and have been trained to decline it.

There is a further layer in India I will name plainly: much of this work is inherited rather than chosen, so the invisibility attaches to the person before the person has done anything at all.

But both systems arrive at the same place. Neither country’s accounts have a line for Dale Ketterman. One hides him behind a wall and the other hides him in plain sight, and both of them book his formation as an expense and write it off in the year it occurs.

Why the rest of my working life

In 1999 a man named Simon Boag walked me across Trim 1 at Buick City, stopped in front of a gray-bearded supervisor named Rick Stork, and told him to form me.

Nineteen ninety-nine. The year the place closed.

Rick did not start with takt time.

He started with sāma, dāna, bheda, daṇḍa. Persuasion, the gift, division, the rod. The four upāyas — the old grammar for how one person moves another. What the instruments are, what order they come in, and how far down the list you have to travel before you have already failed. Persuasion first, because it costs nothing and breaks nothing. Then the gift. Then division — and by the time you are dividing people, you have conceded the thing cannot be done whole. The rod is on the list. The rod is last. Every rung down is an admission.

A supervisor in a Flint assembly plant taught a young engineer Sanskrit statecraft on the trim line, and what he was actually teaching was that I would be answerable for the cars and for the people who built them at the same moment, and would never be permitted to trade one against the other.

That is not what anyone thinks frontline supervision is. That is the thing the mirror will not show you.

And he did it in the last year, with the gate already in sight. Nobody assigned it to him. There was no future in it for him and he knew there wasn’t. He handed a lineage to a young engineer who happened to be standing there, and then the plant closed and he went wherever the rest of them went.

I have the years I have left, and I am spending them on this. Because I have watched what happens when the asset nobody recorded walks out the gate, and I have never once seen anybody stop it — not out of cruelty, but because you cannot defend in a meeting a thing that does not appear in the documents the meeting is held about.

Where I am stuck

I do not know how to fix this cleanly, and I would rather say so than pretend.

I know the accounting is wrong — that formation is capital expenditure booked as operating expense, and that this one misclassification has done more damage to working people than any robot yet built. I am confident about the wound. Much less so about the cure. Who certifies? Who holds the record? What stops a capability account from becoming one more instrument for scoring people? Four years on those questions and I have partial answers at best.

What I am no longer willing to do is let the category go unnamed. You cannot argue for people who are not in the document, and you cannot put people in the document until you have a word for who they are.

So: Frontlines. Take the word, or tell me a better one.

And if you have a Dale — the one whose competence made him furniture, the person you assumed for years without ever once thanking — tell me his name. I am building a record, and it is going to be made out of names or it is not going to be made at all.

That is the long game.

The pipefitter’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

The camera was wrong, and it didn’t know it. At Lansing Delta Township, a vision system read the option code on each car so the right rear glass went in. One shift, it misread. Nothing on any screen turned red. A quality checker at the end of the line stopped at a car wearing a rear glass with an antenna the build didn’t call for.

What happened next is the part nobody wrote down. I’ll get to it.

This week, a Microsoft director of applied science surfaced in the unsealed record of the New York Times’ lawsuit against OpenAI and Microsoft. In a January 2023 memo, he called the data behind his company’s models possibly the “largest theft of labor in human history.”

He wrote labor. Not content, not data, not intellectual property, the words the lawyers will fight over. An insider looked at what went into the machine and named the people.

I have spent thirty-six years walking production lines, from Buick City to Stuttgart to Bengaluru. When someone tells me a new kind of factory has arrived, I walk it from raw material to finished good. At every station I ask two questions. Who formed the people who work here? And where does their formation go when the shift ends?

So let’s walk the AI line. It runs from wafer to token.

AI is a factory, and like every factory it runs on formed people. At three of its five stations, it takes the formation of people it paid little or nothing, crystallizes it into capital, and sells the product back against them. A plant that did that would be eating its own maintenance crew. That’s the argument.

Station one: the plant

The first station is the one a plant manager recognizes. It starts in mines and power plants, runs through chip design, and turns physical at the fab: clean rooms, process windows, yield curves, tools worth hundreds of millions held in tolerance by maintenance crews. The chips then go into data centers built by electricians, pipefitters and controls technicians, who hand the building to a thin operating crew once the lights are on.

The people here are deeply formed. A good fab technician is built over years, the way Lukas was built at Siemens Amberg, roughly 25,000 hours by age thirty. Her judgment keeps yield from collapsing on the night a chamber drifts. The electrician who wired the data center carries what she learned to the next site.

At this station the helix holds. Capital is crystallized labor; labor is capital in formation. The people who build and run it get better every year and take that capability home. Both strands compound, and a sane operator pays for both.

If AI were only this station, I would not be writing this.

Station two: training the model

Training a model is one enormous crystallization event. The raw material is formed human judgment: every well-reported article, every patient explanation, every line of code someone debugged at 2 a.m. and posted so the next person wouldn’t have to. The court filings describe one scraped dataset holding more than two million documents from a single newspaper’s site. Multiply that across every writer, teacher, coder and forum volunteer on the open web.

In a plant, crystallization is ordinary and honest. When my best operator finds a better sequence, we write it into standard work. That is capital formation. But three things hold on a plant floor. She was paid while she formed. She keeps the capability after the process absorbs it. And the process does not go to market against her.

At station two, all three break. The formers were never paid for the formation taken. The model now holds their capability in crystallized form. And the product sells into the same markets they work in. This is where one strand is harvested and cut.

Station three: the rater’s desk

A model fresh from training is raw stock. Before it ships, people finish it. They rate its answers, correct its mistakes, and write the examples it learns manners from.

This station runs on two tiers. At the bottom, annotation and content moderation flow to outsourcing centers in Africa and across the Global South, at a few dollars an hour. At the top, doctors, lawyers, engineers and coders are recruited to judge the model in their own field. One platform that matches these experts to AI labs reports paying them about $95 an hour on average.

This is the only station on the line where formation carries a price tag, and the price tells you how the buyer sees it. A physician’s twenty years is bought one rating at a time, like a gig. Nothing is capitalized in her name. The more carefully she works, the faster she trains the thing that will answer her patients without her.

Station four: the forward-deployed engineer

The fourth station is the one almost nobody talks about, and the one that should worry a plant manager most.

A model knows little about your operation. So the AI companies send forward-deployed engineers to sit with the customer’s operators, planners, schedulers and claims adjusters. They watch how the work really gets done: the workaround, the judgment call, the thing the veteran knows that no procedure holds. Then they encode it.

Any Toyota-trained reader knows the method: go to the gemba and learn from the person doing the work. The difference is where the learning lands. In a real TPS plant, the operator’s insight improves the process and raises her standing; she becomes the one who teaches the next person. At station four, the insight leaves the building, goes into a system the customer rents back, and makes the operator who supplied it less necessary.

Station two harvested millions of anonymous people at once. Station four does it firm by firm, floor by floor, with a friendly engineer and a laptop.

Station five: the token

The last station makes the finished good: the token, sold by the million at a price that keeps falling.

What is unusual is who this finished good competes with. An OpenAI product leader wrote internally that these products would become more substitutive as they improved. Microsoft’s own data, cited in the filing, showed its answer engine cutting click-throughs to the newspaper’s site by as much as 93 percent.

That is a factory whose output replaces the people who supplied its raw material. On any line I have run, we would have flagged that as a supply risk long before anyone called it a strategy.

And the line loops. Every conversation, every thumbs-up, can become raw material for the next model. The finished good rides a return conveyor back to the start, and nobody on that conveyor is forming anything.

The doom loop is a maintenance problem

The same filings describe internal worry about a “doom loop.” Here the metaphor becomes a prediction.

At Lansing Delta Township, a robot laid the urethane bead on the glass, and two people on a deck set it with suction cups and a holding winch. Abdelhak Radjai was a team leader on that line, and on plenty of shifts he helped me run it. He would get on any job on his line to give someone a bathroom break, so he knew every station by hand.

Back to that car at the end of the line. A formed person had caught one. Abdelhak didn’t treat it as one bad car. He walked the line back, car by car, against the flow, and found every wrong back glass, four of them. Then his team went to work. They had to stop the line, more than once. They cut the wrong glass out and cleaned the old urethane off the body. They ran new glass through the bead robot and walked each one back down the line. With no deck and no winch, two people set each glass by hand at line height, careful that no urethane touched the headliner or the inner quarter trim. It was an orchestra: every person knowing their part, nobody waiting to be told. They kept downtime to a minimum, and nobody got hurt.

That camera was crystallized labor, somebody’s knowledge of which glass goes on which car, turned into code. Behind it stood formed people: one who could see a wrong antenna at a glance, one who knew a wrong part at the end means a wrong process upstream, and a crew who could rebuild by hand without hurting anyone. Take them away, and the automation fails in the exact moment it was built for.

Abdelhak went on to run Trim on nights, and he ran it the same way, by forming the people around him. Nobody booked that either.

AI carries the same dependency at scale. Researchers writing in Nature in 2024 showed that models trained repeatedly on machine-generated output degrade, a failure they call model collapse. A system that substitutes for the journalist, the coder and the explainer is eating the formation it depends on. The labs already know this. It is why they keep buying time at the rater’s desk. That is the doom loop: a plant that stopped forming its Abdelhaks and hasn’t yet had the shift that shows it.

What this predicts for labor

Walk the line and the prediction is sharper than “AI will take jobs.”

At the physical ends of the chain, the fab, the trades and the maintenance crews, formed labor gains value. There are not enough formed people to build, run and repair it. In the knowledge-work middle, where formation can be captured at stations two, three or four, labor is crystallized and substituted. The people who supplied the capability lose the asset they built.

That is the chasm. It widens because one strand of the helix compounds while the other is harvested and never re-formed. Redistribution after the fact treats the symptom.

Under oath, Microsoft’s chief executive testified that paywalled content should be licensed for training. A license is a deposit: something flowing back to the people whose formation became capital. The principle is conceded. The ledger is not built.

The court will decide whether station two was fair use. That question isn’t mine. Mine is the one I would ask on any line: where did the formation come from, where did it go, and who is forming the next shift?

I don’t know what the deposit should look like. A license fee is one answer. A capability account that follows the person is another. Someone standing closer to station two may see a better one than I can. What I do know is that every line I have run needed an Abdelhak forming the next shift, and I haven’t found one on this line. That’s the long game.

If you work at one of these stations, tell me: who is forming the next shift where you are?

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

Evidence They Can’t Defend — Essay 12 of 13

I ran a plant in Ohio, in a manufacturing town, competing for labour against employers who had been there for generations and could pay more than we could. What we could offer was close to the local minimum. Turnover ran about 50 percent a year, and a meaningful share of the floor was staffed by temporary workers.

Two of those temps were living in the local United Way shelter.

They came to me for work and I took them gladly. What happened next was the most satisfying thing I have watched in years. They learned. They got good. And then the ordinary furniture of a life began arriving: an apartment. New clothes. A set of wheels. I have run plants on three continents and held titles that sounded important, and watching a man walk in wearing a jacket he bought himself is a different order of thing. It felt close to missionary.

Then I caught one of them stealing from the self-dispense food pantry on the floor. Not subtly. Blatantly.

And the other — the day I gave him permanent employment, with the certainty and benefits that come with it, was the day he parlayed that status into a job somewhere else for a few dollars more.

Missionary without satisfaction.

I have turned that over for a long time, and the easy reading — that two poor men let me down — is wrong.

Here is what it proves. I could form those men. I could not hold them. The one thing genuinely mine to give — the teaching, the standard, the chance to become capable — I gave, and it worked, and it was worth about two dollars an hour less than the thing across town. The formation was real. The wage was the binding constraint, and no plant sets the wage in a vacuum.

Underneath it sits a prior question I could not stop asking. Why were those men temps at all? Why does a person arrive at thirty with no credential, a thin or damaged early record, sometimes a petty conviction, and a settled expectation of living paycheck to paycheck — and why is the category built to receive him called temporary?

I went looking at the research. The answer is worse than I expected.

They are not temporary

More than a third of American temp workers — 35 percent — report their current assignment has already lasted over a year. Nearly one in five say over two.

Read that again. For a large share of the people in it, temping is not a bridge to permanent work. It is the destination. The category exists so that a permanent worker can be classified as something else.

The consequences are not incidental. In production and manufacturing, temps earn roughly 72 to 80 percent of what directly hired workers earn for the same work. A quarter report wage theft. Seventy-one percent report retaliation for raising an issue with a supervisor.

And the safety figures should stop anyone who runs a floor. A study of 1.3 million workers’ compensation claims in Ohio found temp agency workers injured at 11.6 per 100 workers against 4.9 for permanent workers— more than double. Injured temps were about five years younger and had been on site a shorter time before it happened. Research across five other states found injury incidence 36 to 72 percent higher, with temps 68 percent more likely to be working in the most dangerous fifth of all occupations.

That study is Ohio, and Ohio is where I ran temps. The number is about my floor too, and I would rather say so than write an essay pointing at somebody else’s.

Now notice what those properties do to the public balance sheet. A temp is a worker whose employer contributes less to unemployment insurance, whose compensation coverage is contested between agency and host, who receives no health benefit, and who is injured at twice the rate. Every gap there is filled by somebody: the emergency room, Medicaid, the shelter — the same shelter that housed two of my workers before I hired them.

That’s the argument. Extraction does not only move value from workers to shareholders. It moves costs from corporations to communities. And the communities absorbing them are the same ones that educated the workers, built the roads, and kept the lights on in the shelter where my two men were sleeping.

The halving

In 1952, corporate income taxes accounted for roughly 33 percent of federal revenue. By 2023, about 10 percent.

The statutory federal rate fell from 46 percent in the early 1980s to 21 percent after the 2017 Tax Cuts and Jobs Act, and the effective rate — after deductions, credits, deferrals, offshore profit-shifting, accelerated depreciation — runs lower still. Many of the largest corporations pay single digits. Some pay nothing.

The 2017 cut was sold as a stimulus that would produce investment and wage growth. What followed: S&P 500 buybacks passed $800 billion for the first time the next year. The savings bought stock, which lifted earnings per share, which lifted executive pay.

It did not trickle down. It was extracted up.

Who pays instead

The revenue gap does not vanish. It is filled by other taxpayers or it shows up as deteriorated public services, and both have happened.

Individual income and payroll taxes now supply roughly 85 percent of federal revenue, up from about 60 percent in the 1950s. The burden moved from corporations to workers — the same workers whose wages stagnated and whose pensions were eliminated.

At state and local level it appears as crumbling infrastructure, underfunded schools, reduced public health. The American Society of Civil Engineers grades American infrastructure at C-minus or below, with over $2 trillion needed to reach adequacy — a gap that exists partly because the corporations whose supply chains depend on that infrastructure have systematically reduced what they pay toward it.

The externalization machine

When Walmart pays wages low enough that full-time employees qualify for food stamps and Medicaid, the public subsidises its labour costs — an estimated $6.2 billion a year, a taxpayer transfer to one of the most profitable companies on earth.

When warehouse injury rates run above industry averages, the cost of care and lost productivity flows to workers, families and public health systems rather than the company that produced the conditions. When construction firms misclassify workers, the unpaid payroll taxes and compensation premiums create shortfalls in the exact systems those workers will need. When companies eliminate pensions, the retirement shortfall becomes a public problem.

Every extraction mechanism in this series has a matching externalization. The cost the corporation sheds lands on the commons.

The ratio

Oxfam’s analysis of America’s five largest corporations found that their combined buybacks and dividends over a five-year period exceeded $1 trillion — roughly five times their total federal tax payments across the same years.

For every dollar those companies paid in federal tax, they sent five to shareholders.

These are firms whose operations depend on publicly funded roads, airports, water systems and an educated workforce. They use the infrastructure, recruit from the public schools, benefit from the stability — and contribute a fraction of what they draw. That is extraction at the level of the commons: not from an individual worker, but from the shared resources that make private enterprise possible.

Another model exists

The Nordic economies maintain high corporate taxes, comprehensive social services and strong worker protections — alongside some of the highest levels of competitiveness in the world, and companies including IKEA, Volvo, Novo Nordisk and Ericsson.

High taxes and strong protections do not destroy competitiveness. In several cases they strengthen it, because a firm that can rely on publicly funded healthcare, education and infrastructure puts its capital into innovation instead of duplicating basic social functions.

The American model of low corporate tax, minimal public investment, and externalized cost is not the only model. It is not even the most competitive. It is the most extractive.

What I do not know

Here is what I cannot resolve, standing on my own floor.

Would I hire from the shelter again? Yes. In the hope that some would stay — and because the one who left had been formed, and took that formation with him to a floor that will benefit from it. Maybe someone else’s formed man walks through the door next year. That is not a system. It is a hope, and I know the difference.

What I do not know is how to hold anyone at the wage such a plant can pay. It did not set the local labour market and could not beat it. Every solution I believe in — the guaranteed runway, the Meister, the account the worker owns — assumes a firm that can make a multi-year commitment. A business paying local minimum against established competitors can form people and cannot keep them, and I have no answer that survives that payroll.

And I do not know how to reach the gates before mine. Those men arrived with no credential and a damaged early record, and by then the failures were decades old — a school, a home, a first employer, a court. Everything I know how to build starts at the fourth gate. The public systems that could have caught them at the first three are the ones being defunded by the extraction this essay describes. I can name the loop. I cannot break it from where I stand.

So here is a real request rather than a rhetorical one. Find out how many people in your building are classified as temporary, and how long they have been temporary. Then find out what your company contributes for them, and what the public contributes instead.

Then tell me whether the arithmetic in your building looks any different from mine.

Next: The Fork in the Road — AI can accelerate extraction or correct it. Same technology. Opposite outcomes. Everything depends on who decides.

Sources: Congressional Budget Office, historical corporate tax revenue; IRS Statistics of Income; Oxfam corporate tax analysis; Americans for Tax Fairness, Walmart public assistance analysis; American Society of Civil Engineers Infrastructure Report Card; Tax Policy Center, corporate effective rates; National Employment Law Project, temp worker survey and “Lasting Solutions for America’s Temporary Workers”; NIOSH / American Journal of Industrial Medicine, Ohio Bureau of Workers’ Compensation claims analysis 2001–2013; ProPublica, temporary worker injury rate analysis; Illinois temporary staffing wage comparison, PMC.

The plant account is the author’s own.

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

Jamie was a Marine sharpshooter before he was anything on our ledger. The Corps had spent years forming him — not merely training him but forming him, the slow conversion of a person into judgment that holds under load — so that he could be the one they sent first. The tip of the spear. The man you put at the point of contact precisely because he is the one you have invested in most, and would never spend without cause.

Then he came home and walked onto the trim line. Team Leader, Trim 2, seventy-five thousand dollars a year, standing watch over a plant worth fifteen billion. And the first thing our accounting did, the morning he badged in, was file what the Corps had built under a single word.

Headcount.

That distance — between what the country formed and what the firm recorded — is the whole subject of this essay. We didn’t lower Jamie with a policy. We lowered him with a noun.

Words come first

A word carries a buried instruction, written long before any of us arrived, and it runs whether or not we read it. We named the people who do the world’s actual work, and the names quietly did the rest for a century. So look at what we named them.

Employee comes from the Latin implicare: to enfold, to entangle, to fold a thing into something larger than itself — the root that also gives us implicate. An employee, at the level of the word, is one folded into another’s purpose. And the grammar finishes the job. That small suffix — the one in payee, detainee, addressee — marks the person to whom things are done. You are the object of the verb.

Labor is older and harsher. In Latin it never meant the fruit of effort but the pain of it — toil, distress, suffering. We keep the original sense in the “labor” of childbirth, and the French cousin travail descends from trepalium, an instrument of torture. To call a person’s contribution “labor” is to define it at the root as suffering to be purchased, and therefore minimized. The word points at the ache, not the growth.

Workforce sounds neutral until you hear the second half. “Force” turns work into a mass — a quantity to be sized, scaled, and cut. You do not form a workforce; you set its level. And human resource files the person beside ore and timber: a resource is a supply drawn down until it is gone. The academy’s upgrade, human capital, is more honest — it admits that forming a person creates value — but it buys the honesty by moving her onto the other side’s balance sheet, where she sits depreciating.

Then there are the tells we no longer hear: hands, which cuts the working hand away from the head that guides it; headcount, the person as a number you can reduce. And the euphemisms — associate, partner, team member — which dress the relation in the language of alliance. That costume is itself the tell. You only disguise a relation you would rather not name plainly.

What the words are for

These nouns are not descriptions that drifted toward unkindness. They do work. They strip the worker of agency, fixing her as the one acted upon, and they license extraction by filing the human under cost — a thing to draw down, never the growth that compounds. It is the vocabulary of a mine. We own almost no common business word from the vocabulary of a garden.

But the deepest move is the quietest. These words make the relationship between capital and labor disappear as a relationship. “Employee” presumes “employer” and says nothing about the distance between them. The vocabulary names one pole — the human as input — and erases the strand doing the drawing-down. That is how domination gets to sound like description. Capital wins before the argument starts, because it owns the nouns.

The specimen

You don’t have to take this on faith. The vocabulary arrives in your feed, paid for, in a box.

This week it was an ad for a corporate learning platform: warehouse workers in hi-vis on one side, nurses in scrubs on the other. The headline: Get your workforce ready for what’s next. Beneath it: Connect employee learning to your talent needs.

Read it the way it asks to be read. Every sentence is addressed to the owner; the people in the photographs are the object of each one. And the telling line is the one about formation. Learning — the single thing that could lift the human strand — is bent to serve “your talent needs.” The worker is formed so the owner’s pipeline stays full. Development pointed backward, sold as a product, in a box stamped ADVERTISEMENT, and none of us blink.

The front of the front

So here is the word I use instead, and why.

Front is the Latin frons — the brow, the face, the foremost part, the place where you meet what is coming. Three things are written into it before I add anything. It names a position relative to reality — the load, the patient, the weld — rather than to an owner, so dignity moves from who signs the check to who meets the world. It carries authority, because the front is where the truth arrives first. And it puts the rear in service of the front, never the reverse.

The word cuts two ways, and the wrong reading is the dangerous one. There is front as cannon fodder, the expendable line fed into the grinder; if that were the meaning, I would be dignifying being used up. But the front I mean is the Marine — the front of the front, sent first because he is the most formed and the most trusted, never spent carelessly. Frontline as honor, not as attrition.

What the word sets in motion

The old words make the two strands enemies. If the human is cost, capital can only rise by pressing labor down — the extraction is built into the language. “Frontlines” sets them climbing together. Once the worker is the point where value is made or lost, forming her stops being a cost to shave and becomes the investment that compounds. Capital that forms its front grows. Capital that drains its front is eating its seed corn.

That is our founding wager in a single word: capital is crystallized labor, and labor is capital still forming. The front is where one becomes the other.

A nation with no name

So who are they, finally? Not the residual, not what is left after management. Define them by the only test that holds: those who stand where the enterprise touches reality. By that test there are about 360 million of them, across six sectors in two countries — manufacturing, construction, logistics and transport, hospitality, retail, healthcare.

Gather those 360 million into one country and it would rank third on Earth, behind only India and China, larger than the United States itself. The largest unformed, ungoverned nation in history — and it is hiding inside two democracies, filed under your workforce. No flag. No capital. No name.

The largest arbitrage on earth

None of this is a charity case. It is the largest pool of unpriced capability on the planet — and there is already a form of capital that makes its living pricing what the official ledgers cannot see. Venture capital looks at a team with no earnings, no assets and no track record, and underwrites the capacity to create value that does not yet exist. It books the unformed as an asset to grow rather than a cost to cut, and it is paid for that foresight in multiples. Nobody has ever called it generous. Bloom is the same trade run on the frontline — capability capital was never a metaphor.

For half a century we ran the opposite book. Taylor taught the firm to strip the worker’s judgment into a standard and treat the person as a variance to minimize. Friedman taught it that anything not landing in this quarter’s return was indulgence. Together they sold a misallocation as efficiency and left the compounding return on 360 million people lying on the table, because the framework could not see formed capability as capital at all. Bloom is not gentler than Friedman. It is a more complete economics than Friedman, and the value it frees was ours the whole time.

What we don’t yet know

Here is where I stop performing certainty, because the diagnosis is the easy part. The wound is settled: the words are extractive, and “frontline” is truer. The cure is not, and I would be lying if I sealed it shut.

We work from three principles — form people to their full capability; capability is the only profit that compounds; augment people, don’t replace them. They are a direction, not a finished constitution. The hard clauses are still unwritten, and one of them is load-bearing: I can call Jamie an asset all I like, but the accounting standards forbid me from posting him as one. His formed capability disappears into goodwill, unbooked — even though the same profession can already compute his value to the dollar the day a plant changes hands. The value is not missing; it is trapped, invisible to the ledger the way a seed round is invisible to a bank. The clause still unwritten is the one that builds the ledger to let a firm carry Jamie as capital to grow. Solve that and the trade stops being a manifesto and becomes an asset class.

So I am not handing you a doctrine. I am handing you a word and an unfinished argument, and asking you to test it where you work — to tell me where “frontline” breaks, who it leaves out, and what it would take to book a person as the asset she already is. The constitution of that nation with no name is being written now, and it is short a great many authors.

The Indian motto is Satyameva Jayate — truth alone triumphs. Not has triumphed; the verb points forward. And the truth here is a number: what a formed person is actually worth, a figure the ledger can already compute and still refuses to post. Keep naming it correctly — a Marine standing watch over fifteen billion dollars, not “headcount” — and the real figure surfaces in the end. Suppressed value, like suppressed truth, does not stay down. The front is where both show up first.

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

Dear Dario,

Opinions about AI are like a******s. Everybody has one, and since Saturday morning everybody has been showing you theirs.

Within the hour, Chamath Palihapitiya had decided the essay was a bid to kill open source and park the power at Anthropic. Elon Musk quote-posted it with three words: Dario is right. Then Sam Altman agreed the frontier needs pacing and committed to matching your embedded-evaluator step. By now there are a few thousand replies and I have read enough of them to know that every one is written from your end of the wire.

I spent thirty-six years running factory floors. General Motors. Chrysler. Mercedes in Stuttgart. Royal Enfield. Ather. Not writing about manufacturing — running it. That doesn’t make me right about AI. It puts me at the other end, where the things you build actually land, and I have not yet found a response to your essay written from here.

So: one disagreement, and one ask small enough to act on this quarter.

You asked what we would do with the extra time, and you answered four times — operational excellence, alignment, interpretability, testing and evaluation. All four answers are inside your building.

But you also proposed something in step one that is better than you seem to realise, and you aimed it too narrowly. Embedded evaluators are an apprenticeship. You have committed to giving outsiders desks, badges, laptops and employee-like access so they can learn by proximity what no document transmits. That is exactly right, and every one of those seats is currently pointed at people who audit how you build. Not one is pointed at anyone who knows what happens when the thing you built meets the work.

Give some of those seats to practitioners. That’s the argument, and the rest of this letter is why.

The airplane you reached for

When you made the case for operational excellence you needed a precedent — a complex, safety-critical system human beings run millions of times without anyone dying. You chose commercial aviation. Right choice, wrong reading.

Aviation is not safe because of engineering. The engineering was largely solved decades ago. What came after was forty years of building the human system around the hardware: type ratings certifying a pilot on one airframe and no other. Recurrent checkrides forever, no matter how senior. The A&P license that verifies a mechanic’s hands, not just his knowledge. The checklist, which exists because the industry decided memory is not a safety system. And crew resource management, invented after a run of crashes where the metal worked fine and the cockpit didn’t, because the junior officer saw it coming and didn’t say it loudly enough.

You reached for the best formation story in industrial history and used it as an engineering story.

What proximity is for

You already know verification requires proximity. A report cannot carry what presence carries; a second opinion is worth most from someone who was in the room. Your entire first step rests on that. You reached for banking supervisors as the precedent — fine, but there is an older one, and it is the one the trade I came up in still runs on. You cannot certify what you have not stood next to. The guilds settled that in the twelfth century. The journeyman did not read about the Meister’s judgment. He stood at the bench, absorbed it over years, was tested on it, and walked out with a paper that traveled.

You called embedded evaluators radical. In the trades it was Tuesday.

Here is why it matters that the seats face inward. Your evaluators will be able to verify that you followed your own safety practices. They will not be able to tell you that the model is confidently, plausibly, fluently wrong about something a person doing that job for eleven years would catch in four seconds. That is a different failure and it does not live in your building.

What it looks like from here

A line goes down at Lansing Delta Township. Fifty-five jobs an hour — Traverse, Acadia, Enclave — one vehicle off the end every sixty-five seconds. The glass leg is not one line but a set of robot cells: front glass, backlite, left and right quarter glass, moonroof. Each cell has its own robots, its own urethane feed, its own bead pattern, and the bead changes with the build option of the glass going in.

One man fed all of them. A pipefitter. He was in the plant before the shift, purging lines and checking heat cell by cell, and for years nothing happened.

Now put your model on his cart.

It will read the fault log faster than he can and return forty plausible causes in nine seconds, ranked, each with a confident explanation. Whether two hundred and fifty people stand for four more minutes or four more hours comes down entirely to whether the person holding the laptop knows which three of the forty are worth checking first — and knows, when the model is wrong, that it is wrong.

Put a number on it. At fifty-five an hour a minute of downtime is nine-tenths of a vehicle. At five thousand dollars of gross margin a unit, conservative for that mix, the line bleeds near five thousand a minute. A quarter of a million an hour. The distance between four minutes and four hours is about a million dollars. One fault, one line, one shift.

That is the scale of a single judgment made by one person at one machine on one afternoon, and there is no instrument anywhere in your plan that can see it.

The risk that got a hyperlink

Your third paragraph lists the risks: losing control of the systems, misuse for cyberattacks and bioterrorism, serious economic disruption.

Two of those get a plan. The third gets a link, and never returns — not in the four uses of bought time, not in the three steps, not in the bottom line. The closest the essay comes is a line about society having a say and pacing making room for public deliberation. Deliberation is people talking about what is being done to them. Formation is people equipped to do something about it.

Formation is a safety instrument, not a training budget

There is an obvious objection here, and it is the right one: that forming operators is a productivity argument in a safety argument’s coat. Better-trained people get more out of the tool. Fine — but that is efficiency, and you are writing about catastrophe.

So take the other half. Your threat intelligence report, published days before the essay, catalogs what people did with Claude — fraud at scale, surveillance, espionage, thousands of fake profiles, biological misuse. In every case the model was not the actor. A person was.

And your entire answer is model-side. Refusals, classifiers, detection, bans, an evaluator verifying you built them well. That assumes a hundred percent of the discipline must live in the tool, because none can be expected of the person holding it.

No industrial safety regime has ever been built that way.

Ask what actually keeps an electrician alive and it is not the wire and it is not the license. It is lockout-tagout, and it is a dance performed by human beings. You isolate the energy. You put your own lock on the hasp and hang your own tag with your own name on it. Then you try to start the machine, to prove to yourself that it is dead. Every person working on that equipment hangs a lock, and nobody on earth can cut yours off but you.

There is no engineering in that. No artifact enforces it. It is a discipline carried out, every single time, by a formed person who understands why each step is there — and it is the most effective safety practice on any floor in the world.

The same holds further out. Nothing about a forklift prevents the obvious thing a forklift can do; we certify the driver. Hazmat moves down the interstate because a man passed an endorsement, not because the trailer refuses to carry it. Every one of those is a system in which the human is a load-bearing safety component rather than the threat model.

Yours is the first that treats the operator purely as the adversary. And the guilds never only certified what a man could do — they certified what he was permitted to do, and they could take it back. Licensure is two instruments in one: capability and constraint. You are trying to build the constraint entirely into the artifact, and it will not hold, because the artifact will eventually be open-weight, or jailbroken, or built by people who never agreed to your constraints.

Formation is the only constraint that travels with the person instead of the file.

The ask

You have already committed, unilaterally, to putting outsiders inside your building with employee-like access. I am asking you to amend that commitment, not replace it.

Reserve one in four of those embedded seats for practitioners from the sectors your models are being deployed into. Not AI safety researchers. A controls engineer. A charge nurse. A dispatcher. A site superintendent.

Same desks, same badges, same laptops, same permissions, same contract — including the right to publish what they found and what they were not shown. Rotating terms, ninety days, because they have to still be doing the job to be worth anything. And one deliverable: where the model is confidently wrong about their work, and what a person would need to know to catch it.

That is the whole ask. It needs no antitrust waiver, no agreement with Beijing, no new institution, and no budget line you have not already approved.

The second ask is larger and I will make it anyway. One percent of what you spend on training compute, directed at building and certifying the capability to operate what you ship. Not user education, not prompt tips — certification: a defined body of competence, independently assessed, recorded, portable, owned by the person who earned it rather than the employer who paid for it, and revocable. The same architecture you built for evaluators, pointed at the operator.

The first ask is how you would find out whether the second one is worth funding.

What you owe the time

You closed by writing that we owe it to humanity to try, and I take you at your word. A frontier CEO conceding in public that the pace has outrun the ability to keep up is not a cheap sentence to publish.

But notice how fast your rivals agreed — inside a single afternoon, from a man who competes with you for compute and a man who competes with you for everything. The parts of your plan that would actually bind somebody need a waiver from one government and a handshake from another, and nobody signed up for those on Saturday. What got instant assent was the auditor in the lobby.

The other end of the wire is a woman on second shift in Lansing who will be handed one of these systems in about a year, with no training, no certification, no recourse, and no idea what it does when it is wrong. She is not in your essay. She was not in anyone’s.

Now, what I don’t know. Whether one in four is the right fraction, or whether ninety days is long enough to be useful. Who should hold a certification, or audit the issuer, or whether one earned in Lansing means anything in Chennai. I have spent four years on those questions and have partial answers at best.

What I am confident about is narrower. You have built the mechanism already. It has desks and badges and a contract, and every seat faces the same direction.

Turn a few of them around and tell me what you see. If you think that end takes care of itself, say so and tell me why — you would be the first person in this conversation to make that argument out loud, and I would read it closely.

— Venki Padmanabhan

Venki Padmanabhan is a co-founder of the Capability Capital Institute (CapabilityCapital.org) 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.

Ramesh Venkateswaran chaired two convocations this summer, at IIT Palakkad and IIIT Sricity, and shared those stages with Dr. Tessy Thomas of DRDO and Suchitra Ella of Bharat Biotech. He drew a conclusion from the pairing that deserves to be taken seriously, and that is more generous than most of what gets said about Indian education. Both women are products of the Indian system. So are most of the Indian-origin chief executives now running Western firms. Something in that system is working, and we would learn more by identifying it and building on it than by listening to those who only decry.

He is right about the method. Diagnosis is cheap and India has never been short of it. Asking what works is the harder question, and almost nobody asks it.

I want to add three things. One is a caution about the evidence. One is an accusation I can only make against myself. The third is a place to look that no convocation photograph will ever contain.

I. The stage cannot tell us whether the system forms

Everyone on those two stages cleared a filter.

This year 1,79,694 candidates sat JEE Advanced. 56,880 qualified. There were 18,160 undergraduate seats across the twenty-three IITs. Behind those figures stand roughly eighteen lakh who sat JEE Main to get that far.

A country that can pull eighteen thousand out of eighteen lakh, and be right about most of them, possesses an extraordinary sorting instrument. India should say so without embarrassment. Very few nations can do it at that scale, and none can do it at that cost.

But sorting is not forming. A filter measures what arrives at it. It says nothing about who built what arrived.

This is the difficulty with reasoning from the stage. Every person standing on it cleared the filter, which makes every person on it evidence about selection. To learn whether the system forms, we would need to know what became of the 1,22,000 who sat the same exam and did not qualify, and the seventeen lakh below them. Those people are not at the convocation. They are the actual test of the proposition, and they are absent from the sample.

A selection system and a formation system can look identical from the winners’ side of the gate. They are not identical, and only one of them can be built on.

II. The suspicion that it may not be the schools

There is a second possibility, and I raise it because it changes what we would build.

Some of what we credit to Indian education may belong to the Indian household. The transmission of effort, deferral, obligation, and the ordinary domestic assumption that study is not optional. If the home deposits that and the school merely filters on it, then reform aimed at schools will not reach the thing that is working. Worse, we would be crediting an institution for an inheritance.

I cannot settle this from Michigan, and neither can a convocation. But it is testable, and to my knowledge nobody has designed the study.

I will say what I carry myself, because it is the only evidence I own. I was formed at De Nobili School in Dhanbad, under Jesuits, and what I carry from that place is not the syllabus. It is what Father McKenna expected of a boy and would not negotiate. I also still sit with a teacher in the Etawah gharana, in a transmission that is older than any university in this country and appears in no statistic the Ministry publishes. That is the most complete formation system I have personally been inside. It has no accreditation, no examination, and no record. It works anyway.

III. The gate nobody photographs

Now the part I would press on hardest, with respect.

The sample stops at the IITs. India has roughly 14,993 Industrial Training Institutes offering on the order of 26 lakh seats. Add the polytechnics and the apprenticeship system and you have the formation architecture for a working population that dwarfs the entire IIT and IIIT output several times over.

In July the Maharashtra Directorate of Vocational Education and Training published its status report on this year’s centralised ITI admissions. Against a stated intake capacity of about 1,45,144 seats, 79,278 candidates had been admitted. In the same three rounds last year the figure was 1,30,745. The fill rate fell from 86.78 percent to 54.62 percent in a single cycle. Private ITIs took the harder blow, dropping from 40,639 admissions to 16,644.

Before anyone reaches for the obvious explanation, note that 1,70,399 students applied. Applications exceeded seats. This is not young Maharashtrians losing interest in the trades.

Institution operators have attributed the fall to a slower admission process and stricter documentation requirements introduced this year. If that account is even partly right, then roughly fifty thousand young people who wanted a trade did not get one because of domicile papers and processing time.

And it is specific to this gate. Maharashtra’s polytechnics recorded their highest intake in a decade in 2025, over a lakh students at a 93 percent fill rate. Same state, same age group, the adjacent track, moving in the opposite direction.

There is no convocation for any of this. No chief guest, no board chairman on the dais, no photograph. A gate serving many times more young Indians than the IITs lost half its intake in one year, and the loss is recorded in a status document that almost nobody reads.

If the foundation is strong, that is where it is carrying the most weight. And it is the part of the structure nobody is inspecting.

IV. My own account of it

I cannot make this argument from a clean position, so let me make it from a dirty one.

I ran Royal Enfield out of Chennai and took the company from roughly fifty thousand motorcycles a year to about a hundred and thirteen thousand. I am proud of that. It is on my record and it opens doors for me still.

Ask me who formed the people who built the difference, and I have no answer.

I hired, deployed, promoted, and lost thousands of frontline workers. Not one of them left my gate with a document showing what he had become while he was inside it. There was no ledger. There was no line on any statement. A fitter who spent four years with us acquired judgment I could see with my own eyes and could not write down, and when he walked out he carried nothing he could show to the next employer. I was the fourth gate for those people, the first employer, the one gate that pays. I operated it for years and kept no account of what I deposited or what I withdrew.

I can see their faces. The fitter on the line, the man I promoted twice. I cannot tell you what any of them were called.

That is not age. I can still name six men from Buick City and Lansing thirty years on, because I stood beside them at their stations for years. At Royal Enfield I was the chief executive, which meant I saw the faces from a distance, and the plant kept no record that would let me check. The distance made the names unnecessary. The missing ledger made the distance costless. And a company with no ledger is precisely a company where a chief executive can not know the names and still be thought competent. All three are true, and all three are mine.

This is why I do not think the Indian question is only an education question. Three gates precede employment. The fourth gate is where formation either compounds or is spent down, and it sits inside companies, which no education reform reaches.

V. Where it already works, and what it costs to say so honestly

The good news is that the answer already exists in India, roughly two hundred kilometres from where you sat on those two stages, and it was not imported.

ELGi Equipments runs the ELGi Vocational Training School on the grounds of its Air Centre Plant at Kinathukadavu, outside Coimbatore. Eighteen thousand square feet. Classrooms and practical labs for machining, assembly, welding and electrical work. A three-year programme entered after Class 12, selection by written and skill test, interview and medical, a monthly stipend throughout, and teaching done by ELGi’s own employees rather than contracted trainers. Around 130 trainees have come through it. After six years of male-only intake the company inducted nineteen women in 2019 and nineteen more in 2021, and thirty of those graduates now run the airend and top block assembly lines.

The company is candid about its motive. It could not find skilled people to replace an ageing workforce, so it chose, in its own words, to vertically integrate the talent pipeline. This is not philanthropy and should not be praised as though it were. It is a manufacturer deciding that capability has to be produced rather than purchased.

Then, in August 2023, the part that matters most for your question. The school partnered with the Community College of the Central University of Tamil Nadu to deliver a three-year Bachelor of Vocation in Production Technology under the National Skill Qualification Framework. CUTN carries the theoretical curriculum. ELGi delivers about seventy percent of the skill training in its own plant. Graduates are absorbed as employees. The university’s own listings now show the programme running alongside a B.Voc in Manufacturing Technology and a diploma in the same field.

Look at what that structure actually is. The employer forms. A central university certifies. The paper is issued by a body that does not capture the value of the work. That is the firewall aviation has in its examiner and medicine has in its board, built inside Indian manufacturing, under a national framework, by a Coimbatore company and a central university, without anybody calling it a reform.

I would not present it as finished. Because ELGi guarantees absorption, it is also the only destination, so the credential is portable in principle while the career is not. And most of the account of the school comes from ELGi itself, which is the same evidentiary problem as reasoning from the convocation stage. But the mechanism is real, it is running, and it is closer to your two boards than any German or Japanese example anyone will offer you.

VI. What I do not know

I want to be honest about the edge of this, because the diagnosis is the easy half.

I do not know whether the home or the school deserves the credit, and until somebody designs that study, everything either of us says about the strong foundation is inference from survivors.

I do not know who signs. If capability were to be recorded rather than assumed, somebody with standing must confirm it, and that somebody cannot be the employer who benefits from it. Aviation solved this with the instructor and the regulator. Medicine solved it with the board that is not the hospital. Indian manufacturing has no such person, and the guru-shishya tradition, which had exactly that person, was never institutionalised outside the arts.

And I do not know whether India’s scale makes this easier or impossible. Twenty-six lakh ITI seats is either the largest formation asset in the world or the largest stranded one, and I cannot tell from here which.

VII. The discussion you proposed

You floated organising discussions on this, and said you wanted conversations rather than debates. I would like to be in the room, and I would like to suggest a shape.

Put an ITI principal on the same panel as an IIT director. Put a plant head who hires from both on the same panel as a parent from a household that produced a JEE qualifier. One session, four gates, and a single question: which gate actually deposited what this person carries, and who wrote any of it down.

That conversation has not happened in India as far as I know, and the two boards you chair are among the very few platforms with the standing to convene it.

You have asked the right question. I am only suggesting that the answer is standing somewhere outside the photograph.

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.

Sources

Section I. JEE (Advanced) 2026 Results Press Release, 1 June 2026, jeeadv.ac.in: 1,87,389 registered, 1,79,694 appeared in both papers, 56,880 qualified. IIT undergraduate seat figure of 18,160 from JoSAA 2026 seat allocation reporting. The approximate eighteen lakh JEE Main figure is from 2026 press summaries and should be treated as an order of magnitude.

Section III. Directorate of Vocational Education and Training, Maharashtra, centralised ITI admission status report, 27 July 2026, as reported by Careers360 on the same date: 79,278 admitted against an intake capacity of 1,45,144, of which 95,820 government and 49,324 private seats; 54.62 percent of capacity filled. Comparative figures from mahaenews, week of 27 July 2026: 1,30,745 confirmed admissions in the same three rounds in 2025, comprising 90,106 government and 40,639 private, at a fill rate of 86.78 percent; this year 62,634 government and 16,644 private. The 39.36 percent decline is confirmed by the arithmetic of those two totals. One outlet headlined the fall as 32 percent, which does not reconcile with its own figures.

Section III, applications. Free Press Journal, 2 July 2026, citing DVET: 1,70,399 students applied against 1,57,664 seats across 1,004 institutes, 418 government and 586 private, with 78,474 allotted in the first merit list. Note that the seat capacity reported in July (1,57,664) differs from the figure in the 27 July status report (1,45,144). The essay uses the later figure for the fill rate and the earlier one only for the applications comparison.

Section III, polytechnic comparison. Careers360, 12 August 2025, citing the Directorate of Technical Education and Higher and Technical Education Minister Chandrakant Patil: 1,03,115 first-year diploma admissions for 2025-26 at a 93 percent admission rate, described as the highest in a decade.

Section III, national ITI figures. Approximately 14,993 Industrial Training Institutes and about 26 lakh seats for the 2026-27 session, from 2026 state admission reporting. Verify against the DGT dashboard before publication if the number is to be quoted precisely.

Section V. ELGi Equipments, Vocational Training Programme page, elgi.com, accessed 6 August 2026, for the school location at Kinathukadavu, the 18,000 square foot facility, the three-year structure, the selection process and stipend, the 130 trainees inducted, the twenty-year average tenure, and the 2019 and 2021 women’s intakes with thirty graduates now on the airend and top block assembly lines. The vertical integration language is the company’s own.

Section V, the university partnership. Central University of Tamil Nadu, Centre for Skill Development and EVTS-CUTN partnership pages, cutn.ac.in, which independently list the B.Voc in Production Technology, a B.Voc in Manufacturing Technology and a D.Voc in Production Technology delivered through ELGi-EVTS, with enrolment given as 114. Inauguration reported by India Education Diary, 19 August 2023, with Prof. M. Krishnan, Vice Chancellor of CUTN, and Dr. Jairam Varadaraj present. Reported intake differs across sources: ELGi says approximately 100, the 2023 press report says approximately 200, CUTN lists 114. The essay avoids the number for that reason.

Not used, and why. The figure of 400 hours of training per shop floor employee traces to ELGi’s own 2019 Deming Prize communications and is company-sourced. The claim that Dr. Jairam Varadaraj has publicly named middle management as an unsolved problem comes from earlier research in this corpus and was not re-verified; it has been left out rather than risked with a reader in Tamil Nadu.

Ramesh Venkateswaran’s post, LinkedIn, 6 August 2026, on the IIT Palakkad convocation of 18 July 2026 and the IIIT Sricity tenth batch convocation.

Part One ended on a man at a machine, solving the wrong problem well.

He is optimizing. He is a superb optimizer. We just never told him what for.

That is the second failure, and it is the one that does not go away when you fix the board. You can cut twenty metrics to five, name the binding constraint, and publish it in every aisle — and the operator will still be reasoning in the dark, because the number that gives all five their meaning is the one number we have decided he cannot see.

Three Reasons We Don’t Show the Numbers

We do not give the floor the base financials. Ask any plant leadership team why and you will get one of three answers.

They won’t understand it. This one does not survive ten seconds of contact with the people involved. The operator running a household on variable overtime, a mortgage, a car note, and a kid starting community college is performing continuous multi-constraint budgeting under uncertainty, without software, with real consequences for error. He understands margin. He has just never been shown ours.

I want to put a name against that.

Lansing Grand River, during the CTS launch. We had a chronic defect — the instrument panel scuffing the door opening as it went in. The easy move, the one the system rewards, is to blame the operator and log it. Ramon Hernandez, my team leader on instrument-panel install, did the opposite. He took the blame onto himself. He gathered data shift after shift. And he traced the fault back to a tool mechanism on his own team.

Now look at what Ramon was actually doing. A defect caught in-station costs a fraction of one caught at repair, which costs a fraction of one that leaves the plant, which costs a fraction of a warranty claim with a customer’s name on it. That is the cost-of-quality curve, and it drives an enormous share of plant profitability. Ramon worked his way down it, correctly, on nights, while absorbing blame that was not his.

He never used a word of that vocabulary. Money was never in Ramon’s language, and I want to be clear about why: we made sure it wasn’t. We stripped it out deliberately. Cost was management’s business and quality was the floor’s, and we ran the line that way on purpose.

With one exception. The floor did see the cost of scrap parts.

Sit with which number we chose. Not what the plant earned. Not the margin on a car. Not what a defect caught in-station was worth to us. The single piece of financial information we made sure the floor could see was the price of what they had ruined.

That is not withholding out of doubt about their intelligence. That is a curated disclosure, and what it was curated to produce was guilt.

Ramon pulled real money off that plant’s P&L. Nobody ever showed him how much.

They won’t hold it confidentially. Also thin. We hand quality escape data, customer complaint detail, and process specifications to the same people every day. The proprietary material is already in their hands. It is the summary statement we withhold — the one document that would let them see what all their effort adds up to.

If they see the numbers they’ll want more money. This one is honest. And it is worth sitting with what it admits.

It says: our wage will not survive scrutiny. It says we are relying on ignorance to hold the line on labour cost. It says the thing we are protecting is not confidentiality — it is an information asymmetry we have decided is load-bearing.

That is a defensible commercial position. It is not a defensible position for anyone who also claims to be developing people. You cannot form judgment in someone while withholding the only information judgment requires. Formation and information asymmetry are not compatible projects. You have to pick one.

Three Companies That Opened the Books

Springfield Remanufacturing: they taught it from zero, under maximum duress

In 1983, Jack Stack and twelve managers were trying to save 119 jobs at International Harvester’s engine remanufacturing plant in Springfield, Missouri. They put together a hundred thousand dollars in cash and borrowed 8.9 million — at eighteen percent interest. A debt-to-equity ratio of eighty-nine to one, in a dying division of a dying company.

This is the critical fact about SRC and it is usually skipped. Opening the books was not generosity from a comfortable company. It was the only asset they had left. They could not pay competitively, could not guarantee anything, and could not out-equip anyone. So they taught the workforce to read the financial statements and let them play the actual game.

Today SRC Holdings runs 1,900 employees across fourteen business units at roughly $600 million in annual sales. Share value went from ten cents in 1983 to over $764. Hourly employees’ ESOP holdings grew from around $35,000 to about $400,000.

Stack’s description of the business is the sentence I would put on a plant wall: their real business is education, and what they teach people is business. A remanufacturer of diesel engines, in a loud dirty shop where people leave covered in grease, describing itself as an education company. That is the formation thesis, stated by an American plant guy in 1983, without any of the vocabulary.

And on the question of whether the floor can understand a P&L: at SRC, people without a high school diploma can tell you how the toilet paper they use lands on the statement. Not because they were taught accounting. Because they were taught that consumption on the floor arrives somewhere, and where.

ELGi: what formation costs, and what it returns

ELGi Equipments, Coimbatore. In 2019 it became the first globally established industrial air compressor manufacturer outside Japan to win the Deming Prize in over sixty years — an award given to roughly 250 organizations worldwide since 1951.

The managing director, Dr. Jairam Varadaraj, describes the foundations of the ELGi Business System as business processes, measurement-based management, and alignment of individual goals with company goals. That is this essay’s argument, stated by an operator rather than a consultant.

The number that matters: 400 hours of training per person for every shop floor employee, in-house and external, plus a company vocational school. Four hundred hours. Set that against what most plants book as training — a safety video and a signature on an SOP.

The improvement work came through Gemba capability and cross-functional problem-solving teams. Which is to say the intelligence was drawn from the floor, not delivered to it.

And the wage evidence closes the third excuse. ELGi pays among the highest wages in Indian manufacturing; its Coimbatore shop floor workers out-earn much of the white-collar population of the city. Thirty years after Varadaraj came home at liberalization and shut profitable businesses to concentrate on compressors, ELGi is number six in the world.

They paid the money, opened the measures, and took the number six position anyway.

Varadaraj is also honest about where it still fails. Middle management, he has said, is the silent sufferer and the largest source of inertia, and he does not claim to have solved it.

That admission is worth more than any success claim, and I want to be equally plain: neither have I. That layer owns the twenty metrics. It has the most to lose when five become legible, and every scheme I have seen — including some I have run — asks it to dismantle the thing that protects it and offers nothing back. A man who has won a Deming Prize says he has not cracked it. I am not going to stand here and say I have.

So that is the open problem in this essay, and I would rather be sent a story about a plant, a ward, or a jobsite that got through that layer than anything else I could receive.

Siemens: it survives the founder

The obvious objection to the first two is founder dependence — Stack and Varadaraj, remarkable men with unusual convictions. Siemens answers it. The Meister ladder and the German apprenticeship system make formation a structural feature of the firm rather than a trait of the chief executive, and codetermination puts labour in the governance room by law rather than by grace. I spent time embedded in Stuttgart and around IG Metall, and what struck me was not the goodwill. It was that goodwill was not required.

The Loop

Here is the whole thing in one motion.

We assume the floor cannot think. So we build a measurement system that does not require them to think — twenty targets, all mandatory, no objective disclosed, no judgment asked for and none permitted. Predictably they stop thinking, because we have removed both the information and the authority that thinking would need. Their disengagement then confirms the original assumption.

So we do not form them. We do not invest. And when they leave, we go and find another one, and we call the churn a labour market problem.

It is the most efficient self-fulfilling prophecy in industry. Every step is locally rational. The whole thing is a machine for manufacturing the very workforce we complain about.

And the cost is not primarily moral. It is arithmetic. The plant is an optimization problem being solved every hour by the only people standing at the machine — and we have decided, as policy, not to give the solver the objective function. Whatever intelligence exists on that floor is being spent on the wrong problem.

Not a Manufacturing Problem

I have grounded all of this on a plant floor because that is the floor I know. But the three obligations are not an artifact of discrete manufacturing.

A hospital floor: patients to staffing plan, care to standard, nobody gets hurt — and there “nobody” runs in two directions, because the nurse is injured at rates a plant manager would be fired for. A jobsite: work to schedule, work to spec, everybody goes home whole. A dock, a store, a kitchen. Same three, every time. The floor’s irreducible obligations are the same across all six sectors that carry the frontline work of the world.

And so is the board. Ratio compliance, documentation completeness, patient satisfaction, throughput, readmission, never-events, hand hygiene, chart closure — all mandatory, all at target, all at once. That is the same infeasible program, handed to a nurse instead of a machinist, and it produces the same solution: chart to the metric rather than to the patient, because the metric is what gets audited.

Which tells you this is not an industry problem. Six unrelated sectors do not independently invent the same failure. They inherit it. This is management doctrine, taught the same way in the same schools, and applied to roughly three hundred and sixty million people who stand where the work actually happens.

What Changes

The Bloom System’s answer is not a new dashboard. It is that measurement and formation are the same instrument, and we have been holding them as two.

Five metrics, not twenty. A few you are accountable to. Many you are informed by. Never confuse the two — most of the twenty are legitimate diagnostics that were promoted to targets by people who did not understand the difference.

One binding constraint, named. SRC’s Critical Number, in Bloom vocabulary: the thing that binds this year, declared by leadership, changed when it stops binding. Declaring it is leadership’s actual job and almost nobody does it.

The objective function, disclosed. The floor sees what the enterprise is optimizing for. Not a sanitized version. The real one.

Formation sized to make the measures readable. ELGi spends four hundred hours per person. That is the price. The Capability Account exists precisely so that this investment is recorded as what it is — deposited capability, on the books, belonging to the person.

Employment guaranteed for three years, with continuous confirmation. Because you cannot ask someone to optimize for the enterprise’s long-term profitability while their own horizon is the next headcount review. Lincoln understood this in 1958.

Capital is crystallized labor. Labor is capital in formation.

The operator at the machine is already solving the problem. Continuously, competently, every hour of every shift. The only question ever on the table is whether we tell him what he is solving for.

So go and look at your board tomorrow morning.

Count the metrics. If there are more than seven, you are not measuring — you are insuring. Every one of those numbers was added by someone protecting themselves from a question they once got asked badly, and not one of them has ever been retired, because retiring a metric requires somebody to say out loud that it did not matter.

Then find the one that binds. Not the one that is red. The one where relieving it would move everything else. If you cannot find it, you have your answer about what the other nineteen are for.

I have run boards that would not have survived this test. I am not asking anything of you that I did well myself.

First in this pair: “No Feasible Solution.”

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.

Sources: SRC Holdings history and figures — greatgame.com and The Great Game of Business (Stack and Burlingham). ELGi — 2019 Deming Prize announcements and Dr. Jairam Varadaraj’s published interviews.

A manufacturing plant has to do three things.

Make the product to plan. Make it to quality. Send everybody home whole.

That is it. Three. A new hire can hold all three in his head by the end of his first shift, and every one of them is visible from where he stands. He can see whether the line is running. He can see whether the part is right. He can see whether the guy next to him still has all his fingers.

Add cost to budget at the plant manager’s desk and you have four. Add the growth number the business is actually being run for — the one the board asks about — and you have five.

Five. That is the entire measurement architecture of a manufacturing enterprise. It fits on a card.

So how does every plant I have ever worked in end up with twenty?

The Board Was Green

I have stood in front of a board with twenty-two metrics on it, every one of them green, and known that at least six were dressed.

Not lies I told. Lies the system required. Each number carried a target set by someone who had never run the line it measured. Each target was to be hit at one hundred percent. Twenty-two of them. Simultaneously. Anyone who has run a plant knows the arithmetic: hitting all twenty-two is not possible, so what goes on the board is not performance. It is a negotiation between what happened and what can be defended.

I want to be precise about my position in that room. I was not the victim of that board. I built boards like it. I presented them. I coached supervisors on which number to lead with and which to bury in the appendix, and I called it managing upward — which is the phrase we use when we do not want to say the other thing.

The floor knew. They always know. They watch the board go up and can tell you within thirty seconds which numbers are real and which are dressed, because they are the ones who dressed them — which line ran overtime to make plan, which quality escape got recategorized as a customer specification issue.

And they draw the only available conclusion. That the measurement is theater. That management knows it is theater. That the job is to not disturb it.

That is the day formation dies. Not when the training budget gets cut. When the floor learns that the numbers are a performance, and that nobody upstairs wants to know.

Three Metrics, Colliding

Let me show you the three colliding, before I show you twenty.

Lansing Delta Township. We installed a robot at the spare tire load station — a productivity project that took out one person a shift, three across the three shifts. The implementation was a disaster. Controls faulted nightly. And because there were no longer human eyes on cars entering the cell, any part even marginally out of position would stop the robot. That happened almost daily. The line would go down in chassis, and chassis would charge the downtime to my trim shop.

There was a second loss nobody had costed. The spare tire load station was where the car came down off the elevated conveyor — the one spot on the line where a team leader could reach a defect at working height and fix it before the car went out the door. The robot took that spot. It did not appear anywhere in the business case, because the business case counted three heads and nothing else.

So one night a defect comes down from trim needing repair. Kirk Branson, one of my best team leaders and one of the few I had loose that night, walks half a mile down the line, times the car, and gets himself into the enclosure to make the fix without stopping anything.

He knew about the light curtain. He had worked around light curtains for years, and he took careful, intelligent contortions to get in and out without breaking the beam.

What he did not know was that this cell had two of them, in two planes. Nobody in trim had been trained on the safety system around that robot. It was his first time inside it. He had reasoned carefully about a system we had never taught him, and reasoned correctly about the half he knew.

He tripped the other one. Chassis went down about thirty minutes on a line building fifty-five cars an hour. Do that arithmetic.

And on paper, what I now had was a gross safety violation by one of my best people.

Look at what the metric system saw. Quality said fix the defect before it leaves. Throughput said do not stop the line. Safety said do not enter the cell. All three were mandatory. All three were correct. And the one man on the floor with the judgment to hold all three at once had been handed a cell nobody taught him, in a layout that had quietly removed the only place those three used to be reconcilable.

He did not fail the three metrics. We had made them unsatisfiable, and then measured him against them anyway.

He was never in danger. The system worked exactly as designed. It detected a man where a man should not be, and it stopped.

But the downtime was charged to my shop, and so we called a discipline meeting.

I want to be exact about what we were doing, because at the time we told ourselves it was about safety. It was not. We were angry about the downtime, and we reached for the safety file to justify it — because we had terminated skilled trades and maintenance supervisors before for lockout-tagout failures, those cases were sitting right there, and they let us treat a training gap we had created as a violation Kirk had committed.

He heard a discipline meeting was coming and he was shocked. Then he was furious. He had walked half a mile on his own time to save a car, and we were building a file on him.

His supervisor disciplined him in front of the shop committeeman. The committeeman grieved it on the spot, in full.

He was right to. That is the sentence I have taken thirty years to be able to write plainly. The grievance was the only part of that entire process that was working correctly, and at the time I experienced it as an obstacle.

Then came the follow-up action, and this is the part I would ask you to hold onto. We banned repairs in the area. That was the fix. We never did find another station where trim could reach a car without interfering with something.

So the defects went to the end of the line.

Shortly after, Kirk volunteered for the pilot team. He went out of disgust.

I lost one of the strongest team leaders on my floor. The plant recorded it as a transfer.

The Problem We Were Trained to Solve

I am an industrial engineer. Doctorate, Pittsburgh. And in hindsight the most useful course I ever took was linear optimization.

Here is what it teaches. You state an objective — the one thing you are maximizing or minimizing. You state your constraints — the resources you actually have, not the ones you wish for. Then you solve for the best point reachable inside them. Not the ideal point. The best reachable one.

The discipline is in what it forces you to admit. Name what you are optimizing for. Name what limits you. And accept that some constraints will bind — that certain things sit at their ceiling and cannot improve without something else getting worse. That is not failure. That is the shape of the answer.

Now consider what a plant actually is. Finite people, finite machine hours, finite capital, all of it scarcer or dearer every year as the labor market tightens and the capex ask gets halved in October. One of the cleanest optimization problems in the industrial world, running continuously, in real time.

You would think we would be kings of it.

We are not. In thirty-six years across four countries I have almost never seen a plant run as an optimization problem. I have seen hundreds run as a list of demands.

Because here is what twenty metrics at one hundred percent actually is. It is a constraint set with no objective function and no feasible region. Every metric converted from a variable into a constraint, every constraint set at its individual maximum. Hand that to any solver and it returns two words.

No solution.

Not a hard problem. Not a stretch goal. Infeasible. Any first-year student would recognize it in ten seconds. We hand it to the floor every morning and call the resulting failure a discipline problem.

And the impossibility is not the worst of it. In a real optimization the binding constraint is the most valuable information in the system: it tells you where the next dollar goes, because relieving it is the only thing that moves the answer. Everything else is noise until that one gives.

When everything is set at one hundred percent, nothing is identifiable as binding. You have lost the diagnosis — and the diagnosis was the entire reason to measure.

Two Companies That Did the Arithmetic

The counter-argument is that this is idealism — that real plants cannot run on five numbers. So let me put two companies on the stand, and note that neither is a software firm with unlimited margin.

Lincoln Electric: five factors, and the reward flows from them

Lincoln has run the same architecture for the better part of a century. Piecework employees are rated on five performance factors, each worth a maximum of twenty points: productivity, quality, adaptability, dependability, and compliance with environmental, health and safety priorities. Average score set at one hundred. Twice a year.

Five. Not twenty. And each one is legible on a single card.

The tradeoffs are resolved by design rather than exhortation. Piecework alone would tempt volume over quality, so each employee inspects their own parts, corrects defective work on their own time, and records tie every piece to a person. Rating quality separately closes the loop. Rates are held constant when output rises, changed only when technology changes the job — and the worker can challenge the change.

The objective function is disclosed. The board sets aside roughly thirty-two percent of profits before interest and taxes for the bonus pool. Since 1955 the bonus has averaged seventy-seven percent of wages. The 2021 average bonus in Cleveland and Mentor was $25,211, with average full-time production total compensation above $73,000.

Which means the number the floor is optimizing for is the same number the company is optimizing for, and every welder in the building can compute it.

Two more pieces. The Advisory Board, since 1914 — elected representatives from every division, meeting with the president every two weeks, for over a century. And guaranteed employment since 1958, no layoffs since, traded explicitly for flexibility: reduced hours in slow times, transfers to lower-paid work, a thirty-hour weekly floor. Security and flexibility, both named, both binding on both parties.

Springfield Remanufacturing: one number, chosen because it binds

SRC was a dying division of International Harvester in 1983, bought by thirteen people who had no money and no alternative. What they built out of it is a longer story, and I will tell it properly in Part Two. For now, take just one mechanism from them.

The Critical Number — one number per year, chosen because it is what could kill you this year, and changed when the binding changes. That is a shadow price with a name on it. It is the structural opposite of twenty permanent metrics at one hundred percent, and it forces leadership to do the one piece of work that cannot be pushed downward: decide what actually binds.

And the Huddle — weekly, names on the numbers, line owners forecasting their own line forward, in front of their peers, before the result is known. Forecasting forward is the trust move. It treats the operator as someone with judgment about what is coming rather than a recorder of what already happened.

Neither company is running a philosophy. Both are running an arithmetic that closes.

What We Do Instead

Leadership’s work in an optimization problem is to declare the objective function and name the true constraints. Only leadership can see the whole program, so only leadership can do it.

Instead we withhold the objective, declare every constraint binding, and hand an infeasible problem to the operator.

Who then solves the only problem available to him. Which is: what gets me yelled at least today.

He is optimizing. He is a superb optimizer.

We just never told him what for.

Before You Read Part Two

Go and count your board tomorrow morning.

Not to judge it. Just count. Then try to find the one that binds — not the one that is red, the one where relieving it would move everything else. If you can find it in under a minute, you are running a real program. If you cannot, you already know what the others are for.

Then ask the harder question, which is the one I got wrong with Kirk. Is there anything on that board a good person could not satisfy without breaking something else on it? Because that is not a discipline problem when it happens. That is arithmetic, and it is yours, not theirs.

I would rather hear what you find than almost anything else. Especially if it is a board that closes — I have seen very few, and I would like to be shown one.

Part Two — “A Blinded Floor” — takes up the second half: what happens when the board is feasible and the man still cannot see what he is solving for.

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.

Sources: Lincoln Electric merit rating and bonus structure — The Globe and Mail, Lincoln Electric corporate materials, and Lincoln Electric Holdings proxy filings. SRC Holdings history and figures — greatgame.com and The Great Game of Business (Stack and Burlingham).

Evidence They Can’t Defend — Essay 10 of 13

Every vehicle launch has a signature failure, and on a modern car it is electrical.

Not the components. The components arrive working. What goes wrong is the connections — a harness seated proud instead of flush, an optical fibre nicked somewhere during the long violence of installing a vehicle’s insides. The part is fine. The joining of it is not. And the fault often presents nowhere near where it lives, so a customer finds it three months later in a driveway.

For the 2023 launch at Lansing Delta Township, the plan for preventing that had two parts, and I want to describe both precisely, because between them they explain the whole of this essay.

The first was training. Operators would learn the electrical work on a digital facsimile of the vehicle, in the trim shop — clicking on parts, clicking on points in an image of the car, clicking tools, completing sequences on a screen.

I thought then and think now that this is a classic formation error. There is no substitute for a master showing, instructing, observing, and certifying. Toyota has a word for the place where that happens: the dojo, a training area off the line where a new operator practises the actual work under a seasoned hand until someone competent signs off that they can do it to standard. Then they take a station. We had a screen with clickable parts.

I have watched that error take the same shape for thirty-six years. Somebody skips the formation, buys an instrument to catch what the formation would have prevented, and then defends the instrument long after it stops working. What follows is the cleanest example I have, and it cost a million dollars.

The second part was inspection. To stop cars leaving the trim shop with electrical faults, the plant installed about a million dollars of equipment that plugged into the vehicle’s electronics and verified circuit completion and function.

There was one problem, and it was mine. There was not enough cycle time for the test and not enough footprint to put it anywhere except at the end of the trim shop — which is to say, at my waypoint. I had thirty cars between me and chassis. Every blip on that tester ate into the bank, and if the bank emptied, chassis starved and I became Herbie: the constraint the whole plant is waiting on.

So watch what the million dollars actually bought. False alarms. Cycle-time problems from the equipment and from the inspectors operating it. And diagnosis time — because when the tester flagged a circuit, somebody still had to find out what was wrong with it, and finding out is slow.

Every one of those costs pointed back to the same place. Not to the tester. To the training that never happened: no dojo, no verified electrical skills, no repetitions on this particular vehicle under someone who could confirm the work was right.

And here is the ending, which is why I am telling you this in an essay about algorithms. Within about six months of ramp-up, the operators had accumulated enough repetitions to climb the learning curve on their own — the hard way, on live cars, without a dojo. The tester’s catch rate fell to almost nothing. It had become expensive and useless at the same time.

We disabled it. For good.

A million dollars of inspection technology, deployed as a substitute for formation, switched off six months later because the humans had finally learned what nobody had taught them.

That’s the argument. Every extraction mechanism in this series required a human decision. The gig economy automated the deciding — and what it automates, over and over, is the same swap I watched at LDT: a measurement system standing in for a formation system, priced as if it were cheaper, and defended long after it stops working.

Algorithmic wage suppression

Traditional wage suppression works through institutional mechanisms: weakened unions, stagnant minimums, industry norms. It operates over years.

Algorithmic wage suppression operates in real time.

Ride-share platforms can adjust the per-mile rate a driver receives minute by minute. If the algorithm detects that drivers in a market have limited alternatives, it can lower rates without any human deciding to suppress a wage. The suppression sits inside the optimisation function.

A quarter of gig workers earn less than the state or federal minimum wage once all hours, including waiting time, are counted. Because they are classified as independent contractors, minimum wage law does not apply. The misclassification of the previous essay, industrialised.

The Amazon model

Amazon warehouses are where digital extraction meets physical labour. Workers are tracked by algorithms measuring “time off task” — any period not actively scanning, picking, or moving product.

A 2023 OSHA investigation found Amazon warehouse workers suffering serious injuries at more than twice the rate of non-Amazon warehouses.

This is the logical endpoint of Frederick Taylor’s scientific management. Taylor wanted to separate thinking from doing — management determines the one best way, workers execute without deviation. The algorithm completes it: determines the method in real time, monitors compliance continuously, terminates automatically.

Taylor’s stopwatch is now continuous sensor data. His foremen are automated tracking. And his suppression of worker intelligence is perfected by a system that does not need to suppress intelligence, because it has no input channel for it.

That last clause is the one I would underline. My tester had no input channel either. It could tell me a circuit had failed. It could not tell me the operator had never been taught to seat the connector, which was the only fact that mattered.

The extraction arithmetic

Uber went public in 2019 at an $82 billion valuation and reported its first consistent annual profit in 2024. Investors held it through the loss years on its ability to take a dominant platform position and extract from it later.

Where does that profit come from? The difference between what customers pay and what drivers receive. Every fraction of a cent squeezed from routing, pricing, or allocation flows to the platform.

Uber’s CEO made $24.3 million in 2023. The median driver earns between $25,000 and $35,000 before expenses, and after vehicle costs, fuel, insurance, and maintenance — all borne by the “independent contractor” — net earnings often fall below minimum wage.

The platform captures the value. The worker carries the cost. The algorithm optimises the extraction.

I keep coming back to my thirty cars. The bank between me and chassis was visible — I could stand at the waypoint and count it. A driver has a bank too: the distance between the fare and what the car actually costs to run. Nobody puts it where he can count it.

Surveillance as extraction

Ring doorbells, Alexa devices, delivery driver cameras — all generate data. The data trains algorithms, the algorithms cut labour cost, and lower cost lifts margin, stock, and pay at the top. The worker is both data source and labour source, and is paid for the labour only.

Why the machine wins the meeting

The million dollars was a formation error. It was also an accounting one, and the accounting is what made it inevitable.

A tester is equipment. It goes on the balance sheet, it depreciates over years, and the launch quarter absorbs a fraction of the cost. A dojo is people — hours, floor space, a seasoned hand pulled off the line to teach. All of it expenses in the period it happens, which is the launch quarter, when the schedule is tightest and every dollar is visible.

Two investments aimed at the same defect, landing on the books in opposite directions. One buys an asset. The other reduces earnings. Nobody in that room has to prefer a machine over a person for the machine to win. The rule prefers it for them, before anyone opens their mouth.

The dojo was never rejected. It was never priced as an investment, because there is no line on any statement where it becomes one. It only ever appears as a cost.

The algorithm did not invent this. It inherited it.

The counter-example

Costco uses technology to extend worker capability rather than to surveil it. Its systems support the person’s judgment rather than replace it. The results are familiar from both series: lower turnover, higher productivity, better customer experience, superior stock performance.

The technology is neutral. An algorithm can be designed to maximise extraction from workers or to maximise deployment of what they know. The choice reflects values, not technical constraints.

What I do not know

Here is where my own story stops being comfortable, and I would rather say so.

I do not know whether the tester was wrong. It was expensive, it starved my bank, it generated false alarms, and we switched it off — but it also ran during the exact months when the operators were least formed, which is when the risk of shipping a fault was highest. It is entirely possible it caught things in month two that would have reached a customer. I cannot prove it did. Nobody counted, because the whole point of a fault caught early is that nothing happens.

I do not know how to tell, in advance, which of the two investments to make. Naming the accounting rule explains why the machine wins the meeting. It does not prove the meeting was decided wrong. Sometimes the tester is the right call, and I have no instrument that would let anyone tell which case they are in.

And I do not know how you do any of this on a floor where the worker is a contractor and the platform is the employer. Every fix I believe in — the dojo, the master’s sign-off, the reps under observation — assumes somebody has a durable relationship with the person doing the work. The gig platform is designed so that nobody does.

That is the hardest version of the problem in this whole series, and I do not have it solved.

So tell me about a measurement system where you work that was installed to catch what nobody was trained to do right. Is it still running? And if it was ever switched off, what actually changed — the technology, or the people?

Next: The Pension They Promised — how the shift from defined-benefit pensions to 401(k)s transferred all retirement risk to workers while companies captured the savings.

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.

Sources: OSHA Amazon warehouse investigation findings (2023); EPI gig economy and misclassification data; Bureau of Labor Statistics injury rate data; Uber and DoorDash SEC filings; National Employment Law Project gig worker wage analysis; academic research on algorithmic management and worker surveillance; Toyota Motor Manufacturing and Lean Enterprise Institute on the dojo as a pre-line training and certification area.

The Lansing Delta Township launch account is the author’s own.