NP Nithin Palal

Chapter 27: The Investor’s Playbook

Part V: Conscious Strategy

· 36 min read · Discuss · Plain text

AI-Proof: What Machines Can Never Be — Volume II: Application, Part V: Conscious Strategy, Chapter 27: The Investor’s Playbook

Part V looks at strategy. After vision, the middle, confirmation, speed, regulation and business ideas in Chapters 21 to 26, this chapter turns to the money behind them: what an investor can judge about strangers, and what he answers for.

Rajan was asked to judge a start-up that could send his buyers to other mills. The fund’s machine found two of his three objections first. No one knows if a firm will last. An investor can test how people act when wrong, and own his stake.

In Brief For readers who invest in, recommend or vouch for young companies they cannot see the end of. Chapter 27 follows a cloth exporter in Coimbatore who is asked by his nephew’s venture fund to judge a young company that matches buyers to mills, and so could send his own buyers elsewhere. He reads its deck, telephones three objections, flies to Singapore and asks the founders whether their platform recommends the mill he trusts most. Keynes on how little anyone knows about an investment’s future and the Tirukkuṟaḷ on testing people before trusting them help separate what an investor can judge, which is conduct, from what he cannot, which is the future, and show what he answers for when he lends his name.

An investor cannot know whether a company will last; he can test how its people act when they are wrong, say what he himself stands to gain, and answer for the trust he lends, including the trust of people who were not in the room.

Opening: Forty Pages in Coimbatore

The deck ran to forty pages, and Rajan Subramanian had it printed.

He read it at his desk in the office above his factory in Coimbatore, with a pencil, the way he reads contracts. He has run the business since 1997. It makes and exports cloth, and about 140 people weave and hand-finish it on the floor below for buyers in fourteen countries. The yarn is dyed by dye houses in Tiruppur. His nephew Karthik works for a venture firm in Singapore, which he joined in July 2025, and he phones his uncle every Sunday. On that week’s call he had asked Rajan to read the deck. The firm was deciding whether to invest in the company it described, and Karthik’s partners wanted the view of someone who worked in the trade.

The company was not yet two years old. It had two founders, an engineer from Chennai and a former cloth buyer from Zurich, and it was registered in Chennai. Its platform matched buyers’ orders to mills, predicted which lots would fail inspection, planned shipments and scored suppliers for risk. A pilot with about forty mills around Tiruppur claimed to have cut the time a buyer needed to find a supplier by a third.

On page eleven was a map of competitors: a square cut into four, with company logos placed in it, and the start-up’s own logo in the best corner. Every logo belonged to a software company. Rajan went over the page with his pencil, looking for the people buyers had turned to for a hundred years: makers and exporters, agents, brokers, buying houses. There were none. He was not on the map, and nor was anyone he knew.

It was September 2026. Eighteen months earlier he had ordered every system the vendors offered in one phone call, and the haste had cost him three inspectors and four months of a silent floor. Nobody could call him a man who did not know these systems. He knew where they broke, and he knew they were not always the ones that broke.

He also knew, before he had read past page eleven, that he was not a neutral reader. A company like this one could offer his buyers other mills. Jürgen Weiss, who has bought from Rajan for a firm in Hamburg since 2013, now buys some cloth through a sourcing platform that matches buyers to mills across South India. In March 2026 he split a line of yarn-dyed checks between Rajan and that platform. Six months later Weiss had not yet decided which of them to keep.

This chapter asks: When a machine can copy the product and nobody can know whether the company will last, what is an investor judging, and what does he answer for?

The print edition gave an answer that this one cannot give. It said that in a company built on someone else’s model, the defences that matter most are the least measurable. It named the founders’ knowledge of their trade, their relationships and their ability to adapt. “These are assessments of consciousness,” it said. “No screening tool has learned to exercise them.” The fund in this chapter had a screening tool, and the tool found two of Rajan’s three objections before he did. What an investor can judge is not anyone’s consciousness. It is conduct: what people do when they are wrong, and what they did when the world contradicted them. Whether a young company will last is mostly unknowable, and anyone who says otherwise is assuming that the present will continue. And what the investor answers for is his own judgement, the interest behind it, and the trust he lends to people who were never asked.

The chapter keeps its title, and the title needs an argument. A playbook is a set of moves that work whoever makes them. As of 2026, a machine will write one for any reader in a minute. Rajan’s three questions, which come in 27.2, are of another kind. They work only for someone who will be there when the answer turns out to be wrong.

27.1 What the Deck Does Not Say

The deck was polished. Its charts were clean and its sentences were smooth. Rajan distrusted the smoothness more than any number in it. In nearly thirty years of business, nothing had ever been smooth.

Polish used to mean that someone had worked hard. By 2026 it means very little. A machine will lay out forty clean pages from a page of notes in an afternoon. This book has argued that a fluent sentence stopped being a sign of knowledge when machines learned to write fluent sentences. The same is true of a fluent deck. It shows that someone has a machine, and nothing more.

The pilot had measured one thing: how long a buyer took to find a mill. Rajan knew three of the forty mills. He knew that in Tiruppur the time to find a supplier depends on festival weeks and power cuts, a monsoon that comes early, an order already promised to a cousin. It depends on owners who would rather lose a contract than be told by software how to run their mills. The pilot had taken one measure out of that tangle and presented it as the whole.

Then the map. The deck compared the start-up with other software companies. It did not compare it with the people a buyer turns to today instead. They are makers and exporters like Rajan, agents who know which mills are idle, and buying houses that inspect cloth for European chains. None of them was a technology company, yet they were the start-up’s real competition, and the deck treated them as if they did not exist.

Rajan saw this first, and it is worth being honest about why. He saw it because he was one of the people left out. The first thing the deck did not say was who would lose if it worked, and Rajan was one of them. That does not make his reading wrong. A rival often sees a weakness that a friend will not mention. But it means that his reading came from someone with something to lose, and anyone weighing it had to know that.

Nor was he the only one in the matter with an interest. The founders wanted money. The partner who would lead the fund’s investment wanted a return for the fund’s own investors. Karthik had a job at the fund and an uncle at the table. A deck is testimony from people who want to be believed, read by people who want something too. None of that makes it false, but a reader should believe it only as far as he can check it, or as far as someone will stand behind it. Of each claim he should ask what supports it, and who must explain if it is wrong. Fluency can no longer do that work for him, and neither can a friendly face.

The print edition put the deck’s mistake in one line: the founders treated a trust problem as a data problem. That is half right. Rajan had learned in 2025, from his own records, that data about deliveries is part of trust, and sometimes the more honest part. What the deck left out was something else. It had a slide on the platform’s accuracy. It had no slide on what happens to a mill when the platform is wrong about it.

Rajan also knew the other side of that. When his own scheduler began to move loom time away from his speciality cloth in 2025, he thought at first that it could not see what mattered. His floor supervisor, Muthu, told him the machine was not wrong about the numbers: German buyers were ordering fewer speciality pieces, and none of them, Rajan included, had noticed. Rajan read the deck knowing that his first suspicion of a machine had been wrong before.

27.2 The Durability Test

Rajan does not speak the language of venture capital. He does not talk about defensibility or addressable markets. He has three questions, built from nearly thirty years of judging suppliers, buyers and partners.

Is this person serious? Will this last? What happens when things go wrong?

They are simple questions. The title of this section promises a test of durability, and it needs one correction. Rajan’s questions test something narrower than a company’s life: whether a person’s conduct lasts when it is under strain. The second question is the one that nobody can answer.

In 1936 the English economist John Maynard Keynes wrote the most honest paragraph I know about investment. “Our knowledge of the factors which will govern the yield of an investment some years hence is usually very slight and often negligible.” Then he went further. “If we speak frankly, we have to admit that our basis of knowledge for estimating the yield ten years hence of a railway, a copper mine, a textile factory, the goodwill of a patent medicine, an Atlantic liner, a building in the City of London amounts to little and sometimes to nothing; or even five years hence.”1 A textile factory is on his list. If nobody can know what Rajan’s own business will earn in ten years, nobody can know it of a start-up that has been trading for less than two.

The fund’s screening model could do something Rajan could not. It had been built from the decks and later records of hundreds of companies, and it could say how companies like this one had usually ended. That is a real advantage, and it is the machine’s. Rajan’s sense of how start-ups end came from a handful he had heard about. But a rate for companies like this one is not a forecast for this one. It is still the right place to start, and a reader who ignores it is guessing with less. On the second question, the model and the man were both guessing, and only the model knew how often guesses like its own had been wrong. The guessing was not only a limit of one model in 2026. A young company’s future depends partly on choices that nobody has made yet, including the investors’ own. Keynes added that those who seriously try to make such estimates are often so few that they do not govern the market. Professional investors, he argued, spend most of their skill guessing how the crowd’s opinion will move, and some go as far as “anticipating what average opinion expects the average opinion to be”.2

Keynes was not saying that estimates are useless.2 Parts of a forecast can be checked: how many buyers in the pilot came back, what one match costs, and the money needed before the plan can be tested. What cannot be checked is whether those parts will add up to a company ten years on. This book draws a plain rule from that: uncertainty is a reason to show the assumptions, not to skip them. A founder who answers every hard question with “nobody knows” has misread Keynes.

The first question is different. A person can be tested in a way that a future cannot. The Tirukkuṟaḷ, a classical Tamil book of couplets on virtue, wealth and love, gives a whole chapter to testing people before trusting them. In V. V. S. Aiyar’s translation (1916), one couplet reads: “Dost thou want to find out whether a man is noble or little-minded? Know that conduct is the touchstone of character.”3 The Tamil word for conduct here, karumam, means a person’s deeds. The chapter is counsel to a king choosing men for his business, not to an investor, and I borrow only its test. What counts is what a person does.

A pitch is not conduct. It is words about conduct, rehearsed. So the conduct an investor can see in a young company is small. He can see what the founders did when the world contradicted them. He can hear how they reply when asked about what they do not control. And he can try to learn whether, when something goes wrong, they will call before they are called, as Senthil, the mill owner Rajan trusts most, once did for him. That is where Rajan would look.

That is also what Rajan means by “serious”. A serious person, in his trade, is someone who treats his own promise as binding on himself, and who will still be there to be asked about it. Two years of a founder’s life is a real stake, though it shows nothing yet of what he does when the stake is at risk. Seriousness is also not skill: a founder can treat his promise as binding and still be unable to keep it. And conduct is not the whole decision. A candid founder can build a business that does not pay. Conduct tells an investor whom he is trusting with the money. Whether buyers want the thing, whether it pays its way and whether the price is fair still have to be asked. Rajan’s three questions cannot settle those. He would press on some of them himself, but the fund’s committee had to answer all of them.

The third question is where the two voices meet. The couplet’s chapter trusts a person once he has been tried, and it ends by warning against distrusting the tried as much as trusting the untried.4 Keynes trusts no long forecast, and he describes investors who act on convention and on what he called “animal spirits”, “a spontaneous urge to action rather than inaction”.5 They are answering different questions, one about a person and one about an enterprise’s future, and it would be false to set them against each other. But an investor in a young company must ask both at once, and can test only the first. Rajan’s questions split exactly there. The first is the couplet’s. The second is Keynes’s. The third asks what each side will do when the forecast fails. Only staying long enough shows it.

The other side

Rajan’s first question bets on people. The strongest evidence against that bet comes from three finance economists, Steven Kaplan, Berk Sensoy and Per Strömberg. Their paper of 2009 borrows an image from racing, in which the jockey is the team and the horse is the business. They followed 50 companies backed by venture capital from an early business plan to a public listing, and on to about three years after it.5 The businesses stayed “remarkably stable”. The people did not. Only 44 per cent of the chief executives at the end had been chief executive at the business plan. At the margin, the authors concluded, investors should weigh the business more heavily than the team.

That partly wins. A serious founder, in Rajan’s sense, is one who will still be there to be asked. The study shows that many are not, even in companies that succeed. But its authors state one limit themselves: “the results do not imply that good management is not important.” And it followed only companies that reached a listing, so it says little about those that failed on the way. On my reading, that is where conduct under strain matters most.

The rest of Rajan’s judgement does not depend on the founders staying. The conduct he tested decides whom he trusts with money now. What he lends his name to, and what he will answer for, stay his whichever rider is on the horse.

27.3 Three Phone Calls

Rajan spent a week with the deck. He telephoned Karthik three times.

The first call came on the evening he found the map. “They are not serious people,” Rajan said. “They have mapped the companies that do not matter and ignored the people who do.”

“That is exactly why you should come,” Karthik said.

“To tell them their deck is incomplete?”

“To tell them what is missing. That is a different thing.”

Rajan put the phone down with the feeling that he was being handled, patiently, the way he handled mill owners who resisted a change they needed to make. The feeling was accurate. Karthik wanted his firm’s decision to be a good one, and he wanted his uncle in the room. Both wants were his own, and the second was a nephew’s as much as an employee’s.

Two days later Rajan called again. He had worked out that the growth projections assumed mill owners would adopt the platform faster than most of them had adopted mobile banking. “They are dreamers,” he said. “They do not understand how slowly this industry moves.”

“Come anyway,” Karthik said.

Before he rang off, Rajan asked a question he had been carrying since Sunday. Did Karthik’s partners know that the man advising them was his uncle? They knew, Karthik said; everyone in the office knew. Was it written down anywhere? It was not. Karthik wrote it that night, in a note to the partner who would lead the investment, and copied his uncle. It cost him a little. In writing, part of his enthusiasm looked like what it was: a nephew’s.

At the end of the week Rajan called a third time. He had read the technology section four times. He still could not tell which parts the founders had built and which they rented from the company that made their model. “They are unclear,” he said.

Karthik was quiet for a moment, and then he told his uncle the truth. The fund’s screening model had flagged two of Rajan’s three problems on the first day. It had marked the adoption curve as faster than any comparable company had managed. It had marked the blur between what was built and what was licensed. “The map was yours,” Karthik said. “Our model compares them with software companies. So does their deck. Neither of them has a square for you.”

The model had been given the deck’s own question: how does this company compare with others like it? It answered that question well. Nobody had asked it who would lose if the company worked. Someone at the fund had also chosen what the model compared a company with, and what it counted as a good outcome. That choice was the fund’s own map, and nobody had put it on the table. The model judged the start-up much as the platform would judge mills: it compared the start-up with others, from records the start-up had never seen. The difference was that the start-up had asked to be judged.

So the print edition’s line, in which Karthik tells his uncle that he has found “three problems that my entire team missed”, is withdrawn. The machine had done most of the reading, and done it faster. What Rajan saw that the machine did not was the thing that threatened him. That is a real contribution, and it is also a warning. The part of a judgement that a man adds to the machine’s may be exactly the part where his own interest lives.

He booked the flight anyway. His wife asked what had changed. He said he wanted to see the people who might take his buyers. It was true, and it was not the whole reason. Part of it was Karthik. Part of it was close to what Keynes meant by animal spirits. It was the wish to do something rather than nothing, which no spreadsheet could justify.

27.4 The Dependency Beneath the Deck

The fund’s office was high above Singapore’s harbour, and the container ships below looked small enough to pick up. The founders were young, prepared and polite. A partner of the fund sat at the end of the table, and Karthik beside his uncle.

Rajan began with what he was. “Before I ask you anything, you should know this,” he said. “I sell cloth to buyers who could find other mills through you. If your company does well, some of my buyers may not need me. Weigh what I say with that in mind.”

The founders looked at each other, and the engineer from Chennai thanked him. Nobody else at the table had said anything like it. An interested judge can still be a good one. A judgement that hides its interest, though, asks to be trusted on terms it has not disclosed, and Rajan had just refused to ask that.

Then he asked about the model. The deck described its abilities over many slides and never said whose it was. “If your model provider decides to build what you have built,” he said, “what do you have left?”

The founders’ answer was honest. They had their data, their customers and the ways of working they had built around the trade. Whether those would be enough without the model underneath was, they said, an open question.

“That is the right answer,” Rajan said. “The wrong answer is ‘that will not happen.’”

It happens often enough to have a name. When a platform’s owner builds into its own system what a small company on that platform was selling, developers call it being “Sherlocked”, after a search tool that Apple built which was strikingly like a small company’s program.6 A company that rents its intelligence from a model maker depends on that maker’s prices, its next version and its plans. The maker shares its plans with nobody. Rajan’s version was plainer. He had asked it for nearly thirty years about cotton mills, shipping agents and dye houses. “If your key supplier disappears, what do you have?”

But the deck had a second supplier that it did not mention either: the mills. Everything the platform knew about a mill came from the mill’s own lots, deliveries and records, gathered in the pilot. The mills could refuse to be recorded. They could leave. They could be misread and never know it. And “their data” was partly the mills’ data. The deck did not show whether the company could keep it after a mill left, use it to improve its scores, or pass it to the model maker. The model maker could withdraw its machine, and the mills could withdraw their trust, and the deck named neither dependence. Depending on others is ordinary: every business depends on someone, and Rajan depended on Weiss. Hiding it is the fault.

The founders’ reply about the model was a small piece of conduct, and the only kind a meeting can show. It named a dependence they could not control and did not pretend to settle it. A founder who says “that will not happen” is promising something that is not in his power. A promise is worth something only when the thing promised is. Reducing the dependence is work, and that was the next question: could the platform move to another model, and what would make the mills stay?

Rajan then said the thing he believed most firmly. The platform had data about the mills, he said: capacity, defect rates, prices. It did not have what comes from walking through a mill at dawn and watching how the floor manager speaks to the workers. It could not know which owner will keep a promise in a crisis. “That knowledge,” he said, “is not in any database.”

The engineer from Chennai asked him how he knew it. He said he would tell them at the end, with a question.

27.5 The Question of Value

The slide on valuation put the company at 18 million US dollars.

Rajan looked at the number for a long time. He had spent nearly thirty years building a business that employs about 140 people and ships to fourteen countries. He suspected it would not fetch that figure. Two founders with a platform and a pilot were asking to be valued above what he had built in a working life.

He was not bitter, and he did not need to be. A price is what someone will pay today for a guess about tomorrow. It passes no verdict on a life’s work and says nothing about what anyone is worth. This book keeps those words apart. What a thing costs, what it sells for and what it is good for are different things, and they can move in different directions; a wage is a fourth. What a person is worth is none of them. Nor were the two prices alike. A buyer would price his business mostly on what it earns now. The start-up was priced on what it might earn if it changed a trade. The price of a share also depends on the terms that come with it, such as who is paid back first if the company is sold. A good company bought at too high a price can still be a bad investment.

There was something else in the number. A valuation of a company that matches buyers to mills is, among other things, a forecast. It guesses how many buyers will find other mills through software instead of ordering from makers like Rajan. Part of the 18 million was a bet that Rajan would be chosen less. He was being asked to judge a price that contained a guess about his own trade, and he could not read it as a stranger would.

What troubled him was what the price rested on. Keynes described how markets value what nobody can know. They follow a convention, and its essence, he wrote, lies “in assuming that the existing state of affairs will continue indefinitely, except in so far as we have specific reasons to expect a change”.7 A smooth curve on a slide is that convention drawn as a line. This quarter’s growth is carried forward to the end of the page.

Rajan traced the number back. The valuation rested on the number of customers. That rested on the belief that mills and buyers would adopt the platform because it saved time. That rested on the model’s performance, and the model belonged to someone else. At the bottom was a question that no slide could settle. Would real people, in a real trade, with their habits, loyalties and doubts, change how they worked because a platform told them to? Each step was a place where the number could break. An honest slide would have shown what happened to the price if any one of them did.

Rajan’s answer was no, and he knew why he wanted it to be. He also had a record against him. For years he had doubted Senthil, the owner of the mill he now trusted most, on an instinct he never examined, until his own records showed him wrong. This book has said before that stubbornness is not the same as discernment: it tells a man that he is still here, so his ways must be right. Rajan was watching himself resist. He was asking whether the resistance was wisdom or habit. The print edition called this moment “consciousness in operation”, and this edition withdraws the phrase. Consciousness, in this book, is experience known as yours; checking your own thinking is something a person does, and machines now do a version of it too. What Rajan had was a question about his own doubt. Nothing in the room could settle it.

27.6 Senthil’s Mill

The meeting had lasted two hours when Rajan asked the question he had come to ask.

“Does your platform recommend Senthil’s mill?”

The founders looked at each other. They had never heard of Senthil’s mill. They did not know Senthil.

Rajan explained. Senthil runs a processing mill in Tiruppur, which dyes and finishes cloth. His father first supplied Rajan in 1998, when Rajan was twenty-six. Senthil went to college, took over the business, and spent his money on temperature controls, a quality engineer and written procedures. In February 2025 a lot from his mill failed a wash test that Rajan had asked for. Senthil phoned Rajan before Rajan could phone him, explained what had gone wrong, and re-ran the whole order at his own cost.

“Your platform measures what can be measured,” Rajan said. “The things that make Senthil the best supplier I have ever worked with cannot be measured.”

The founder from Zurich asked one question. “Do you keep his delivery records?”

Rajan did. In early 2025 Karthik had run ten years of them through a tool that scored suppliers by late deliveries, rejected lots and complaints. Senthil’s mill had come out at the top. It had been the most reliable mill on Rajan’s books for a decade, including the years in which Rajan doubted him. The records had judged Senthil better than Rajan’s feeling had. Karthik, sitting beside him, did not say so. He did not need to.

So Rajan’s claim did not survive his own story. Much of what made Senthil good could be measured, and had been. The founders had never heard of the mill for a plainer reason. Its record was in Rajan’s books and in Senthil’s files, and nobody had asked Senthil whether it could go anywhere else. A platform holding those records would have put Senthil’s mill at or near the top. Rajan’s dawn walks had misjudged Senthil for years. The numbers had not. They had limits of their own. They held only the orders Rajan had placed, and a mill that takes only easy orders can look as good on paper as one that takes hard ones. A ranking also has to say best for what: the right mill for a plain order may be the wrong one for a difficult dye.

Something remained that the records could not hold, and the February lot showed what it was. The records showed that Senthil’s lots were rarely late or rejected. They did not decide to phone Rajan first and explain a fault that the buyer might never have found, and they did not pay for the re-run. That was an act, done by a man who could be found and asked why. A platform can record that such acts happened, and it can do acts of its own. It cannot make them Senthil’s.

Then one of the founders said: “We know.”

The founder from Zurich described the pilot. For three months the mills had kept overruling the platform’s recommendations, and for three months the team had treated this as noise: owners clinging to habit against better data. Then they had sat down with the overrides one by one. Some were knowledge the platform lacked: a festival week, a mill whose power failed every afternoon in May, a dye house that was short of water. Some were kinship and habit: an order sent to a cousin’s mill at a higher price and a later date. On those, the founders said, the platform had been right. The founders did not yet know how to tell the two kinds apart without asking each owner, and asking took time they had not budgeted.

The sorting mattered more than either half. An override that came from knowledge could become a record: a festival calendar, a log of afternoon power cuts, a note that a dye house runs short of water in May. Kept, such a record would let the platform learn what the owners knew, and would let a newcomer see it too. So Rajan’s “not in any database” failed a second time: some of what he meant had simply never been written down. An override that came from kinship looked like the habit the platform existed to question, and often it was. But an order to a cousin can repay help in a bad year, or buy credit that no bank would give. No delivery record shows that. As the founders had built it in 2026, the platform could not tell the two apart by itself. Only an owner could say why he had sent the order to his cousin, and some owners would not say. Asking each of them was slow, somebody had to pay for the slowness, and the founders’ plan had not said who.

The Tirukkuṟaḷ has a couplet, written about men tested for a king’s trust, that I would set beside the first thing the founders admitted: “Even men of rare learning and of pure hearts will not be found, when thou dost test them, to be absolutely exempt from all ignorance.”8 That the founders did not know something was ordinary. What could be judged was what they had done with it. For three months they had treated the people who disagreed with their machine as an error. Then they had changed their minds and gone to find out. That was conduct, the only kind a young company has, and it pointed both ways.

“We do not know how to encode trust into software,” the founder said. “That is why we need someone like you.”

The print edition took this sentence as the founders’ wisdom and ended its section on a lesson: that the pitches which earn trust admit what they do not know, and that “a vision grounded in honest self-assessment is harder to replicate.” Both halves are withdrawn. The second makes candour a moat, which this book has refused. The first turns candour into a technique. Once founders are taught that admitting ignorance wins investors, the admission starts to tell an investor less. This is the rule that economists call Goodhart’s law, in the anthropologist Marilyn Strathern’s words: “When a measure becomes a target, it ceases to be a good measure.”9 An honest sentence in a pitch is worth most while nobody is saying it to win. What is harder to fake is the practice behind the sentence: whether people changed what they did when the evidence changed. These founders said they had done so with the overrides, and an investor can check that by asking the mills.

What the founders needed was more ordinary than trust encoded: Rajan’s name with people who had never been asked. The mills the platform would rank had no seat at that table. They would be scored, recommended or passed over by a company they had not chosen, for buyers they might never meet. If buyers came to expect every mill to be on the platform, a mill would be free to leave in law but not in practice. An investor might count that as a moat. The mills were among the people who pay for a product without using it or being asked. A record that judges such people owes them a way to see what it says about them and to reply. None of that was in the deck.

The Tirukkuṟaḷ’s chapter has a harder couplet, and it cuts both ways. “Beware of trusting men that have no kindred: for their hearts will be without attachment and they will be callous to shame.”10 The couplet is about choosing confidants. Read as a rule of trade, which is my application, it is the old rule of trading families. It trusts those whose families can be shamed, and it is a rule that has kept newcomers with clean records and no father in the trade waiting outside for generations. It is also exactly what the platform was built to get past: a way for strangers to trade without kin. The founder from Zurich had no kindred in Tiruppur. That was the point of his company.

27.7 An Act of Judgement

The fund’s committee would decide the fund’s money, at its own meeting, without Rajan. That was right. He had been asked for a view; the vote was theirs. What he could decide was his own money and his own name.

He invested a modest sum, one he could lose without changing anything at home. He agreed to introduce the founders to mills, beginning with the curious owners and working towards the sceptical ones, because trust cannot be transferred in bulk. That evening, from his hotel, he wrote down three conditions and sent them to the founders, with Karthik and the partner copied. He would introduce a mill only after its owner had been told what the platform would record about it and who could read it. Each mill would be able to see its own entry and write a reply beneath it. And every introduction would say that Rajan had put money into the company.

The founders accepted the first two for the mills he introduced, and the engineer from Chennai said they would cost months. She was right, and the months were theirs to pay. Nobody accepted anything for the mills Rajan would never meet.

The partner replied that night in two lines. The fund would not make the conditions a term of its own investment, because they would slow growth, and growth was what the fund’s own investors had paid for. He did not say the conditions were wrong. He said they were Rajan’s to pay for. That was fair as far as it went. It also showed where the fund’s answering stopped: at the money it put in, and the people it owed a return. The fund’s terms set what it asked in return for its money. They did not settle what the company owed the mills.

A fund can ask for more than growth. It can ask whether the founders change course when they are wrong, what would make the mills stay, and how a mill can see its entry and reply to it. It must then accept slower growth, and this fund’s investors had not paid for that. What the fund would owe the founders, if it invested, was narrow: the money on the stated terms and any help it had offered, not a rescue. If growth stalled, it would owe them an early, plain word on whether it would put in more money, so that they could look elsewhere in time.

The third condition mattered most to Rajan. An introduction is a kind of testimony. When he tells a mill owner that these founders are worth an hour, the owner hears a man he has known for years vouching for strangers. Once Rajan owned part of the company, his vouching was no longer free of interest. He could not make it free. He could only say so in the sentence that carried it. Saying so does not cancel an interest. In experiments, people who were told of an adviser’s conflict of interest still did not discount his advice enough, and disclosing it could make the advice more biased. Later experiments found that being told could lower trust in the advice and still raise the pressure to follow it, when people felt obliged to serve the adviser’s interest.11 An introduction can also be heard as more than it says. A mill owner may take “worth an hour” to mean “I have checked them”, and Rajan had not checked everything. As I read it, the third condition could not do its work alone. The first two gave an owner something to weigh besides Rajan’s word: what the platform would record about his mill, and a way to read it and reply.

His conditions gave a mill a way to reply, and nothing more. A reply beneath a wrong entry explains, and pays for none of the orders a mill lost while the entry was wrong. Nobody in the room offered that, and Rajan did not ask for it.

Keynes noticed something about investors that bears on Rajan’s case. A share in a listed company can be sold tomorrow. “Investments which are ‘fixed’ for the community are thus made ‘liquid’ for the individual.”7 The ease of selling worried him. “The spectacle of modern investment markets has sometimes moved me towards the conclusion that to make the purchase of an investment permanent and indissoluble, like marriage, except by reason of death or other grave cause, might be a useful remedy for our contemporary evils.”11 He did not recommend it. In the next sentences he saw the dilemma: if investors could not sell, many might keep their savings in money instead, and new investment would suffer. But the wish says something. A man who can leave at any time does not have to look far ahead.

Rajan’s stake was different. A small share in a young private company is hard to sell, since such a company restricts the transfer of its shares.12 And an introduction cannot be taken back once it is made. He was bound in a way that Keynes half wished investors were. Being bound is not the same as behaving well: an owner who cannot sell can still neglect what he owns. And his small stake bought him no seat on the board and little say. Once the money was in, his main hold on the company was the introductions he had not yet made. A fund’s portfolio spreads its money across many companies; nothing spreads a name. His shares were fully paid, so if the company failed, the law would limit his loss to what he had paid.12 Nothing limited what a mill would lose if a ranking it could not see sent its buyers elsewhere.

If the platform ever ranked a mill wrongly, many hands would have helped. The founders would have designed the ranking, and the model maker’s machine would have read the records. The fund would have paid for speed, and a buyer would have acted on the result. And one man would have made the introduction. Each would have a different part to explain and a different part to put right. Rajan’s part was small, and it was his own: the trust he had lent.

Here the two voices pull together, and then apart. Both want the one who trusts to be bound. The couplet distrusts the man with no kindred, because nothing attaches him and he has no shame to lose; read as a rule of trade, it trusts kin. Keynes regretted that markets let an investor stay unattached. They part on where the binding should come from. The couplet finds it in family, and so it taxes the stranger. Keynes would find it in commitment, and saw that commitment imposed by law might stop people investing. This book’s answer is narrower than either. What binds is a named person who can be found when it goes wrong, who must explain, and who must put right what can be put right.

So the price of Rajan’s decision is easy to list. A small sum. His hours, which nobody was paying for. His name with every owner he introduced. And his help to a company that might one day send some of his buyers to other mills. His introductions would also help the forecast he doubted to come true. Money and a trusted name can change the outcome they are betting on. A platform had already taken part of Weiss’s yarn-dyed checks from him. Whether the decision was wise belongs to the future. If the company does well, that will not show the decision was good, and if it fails, that will not show it was bad. A judgement is weighed by what was known when it was made.

At Changi, before boarding, he called Karthik.

“Did their numbers support it?” Karthik asked.

“Not entirely.”

“Then what did?”

“They knew what they were missing,” Rajan said. “That is rare.”

It is a good sentence, and one meeting is not enough to stand behind it. He had heard the founders say what they did not know, once, in a room where saying it helped them. The couplet’s test is conduct over time, and he had heard of three months of it, from the founders themselves.

Ending: Three Lines Not Yet Sent

Late in September he drove to Tiruppur and found Senthil in his laboratory, between the wash-test machine and a board of swatches pinned in rows. Steam came in from the dyeing shed across the yard.

Rajan told him what he had done: the deck, the flight, the money, the conditions. Senthil listened with his arms folded. Then he asked to see what the platform would write about his mill before anyone else saw it.

“That is one of the conditions,” Rajan said.

“Then they can show it to me here,” Senthil said. “Not in Singapore. I will read it against my own records. They are better than yours.”

Rajan did not argue. Karthik’s scoring had used Rajan’s books, and Senthil’s would have been better.

That evening, in his office above the floor, Rajan wrote the introduction. It was three lines long. The first said who Senthil was, and the second who the founders were. The third said that Rajan had put his own money into their company. He did not send it. He sent the founders one line instead, asking what the platform would write about Senthil’s mill. The introduction would wait until Senthil had read the answer and said it could go.

Senthil had not yet said. If it went, and the platform ever got Senthil wrong, the first call would not go to Chennai or Zurich. It would come to Coimbatore.

Practice

Before you lend your name. When someone asks you to invest in, recommend or introduce a person or a company, write six lines before you reply. First, what you stand to gain or lose if they succeed, including as a rival, a relative or an employee. Second, which of your reasons rest on things you have seen them do, and which on their manner, their polish or their family. Third, if it is a company, what evidence you have that buyers want the thing, that it pays its way and that the price is fair, keeping what you have seen apart from what you were told. Fourth, what they depend on that they do not control. Fifth, whom your name will bring in with it, what they will take your introduction to mean, and whether they have been told what they are being brought into. Sixth, what you will do if it goes wrong, and who will call you. Send the first line to the person who asked. This costs an hour, and it may cost you the deal. It will not make the bet a good one, and it promises nothing about money. It makes what you are lending visible to the people you lend it to.

Words to keep

  • animal spirits (Keynes): the urge to act rather than wait, on which, Keynes thought, most decisions with long consequences rest.
  • convention (Keynes): the market’s habit of valuing an investment by assuming that the present will continue unless there is a specific reason to expect a change.
  • deck: the slides a company shows to investors.
  • Goodhart’s law: “When a measure becomes a target, it ceases to be a good measure” (Strathern’s wording).
  • karumam: in Tamil, a person’s deeds or conduct.
  • liquid: easy to sell for money; a listed share is liquid, a share in a young private company is not.
  • moat: in business talk, whatever makes it hard for rivals to take a company’s customers.
  • pilot: a small trial of a product with real users, run before it is offered widely.
  • Sherlocked: developers’ word for having your product’s work built into the platform it runs on.
  • Tirukkuṟaḷ: a classical Tamil work of short couplets on virtue, wealth and love, traditionally attributed to Tiruvalluvar.
  • valuation: the price at which investors agree to buy part of a company, which implies a price for the whole; what it means also depends on the terms that come with the shares.

Sources

  1. John Maynard Keynes, The General Theory of Employment, Interest and Money (London: Macmillan, 1936), chapter 12, “The State of Long-Term Expectation”, section III (Marxists Internet Archive text). The next sentence: “In fact, those who seriously attempt to make any such estimate are often so much in the minority that their behaviour does not govern the market.” Cited by chapter and section; the 1936 page numbers were not checked against a scan.
  2. Keynes, General Theory, chapter 12, section V: “anticipating what average opinion expects the average opinion to be”. The passage is about “professional investment”, which Keynes likens to a newspaper competition in picking the faces others will pick. Section VII guards against the opposite reading: “We should not conclude from this that everything depends on waves of irrational psychology”; Keynes speaks of “our rational selves choosing between the alternatives as best we are able, calculating where we can”.
  3. Tirukkuṟaḷ 505, chapter 51, “Testing of Men for Confidence”, trans. V. V. S. Aiyar, The Kural or the Maxims of Tiruvalluvar (Madras, 1916; Standard Ebooks). Tamil: perumaikkum ēṉaic ciṟumaikkum tattam karumamē kaṭṭaḷaik kal.
  4. Tirukkuṟaḷ 509 and 510, the same chapter and translation, in paraphrase: 509 asks for men to be tried before they are trusted; 510 warns that trusting the untried and suspecting the tried lead alike to harm.
  5. Keynes, General Theory, chapter 12, section VII: “a result of animal spirits” and “a spontaneous urge to action rather than inaction”. The original sets the two phrases off with a dash; they are quoted separately. For The other side in 27.2 (added in this edition): Steven N. Kaplan, Berk A. Sensoy and Per Strömberg, “Should Investors Bet on the Jockey or the Horse? Evidence from the Evolution of Firms from Early Business Plans to Public Companies,” Journal of Finance 64, no. 1 (2009): 75–115: 50 companies financed by venture capital; “Firm business lines remain remarkably stable while management turnover is substantial” (abstract); “only 44% of the CEOs at the annual report were CEOs at the business plan” (p. 78; the annual report is the one closest to 36 months after the listing); “at the margin, investors in start-ups should place more weight on the business (‘the horse’) than on the management team (‘the jockey’)” (abstract); “the results do not imply that good management is not important” (p. 79).
  6. Bobby Allyn, “Apple just made your app obsolete? You’ve been ‘Sherlocked’,” NPR, 17 June 2024 (republished by OPB): the practice “has become known as ‘Sherlocking’”; Apple’s search tool Sherlock was “strikingly similar” to a program called Watson. NPR dates this loosely to “the 1990s”: Watson was released in 2001, and Sherlock 3, which overlapped it, in 2002 (Karelia Software, “Watson FAQ”, karelia.com).
  7. Keynes, General Theory, chapter 12, section IV. The sentence opens “The essence of this convention” and sets a qualifying clause off with dashes; only its end is quoted. The same section has: “Investments which are ‘fixed’ for the community are thus made ‘liquid’ for the individual.”
  8. Tirukkuṟaḷ 503, the same chapter and translation. Tamil: ariyakaṟṟu ācaṟṟār kaṇṇum teriyuṅkāl iṉmai aritē veḷiṟu.
  9. Marilyn Strathern, “‘Improving ratings’: audit in the British University system,” European Review 5, no. 3 (1997): 305–321, at 308. Goodhart’s original observation: C. A. E. Goodhart, “Problems of Monetary Management: The U.K. Experience” (1975).
  10. Tirukkuṟaḷ 506, the same chapter and translation. Tamil: aṟṟārait tēṟutal ōmpuka maṟṟavar paṟṟilar nāṇār paḻi.
  11. Keynes, General Theory, chapter 12, section VI. The sentences that follow describe the dilemma: “If individual purchases of investments were rendered illiquid, this might seriously impede new investment”. The same mark, earlier in 27.7, cites two studies of disclosure. Daylian M. Cain, George Loewenstein and Don A. Moore, “The Dirt on Coming Clean: Perverse Effects of Disclosing Conflicts of Interest,” Journal of Legal Studies 34, no. 1 (2005): 1–25, abstract: people “do not discount advice from biased advisors as much as they should, even when advisors’ conflicts of interest are disclosed”, and disclosure “can increase the bias in advice”. Sunita Sah, George Loewenstein and Daylian M. Cain, “The Burden of Disclosure: Increased Compliance with Distrusted Advice,” Journal of Personality and Social Psychology 104, no. 2 (2013): 289–304, abstract (six experiments): disclosure “can decrease advisees’ trust in the advice”, but “can also increase pressure to comply with that advice if advisees feel obliged to satisfy their advisors’ personal interests”.
  12. Companies Act, 2013 (India; Act No. 18 of 2013), s. 2(22): a “company limited by shares” is one “having the liability of its members limited by the memorandum to the amount, if any, unpaid on the shares respectively held by them”. Section 2(68): a private company is one which, “by its articles”, “restricts the right to transfer its shares” (sub-clause (i)).

Named people in this chapter are composites, and their conversations and circumstances are constructed; the studies, laws and figures cited are real and sourced.

Revised from the print edition (2026)

  • What the print edition claimed: that the defences that matter most in an AI company “are assessments of consciousness. No screening tool has learned to exercise them”; that Rajan’s doubt about his own doubt was “consciousness in operation”; that the gap between “we have built the future” and an honest admission “is the gap between confidence and consciousness”; that what makes Senthil good “cannot be measured” and that by every measurable criterion his mill would “rank in the middle”; that “trust is not a data point”; that the pitches that earn trust teach “the counterintuitive lesson that the best investors have learned”, and that “Products can be copied. A vision grounded in honest self-assessment is harder to replicate”; that relationships and expertise give a company “a useful life measured in decades”; that a Tokyo buyer’s disappointment, which “no sensor could detect”, corrected Rajan’s scheduler; that Karthik told Rajan he had found “three problems that my entire team missed”; that Rajan’s recovery came through a son, Vikram, who does not exist in this book; that Rajan had been in business “twenty-seven years”; that Karthik called from “an unfamiliar Singapore number”; that the start-up carried a name that belongs to a real company in the same trade; that Rajan was reviewing papers for a consignment to Copenhagen; and that he telephoned home in flight over the Strait of Malacca. It also ended its sections on lessons for founders.
  • What I now hold: an investor can judge conduct, what people do when they are wrong and what they did when the world contradicted them, and not consciousness. Conduct is not the whole decision: whether buyers want the thing, whether it pays its way and whether the price is fair are separate questions, and a good company bought at too high a price can be a bad investment. Whether a young company will last is mostly unknowable (Keynes), and a smooth curve is the convention that the present will continue; that is a reason to show the assumptions, not to skip them. A model built from the records of hundreds of companies can say how companies like this one have usually ended, and may find most of a reader’s objections first; a rate for the kind is not a forecast for the case, and the model carries its makers’ choice of what counts as a good outcome. An interested judge is not a worthless one, but he must say his interest, and saying it does not cancel it. What made Senthil good was measured, and his records judged him better than Rajan’s feeling did; records still have limits (only the orders placed; best for what?), and an order to kin can carry knowledge they do not hold. A platform that ranks mills owes them a way to see and reply, and a reply is not a remedy. A fund owes the founders it backs its stated terms and an early word on further money, not a rescue. Candour used to win tells an investor less, and the practice behind it, which can be checked, is harder to fake. A portfolio spreads money; nothing spreads a name. The investor’s loss on fully paid shares is limited by law; the loss of those he brings in is not.
  • What changed my mind: Keynes’s chapter on long-term expectation, which puts a textile factory on its list of things nobody can value ten years ahead; the Tirukkuṟaḷ’s chapter on testing people, read with its own warning about kin; this book’s earlier finding that Rajan’s records judged Senthil better than he did; the rule that a machine’s ability is dated and that consciousness is never the hinge; the Part IV finding that a product is judged by what it does to those who pay for it without being asked; Goodhart’s law applied to honesty; and an outside reading of an earlier version of this chapter, in October 2026, which pointed out that conduct is not the whole of an investment’s merit, that uncertainty does not excuse weak estimates, and that Goodhart’s law weakens a signal without erasing it.
  • What survives: the deck that is “smooth”; Rajan’s three questions, “Is this person serious? Will this last? What happens when things go wrong?”, now read as a test of conduct, not of durability; the three phone calls and “They have mapped the companies that do not matter and ignored the people who do”; “the growth projections assumed mill owners would adopt the platform faster than most of them had adopted mobile banking”; “If your model provider decides to build what you have built, what do you have left?” and “The wrong answer is ‘that will not happen’”; the question “Does your platform recommend Senthil’s mill?”, to which the founders still have no answer, for a different reason; “We know”; “trust cannot be transferred in bulk”; Rajan’s modest investment; and “They knew what they were missing. That is rare.”, now with its limit. The company stays unnamed.

↑ Back to top

Discuss

Loading the discussion…

More in Artificial Intelligence
  •  ·  26 min read

    Epilogue

    Late in the Gītā, the one who knows most says: ponder it thoroughly, then act as you like. A machine can now set out the reasons on every side. Saying “I answer for this” then is not pride. It is a reply, for your part, to those owed one.

  •  ·  37 min read

    Chapter 40: What This Book Cannot Tell You

    The print edition’s last chapter reduced the book to one sentence, and rested human worth on what machines cannot do. I withdraw both. A book can date its claims and correct them in public. It cannot sit in the room where you decide.

  •  ·  6 min read

    Part IX: The Unfinished Chapter

    The print edition’s last Part said that consciousness never concludes. This one is a single chapter that turns the book on itself: what a book can stand behind, what it owes those who read the print edition, and what it leaves open.