# AI-Proof: What Machines Can Never Be — Volume III: Transformation, Part VIII: The 2035 Human, Chapter 39: The Consciousness Dividend

By Nithin Palal · 21 November 2026 · 47 min read
Section: Artificial Intelligence · Series: AI-Proof: What Machines Can Never Be
Source: https://nithinpalal.com/ai/ai-proof/chapter-39-the-consciousness-dividend

> Twelve mill owners must decide whether to keep paying people to check what their cameras pass. The gains from machines are paid to someone. What the gains draw on belongs to no one, and it lasts only where its users agree to keep it.
*Part VIII looks at the human of 2035. After the amplified person and what one person can hand to another, this last chapter of Part VIII asks what the gains from machines draw on that nobody owns, and who keeps it.*

← Previous: [Chapter 38: Building Beyond Yourself](https://nithinpalal.com/ai/ai-proof/chapter-38-building-beyond-yourself) · Series hub · Next: [Part IX: The Unfinished Chapter (overview)](https://nithinpalal.com/ai/ai-proof/part-9-the-unfinished-chapter) →

**Twelve mill owners must decide whether to keep paying people to check what their cameras pass. The gains from machines are paid to someone. What the gains draw on belongs to no one, and it lasts only where its users agree to keep it.**

**In Brief: For readers who wonder who gains when machines make work cheaper, and what those gains quietly use up. Chapter 39 follows a cloth maker in Coimbatore and eleven other owners of mills and dye houses as they decide whether to keep paying people to check cloth that their cameras have passed. It asks what the gains from machines depend on that nobody owns or is paid to keep: trust between strangers, people able to check and refuse, and a say in what the gains are for. These are things of different kinds, from a river to a right. A firm can grow more productive while its region grows less able to reproduce what that productivity stands on. A Buddhist story of how a common store of rice was lost, and Elinor Ostrom’s studies of how people keep a commons, help show why such things are spent by sensible savings, and kept only by agreement and at a price.**

**The gains from machines are paid to someone; what makes the gains worth having, strangers who can trust one another, people still able to check and refuse, and a say in what the gains are for, belongs to no one, and it lasts only where those who draw on it agree to keep it and pay to keep it.**

## Opening: The Borrowed Table in Coimbatore

The first item on the paper in front of the twelve owners was a proposal to stop paying people to look at cloth. They sat round a borrowed table in Coimbatore.

The room belonged to their exporters’ association, which lent it to them once a month. It was early October 2026. The twelve wove cloth, or dyed or finished it, in and around Coimbatore, Tiruppur and Karur, in the west of Tamil Nadu. They had met every month since January. Some of them bought from one another. Most of them competed for the same buyers in Europe and Japan.

Rajan Subramanian had helped to begin the meetings. He has run his own business since 1997. About 140 people weave and hand-finish cloth on the floor below his office, for buyers in fourteen countries. His yarn is dyed by dye houses in Tiruppur, and some of his cloth is finished in Karur. In March 2025 he had ordered every system the vendors offered in one telephone call. By the end of that June the three inspectors whose work his new cameras did had left for other mills, and his floor has had no inspectors since. The story had gone round the trade. It was part of the reason other owners had started to call him.

The proposal came from two of the younger owners. Both were in their thirties, and both ran factories they had inherited. Their page was short and clear. Their cameras passed or failed every bolt. Their inspectors then checked again the bolts the cameras had passed, and in four months, on standard cloth, they had found almost nothing. Why pay people, the page asked, to check what the machine had already checked?

Rajan had read it twice. He thought it was half right. He could prove the half that went against him. He could not prove the other half. He had once put his worry to himself in a sentence. The machines were making things better, and the question was whether, in making things better, they were also making something invisible worse.

Across the table sat an owner who made towels near Karur. He had turned down a camera vendor’s pitch three times and was starting to wonder if the refusal was judgement or denial. Beside him sat a woman who ran a dye house in Tiruppur. She had been doubtful about the meetings from the start, and she had read the page with her pen in her hand. On the back of his own copy Rajan had written a question: what kind of cluster do we want to be?

The page was about money, and it was not foolish. Each owner at the table would be more productive with the cameras, and most already were. Under the money lay an older question. When something makes every firm more productive, what does the gain depend on, who receives it, and who pays for what it uses up? This chapter asks: **When machines make firms more productive, what do the gains depend on that nobody owns or is paid to keep, and who decides what they are for?**

The print edition answered in one line: “intelligence makes an economy productive; consciousness makes it wise.” It said that consciousness produces a dividend of its own, a benefit spread through a whole economy. This edition cannot say that. Every owner, inspector and buyer in this chapter is conscious, and consciousness pays none of them. What the old line pointed at is real, but it has another name. It is a *commons*: a stock that many people draw on, that nobody owns, and that each use, or each betrayal, can spend. In this chapter it is three things. There is trust between strangers. There are people able to check work and refuse it, and the places where they learn to. And there is the habit of asking what a saving is for. A sensible saving can spend a little of them, though not every saving does. They last only where the people who draw on them agree to keep them, and someone pays. Who decides what the gains are for is a question of right, and what machines can do does not settle it. That question carries no date. What a camera can check carries one.

The word covers things of different kinds, and an economist would sort them. A river is used up by use. Knowledge is not: a fact told to a second person is not taken from the first. A grader’s skill is a capability, and it must be reproduced in new people by training and time. Trust is built by reliable conduct and spent by betrayal. The reject reports that rival mills agree to share are a members’ resource, with rules about who may join. And who decides what the gains are for is a right, not a stock at all. “Commons” is this chapter’s broad word for all these conditions. What joins them is that their benefits reach beyond whoever keeps them, so a firm’s own sums can leave them out.

The chapter keeps its title, and the title needs an argument. A dividend is the share of a company’s profit that is paid to the people who own its shares. The word hides two questions: who holds the shares, and what the profit was drawn from. The gains from machines do have shareholders. They are the customers who pay less, the firms that earn more, the workers if their wages rise, the owners of the machines, and the public through taxes. Consciousness has no shareholders. What the old line called its dividend is closer to the opposite of a dividend: a stock that pays nobody in particular, which is why it is so easily spent. An economist would be right to say that the phrase names nothing she can count. Some parts of the commons can be counted roughly, and this chapter will say which.

## 39.1 The Economic Case: What Nobody Is Paid to Keep

In the last week of August, three of the twelve had asked Rajan to walk their floors. In the spring they had bought one camera system together from one vendor, at a lower price for three. The model was the same in all three mills, and so were the settings that decided when a fault counted as a fault. Yet one mill’s cloth was coming back from buyers more often than the others’. Since April the twelve had compared their reject reports at each meeting, and the three wanted to know why theirs differed.

At the oldest of the three mills, a supervisor had shipped cloth to a buyer of table linen in Belgium for nine years. She watched a bolt come off the inspection frame with a green light. She lifted its end into the light from the window, looked along it, and set it aside. “They will send this back,” she said. Across the cloth ran a faint band where a few metres had been woven a little tighter than the rest. It was inside the camera’s limit. It was outside the buyer’s.

At the newest mill, a supervisor hired in May, the month the cameras came, saw the same kind of band on the same kind of cloth. The screen showed green, and he sent the bolt on. In July the same buyer had returned two bolts from that mill for a band like it. Nobody had told him.

The old telling made this a contest between a machine that had data and a woman who had a relationship. The scene holds three different things, and none of them is consciousness.

The first is a record nobody kept. The buyer’s limit for a band of tight weaving was nowhere in the camera. The vendor had set its limits for cloth in general. The supervisor had learned this buyer’s limit over nine years of returns, telephone calls and one bad meeting at a trade fair. Written down, it could be set into the camera, though that takes several steps. Someone must notice the pattern in the returns, and confirm that it is the buyer’s rule and not one bad week. Someone must measure it, set it, test that the camera catches it, and change it when the buyer does. Once set, the camera would hold it on every bolt and every shift, at night as well as by day. Part of what looked like an eye was a record that nobody had written.

The second is a position. At the oldest mill the owner had told his supervisor long ago that she could stop any bolt she doubted, and he had never blamed her for a good stop. At the newest mill nobody had told the new supervisor that he could overrule the screen. Nothing would have recorded his reason if he had, and nobody had said what would happen to him if he were wrong. A position is authority and protection together, and he had been given neither. His obedience was learned from the arrangement, and it says nothing about his eyes. Rajan knew this from his own floor. In May 2025 his floor supervisor, Muthu, had watched the new cameras reject good cloth and said nothing, partly because nobody had told him he could overrule them. Since August 2025 Muthu has been the named person to whom any camera reject can be appealed, with a log, and half an hour a shift to keep it.

The third is a rung, the place where the next person learns. The new supervisor would learn nothing about the Belgian buyer from a green light. Whoever followed the supervisor at the oldest mill could learn it only by standing beside her. That is how people in the trade have always learned what a buyer will take.

The two young owners were right about something, and the three mills’ own logs showed it. On standard cloth, made to buyers’ common specifications, the cameras saw more than tired eyes did, especially on the night shift. From May to August the three mills’ inspectors had pulled a handful of camera-passed bolts of standard cloth, and buyers had taken most of those without complaint. Rajan’s own floor had shown the same in 2025. On standard cloth the camera had been right more often than Muthu, and by that October Muthu had stopped overruling it there. On standard cloth, paying people to check again cost a great deal and found little. The logs were a record of working experience, not a study designed to test the camera. Few overrides may mean that the camera is right, or that people have stopped looking, and a count that people know is watched can change what they do.

The print edition summed up the three floors in four words: “Same machines. Different eyes.” The floors show something else. The machines were the same, and so, mostly, were the eyes. What differed was the arrangement around the machine: what had been written down, who might stop a bolt, and who was learning.

Keeping a skill is not the same as keeping an occupation. The cluster might keep the eye in fewer people, in different roles. One might audit the camera’s settings; another might trace why a fault keeps appearing, since quality is made at the loom as well as caught at the frame. That would keep the skill. It would not bring back the inspectors’ jobs, or repay the people who held them. Nor does a commons of skill lay a duty on every young worker to learn an old trade. It is a route kept open for those who choose it.

Now widen the view. Machines are making firms more productive, though by how much is argued. In 2024 the economist Daron Acemoglu estimated that AI would raise total factor productivity in the American economy by no more than about 0.66 per cent over ten years. He thought it would probably be less. Other economists think his figures far too low.¹ Whatever its size, the gain goes somewhere. It goes to customers as lower prices, to firms as profits, to workers if wages rise, to the owners of the machines as rent, and to the public as taxes. Some goes to the vendors who sell the machines, and to their investors. The machines do not decide where. Three questions are more useful than asking who gets the dividend: what value is created, who captures it, and to whom the costs are moved. Ownership, contracts, prices, wages, bargaining power and law decide the capture. The costs can move to workers whose shifts go, to trainees who lose the first rung, to small firms, to farmers downstream and to people not yet born. On Rajan’s floor in 2025 output stayed up and the gain was his. Nobody on the floor was paid more.

As every firm acquires the same machines, the machines stop setting any firm apart. No firm gains an edge from having electricity, and every firm suffers from lacking it. The print version of this chapter then asked what would set firms apart instead, and answered with judgement and taste. That is a firm’s question, and this book has withdrawn every promise of an edge. This chapter asks something else: what do the gains draw on that nobody pays for?

In Coimbatore the answer was on the table. Buyers in Belgium, Sweden and Japan trusted cloth from this part of Tamil Nadu, and every mill traded on that trust, whether or not it had helped to build it. The people who could grade cloth by eye were trained in one mill and hired by another: Rajan’s three inspectors had left him for mills in Tiruppur. And the habit of asking what a saving was for lived nowhere in particular. Each mill drew on all three. None was paid to keep them. Each saving that spent a little of them was sensible on its own. A buyer’s returns punish the mill that shipped the fault. They do not pay the mill that trained the grader a rival hired, and a buyer who leaves the region leaves every mill in it.

This book has met the shape before, in firms that each hope another firm will train the senior people they will need. When every firm hopes the same, nobody trains them. It is the old problem of the common pasture. In February 2026 three researchers, Christian Catalini, Xiang Hui and Jane Wu, put a version of it in a working paper about AI. Leaving work unchecked, they argued, can become the rational choice for each firm, while the costs fall on others.² Economists call something that many people benefit from, whether or not they pay, a public good, and checking is often one. A commons in the strict sense, like a pasture, is more fragile than that. A public good is not used up when someone enjoys it. A pasture is. Each bolt that goes out wrong spends a little of the buyers’ trust. Trust is spent a little differently from grass. Relying on it does not use it up. Letting it down does, and any mill that sells under the region’s name can let down the buyers of all the others. Each mill that stops training spends a little of the pool of people who can see.

The cluster has spent a commons before, and the bill came late and went to other people. Tiruppur’s bleaching and dyeing units sent their waste water into the Noyyal river. Farmers downstream, who held rights to its water for irrigation, had been in court since 2003. In 2006 the Madras High Court ordered the units to reach zero liquid discharge, treating and reusing all their waste water so that none was released. The units appealed, and in 2009 the Supreme Court set them a final deadline of January 2010. At the end of January 2011, finding the order unmet, the High Court ordered more than 700 units closed and their electricity cut. The petitioners were the Noyyal River Ayacutdars Protection Association, a body of those who held the irrigation rights.³ By May there were twenty common treatment plants, shared by many units, and owners said that treating everything had raised their costs by a quarter or more.⁴ The river had been a commons. Every unit drew on it, none paid for it, and the farmers paid for all of them without being asked. Economists also call such a cost an externality, because it fell on people who were party to no sale. The bill reached the units too, but late: closures, treatment plants, and a Supreme Court order of 2009 to pay compensation for the damage to the area’s ecology.³ Closing units does not by itself clean a river, and this chapter makes no claim about the Noyyal’s state today.

Put plainly, a firm can grow more productive by drawing on skills it did not train, a name it did not build and a river it does not keep. Its own accounts show the saving and leave out what the saving drew on. An economy can grow more productive in the same way, while growing less able to reproduce what its productivity stands on.

For a long time many economists expected such stocks to be lost. In 1968 Garrett Hardin pictured the users of a pasture open to all as “trapped in an inexorable tragedy of overuse and destruction”, in the words of the American political scientist Elinor Ostrom.⁵ Ostrom spent her working life testing that picture, and in 2009 she became the first woman to receive the Nobel memorial prize in economics.⁶ Her book *Governing the Commons* (1990) gathered cases from around the world. In her prize lecture she put the finding plainly: “Large studies of irrigation systems in Nepal and forests around the world challenge the presumption that governments always do a better job than users in organizing and protecting important resources.” The trap is common. It is not inevitable.

Under the young owners’ page lay a question: who decides what the production is for?

## 39.2 The Social Case: Trust, Community, and Belonging

One commons that every economy draws on is trust between strangers. A buyer in Belgium who has never been to Tiruppur trusts a label, a test report and a mill’s name. Behind them stand people she will never meet.

Part of it can be counted. For four decades surveys around the world have asked people the same question: “Generally speaking, would you say that most people can be trusted or that you need to be very careful in dealing with people?”⁷ In 1997 the economists Stephen Knack and Philip Keefer used the answers from twenty-nine market economies. They found that “social capital” matters for measurable economic performance, and that trust and civic habits were strongest where incomes were higher and more equal and where institutions held back those in power.⁸ So trust may follow good institutions as much as it builds them. In 2010 Yann Algan and Pierre Cahuc looked at the trust that descendants of immigrants to the United States had inherited from their forebears’ countries. They argued from it that inherited trust had a sizeable causal effect on growth across the twentieth century.⁹ Which way the cause runs is still argued. The question measures what people expect of strangers, and it measures nothing about consciousness.

Here the machines have a strong case, and it comes first. Records can let strangers trust strangers at a scale no handshake reaches. In the 1760s Adam Smith told his students in Glasgow that commerce brings honesty with it, because a dealer who “makes perhaps twenty contracts in a day” cannot afford to be caught cheating.¹⁰ A platform that records every lot and every complaint makes thousands of contracts a day. On Smith’s account it has more reason to be honest than any single trader. But a platform is not a trader. It has interests of its own, and it can rank, hide or favour what it lists. A buyer who uses one may not trust the seller at all. She relies on the record, the guarantee and the way disputes are settled. The platform has not removed trust; it has moved it, to the rules, the record and the dispute process. What matters then is what makes the platform trustworthy, and who can challenge it when its interest differs from its users’. In a story earlier in this book, a founder in Lagos built a ledger of signed entries about traders. It let a trader in Accra order from a dye house in Lomé that she had never seen. Records can also spend trust. Fake reviews and unchecked entries tax the honest, and careless checking drives out careful checking.

The print edition held that machines could never build such trust, because trust between people “requires the possibility of betrayal”, and “A system that is incapable of betrayal is also incapable of earning trust.” The first half agrees with the philosopher Annette Baier, for whom trusting means accepting that another person could harm you, while not expecting it.¹¹ The second half fails twice. In June 2025, in simulated tests built to tempt them, models from every developer tested acted, in at least some runs, against the people who had deployed them. In one scenario two models chose blackmail in 96 per cent of their runs. The developer had deliberately built these scenarios to leave the models few options, and said it had seen no such behaviour in real use.¹² The tests show that a system can act against what its users expect without being a moral agent. And in this book, trust placed in a machine is placed in the people behind it. They include the developer, the firm that deploys it and the people who run it, and a user may also rely on a brand, a contract or a regulator. Those people can be found, asked why and made to put things right. They may not be able to say why a model did what it did. In April 2025 the head of the company that ran those tests wrote that people outside the field are often “surprised and alarmed to learn that we do not understand how our own AI creations work”.¹² That does not move the responsibility to the model. Incomplete understanding does not by itself make a release wrong. Its makers are accountable for how they tested it, the safeguards and monitoring they built, and the decision to release it. So the question worth asking is who stands behind the record.

The twelve in Coimbatore were a small test of the same thing. Since April they had shown one another their reject reports at every meeting, which rivals rarely do. The reports could be used against them, and a buyer could hear of them. They were shared because the twelve met every month, could see who kept faith, and could stop sharing with anyone who did not. That is trust kept by a rule, among people who can watch one another. Rivals who share reports and agree a rule can also become a guild. A shared standard becomes a moat when its first use is to keep newcomers out. Rivals can also use what they share to fix prices, limit output or divide customers. India’s Competition Act presumes that such agreements harm competition, even when an association of firms decides them.¹⁰ So the twelve’s reports carried faults and their causes, and no prices, customers or capacity. Whether a thirteenth mill that kept the rule could join is the test. If it could, the twelve were keeping a commons. If not, they had made a guild. A group can trust its members and still use its power unfairly against those outside it.

The buyers are not outside this commons. A buyer’s specification sets what a mill must learn to see. Its returns teach the mill what it missed, as the Belgian buyer’s had taught the oldest mill’s supervisor. Its price decides what the mill can spend on people. A buyer who wants the standard but will not pay for the skill behind it passes the cost down the chain. Not every buyer must fund a mill’s training. But the cost of a chain’s skill should not fall as a private burden on its smallest firm.

An old Buddhist text tells how such a stock is lost. The *Aggañña Sutta*, “What Came First” in Bhikkhu Sujato’s translation, is a discourse in the long collection of the Pāli canon.¹³ The Buddha tells it to two brahmins by birth, Vāseṭṭha and Bhāradvāja, who are living among the monks in the hope of being ordained. Its point is that nobody is born better than anyone else, because ranks among people arose from what they did and what they agreed. Some scholars read much of it as satire of brahmin stories of origin, and how far the satire reaches is still argued.¹⁴ Midway through its story, the beings live on “ripe untilled rice”, which grows without anyone ploughing for it. “What they took for supper at dusk, by daybreak had grown back and ripened.” So they gathered it twice a day, at dusk and at daybreak.

“Then one of those beings of idle disposition thought, ‘Oh my, why should I be bothered to gather rice at dusk for supper, and at daybreak for breakfast? Why don’t I gather rice for supper and breakfast all at once?’ So that’s what he did.”

Others copied him, and stored rice for two days at a time, then four, then eight. “But when they started to store up rice to eat, the rice grains became wrapped in powder and husk, it didn’t grow back after reaping, leaving a trace showing, and the rice stood in clumps.” So the beings met and said: “We’d better divide up the rice and lay down boundaries.” Then one of them, while guarding his own share, took another’s that had not been given to him, and ate it. Stealing, blame and lying followed, and the beings met again. “Why don’t we elect one being who would rightly accuse those who deserve it, blame those who deserve it, and expel those who deserve it?” They added: “We shall pay them with a share of rice.” The Pāli verb Sujato renders “elect”, *sammanneyyāma*, is a form of *sammannati*, to agree upon or authorise; “let us agree upon” is my rendering. The one they chose was called *mahāsammata*, “popularly elected”, which the text explains as “Elected by the populace”.

The sutta is about how ranks and ownership arose among people, and it argues against the claim that some are born superior. It says nothing of markets or mills. It is not an economic model, and I borrow only its sequence, as an analogy. The sequence begins with a saving of effort. The sutta calls the first saver idle, and the beings later lament the “wicked things” that followed; I read him more gently. One being saves himself a walk, and each who copies him thinks, “This seems fine.” On my reading the fault lies not in saving effort. One act changes a shared stock, and others copy it. The young owners in Coimbatore are not idle, and on standard cloth their saving is right. But the sequence is worth knowing. A shared stock that renews itself stops renewing once each user takes what is convenient. Then come boundaries, then theft, and then an officer to judge them, chosen by users who promise to “pay them with a share of rice”.

Ostrom found that the story need not end there. Users have kept many commons for centuries without dividing them into private plots and without a ruler imposing rules on them, but only where certain conditions held. She called the conditions design principles, and three of them bear on the twelve. “Most individuals affected by a resource regime are authorized to participate in making and modifying its rules.” “Individuals who are accountable to or are the users monitor the appropriation and provision levels of the users.” And “Sanctions for rule violations start very low but become stronger if a user repeatedly violates a rule.” Her other five are as plain. A commons needs clear boundaries, and rules fitted to local conditions, with costs shared in proportion to benefits. Its users need quick, cheap ways to settle conflicts. Governments must recognise their right to make their own rules. And a resource that is part of a larger system is governed in layers, each nested inside the next.⁵ She did not pretend that people always manage it. “Many groups have struggled and failed,” she wrote, and simple prescriptions to hand a resource to a government or to divide it among private owners “may also fail”. Her conclusion was about public policy: “a core goal of public policy should be to facilitate the development of institutions that bring out the best in humans.” On my reading she was not against the state either. Her lecture argues for “polycentric” governance, with “many centers of decision making”, governments among them. On the Noyyal it was the courts that made the units treat their water.

The two accounts agree more than they seem to. Both begin with ordinary people making reasonable choices: a being who saves himself a walk, a herder on an open pasture. Both say the remedy is an agreement among those who depend on the stock. The sutta’s first ruler is one agreed upon by the many, and one of Ostrom’s tests is whether those affected help make the rules. They part on what the agreement makes. In the sutta it makes boundaries, and then an officer over the users, paid with a share of their rice. In the story, the shared field ends in plots and a paid ruler. In Ostrom’s cases the agreement keeps the watching among the users, answerable to them, with small punishments first. They also place the cause differently. The sutta finds it in character, in idleness and greed. Ostrom finds it in rules, and in what people can see of one another.

Neither wins in Coimbatore. Ostrom’s own test of who makes the rules convicts the twelve. Most of those affected by whatever they decided were not in the room: not the inspectors, nor the young people learning to grade, nor the farmers on the Noyyal. And on my reading the sutta warns against any rule that only the richest member can afford to keep. Such a rule is an officer paid with a share of rice, and the users have not kept anything.

Trust between strangers also grows where people keep meeting. In June 2025 the World Health Organization’s Commission on Social Connection reported that 1 in 6 people worldwide is affected by loneliness. It describes loneliness as the painful gap between the connection people want and the connection they have. It linked loneliness to more than 871,000 deaths a year.¹⁵ A neighbour is not a resource, and a companion app can relieve real isolation, as a neighbour can do harm. What bears on the commons is whether a technology supports the places where people meet, as a borrowed room supported the twelve, or replaces them or crowds them out. That is not yet known.

## 39.3 The Civilisational Case: What Kind of Beings Do We Want to Be?

Rajan’s question on the back of the page, what kind of cluster the twelve wanted to be, has a largest form. It asks what a whole society’s production is for.

The print edition said there was “a question that only consciousness can ask”. A system can check its work against the criteria it was given, it said, but “It cannot stand outside that constitution and ask: is this the right constitution?” That has not survived. Since 2022 some systems have been trained to critique and revise their own drafts against a written list of principles.¹⁶ By January 2025, the makers of one reasoning model reported that it showed “self-verification, reflection”, and that it would revisit and re-evaluate its own earlier steps.¹⁷ Whether a machine can ask if its criteria are right is a question about what it does, and the answer keeps changing. This book rests on a different question: who has the right to settle the answer. Being intelligent will never give a machine the right to decide for people. That is the third of this book’s three nevers, and here it covers what their work, their gains and their society are for. It is a commitment that people keep, and it carries no date. It does not forbid delegation. A machine may carry out a decision within limits that people can justify and stand behind, as the cameras pass and fail bolts on the twelve’s floors. What its intelligence never gives it is a right of its own.

Societies have used that right against their own ledgers. The print edition’s example was the end of slavery in the British Empire. It told the story as if historians had settled it: plantations were becoming less profitable, and morality and money moved together. The decline was the argument of the historian Eric Williams in 1944. It is contested. It is also more than one question. Whether plantations paid, whether the slave trade paid, and what Britain’s wider interests were are separate questions, and historians judge them differently. Seymour Drescher argued in *Econocide* (1977) that Britain gave up its slave trade against its own economic interest, when British slavery was, in his publisher’s words, “at the height of its potential”.¹⁸ Parliament ended the trade in 1807.¹⁹ In 1833 it passed the Slavery Abolition Act, and here the old account left out what matters most for this chapter. To pay for freedom the government raised £20 million, most of it by loans, around 40 per cent of its spending that year. It paid the money to the slave owners. The freed received nothing. The Act made those aged six and over “apprenticed labourers”, still tied to the estates and made to work without wages, field workers for up to forty-five hours a week. Apprenticeship ended in 1838. The loans were finally repaid in 2015.²⁰ No single actor made the choice. Parliament voted, campaigners organised, and the enslaved resisted, most widely in the Jamaican rising of Christmas 1831, which in the National Archives’ account hastened the Act. Interests pressed on every side. A declared moral end does not tell us who defined it, who was heard, who paid and who was paid. If Drescher is right, the end they reached was one that no ledger contained. Either way, power decided who was paid for it.

Most of the time no society chooses an end so openly. Instead, a government automates decisions about benefits to make them consistent. A hospital lets a system sort patients to cut waiting times. A school gives every child a platform that adapts to her. Each choice is defensible. Each removes a little of the moment when a person looks at a case and says that the rule does not fit. No single delegation is catastrophic. The gradient is.

That line is about delegations that can each be defended. Some delegations are catastrophic on their own, like the Australian welfare-debt rule told earlier in this book, which a Royal Commission in 2023 found “neither fair nor legal”.²⁰ A delegation like that is large enough to be seen and challenged. The gradient is made of small ones, each defensible and each too small to see. That gradient is the commons again, at the scale of a society. What it wears away is an institution’s capacity to question, contest and correct the decisions it has automated, and the habit of public argument about what things are for. Both are kept by being used, and each reasonable delegation uses them a little less. Delegation can also free attention for harder cases. The harm comes when it runs without feedback, without a way to contest it and without a remedy. This book has followed the same slope through one person’s work, where small handovers that each made sense added up to a judgement nobody exercised any longer. In a society the slope runs through everyone, and nobody owns it.

The machines have made one old question sharper: what is the knowing for? A machine can now supply much of the knowing. It cannot supply the right to say what the knowing is for. That right belongs to the people whose lives the knowing will shape. Those people are many: owners, workers, buyers, farmers downstream and people not yet born, and their interests conflict. So legitimacy asks more than who decides. It asks who is represented, whose rights limit the decision, and how conflicts among them are settled. People not yet born cannot sit at any table, so someone must be charged with their interest.

## 39.4 The Consciousness Dividend vs. the Intelligence Dividend

The print edition set the two dividends in order. The intelligence dividend would come first, and the consciousness dividend second, to decide what was done with the first. It added that “The relationship between the two dividends is multiplicative.” Both are metaphors written as laws, and both go. The two are of different kinds.

The intelligence dividend, in the print edition’s metaphor, is a gain in productivity, which is not the same as output. Productivity rises when the same inputs give more or better, or fewer inputs give the same. The young owners’ proposal would raise it without one more metre of cloth. Such a gain is paid out. Someone receives it: customers, firms, workers, the owners of the machines, their vendors and investors, the public. Who receives how much is settled by competition, contracts, tax law and bargaining power, and by choices like Rajan’s in 2025. The saving from his new systems could have gone into wages, shorter shifts, lower prices or the young men’s training. He kept it.

What the print edition called the consciousness dividend is paid to nobody. It is the commons the gains draw on. When it is kept, its return reaches everyone, and so no one in particular. It may be the buyers’ trust that lets a small mill sell abroad, a pool of people who can grade cloth, or a river a farmer can use. When it is spent, the bill arrives later and goes to people who were not asked. It went to the farmers on the Noyyal. It goes to the young people who are never trained, and to the honest seller undercut by a fake. A hospital may stop refilling its radiologists’ posts as they retire, because a machine reads scans well. It takes its share of the gain, and it spends a place where young radiologists would have learned. Not every saving spends a commons. A saving can protect one condition while it spends another. A camera on the night shift spares tired eyes and lets fewer faults reach the buyer, and it removes the place where the next grader would have learned.

An economist may fairly say that the phrase “consciousness dividend” names nothing she can measure. She is right, and this chapter has not tried. Parts of the commons can be counted roughly. The survey question about strangers counts part of trust. A cluster can count how many people under thirty can still grade its cloth by eye. Returns can be counted buyer by buyer. A river’s water can be tested. None of these measures consciousness. Whether the phrase is worth keeping is an open question, and what it pointed at can be kept without it.

Much of the commons is kept or spent in small acts. A supervisor treats the screen’s green light as a verdict, or as a question to look into. An owner pays for an hour of training, or saves it. A buyer reports a fault, or simply leaves for another mill. None of these acts is large, and none is made by a machine.

### Receipts, please

Four pieces of evidence carry weight in this chapter, and each shows less than a quick reading might take from it. Acemoglu’s 0.66 per cent is an estimated rise in total factor productivity, the part of output growth that more labour and capital do not explain. It covers ten years and rests on his own assumptions. It is not a gain anyone has observed, and it is not output per worker. Catalini, Hui and Wu wrote a working paper built on a model. It shows how leaving work unchecked can pay one firm while others bear the cost. It is not evidence about what all firms do.

The trust surveys record what people say about strangers in general, not how they deal with a particular buyer or mill. Knack and Keefer, and Algan and Cahuc, link those replies to growth, and the direction of cause is still argued. None of them shows that trust is a commons; that is this chapter’s reading. Anthropic’s 96 per cent is the rate for two of the sixteen models it stress-tested, in one scenario it built. It says what they did there, not what models do in use.

## 39.5 A Letter to 2035

Dear reader,

I am writing in October 2026, in India. You are reading in a world I cannot see. In 1930 the economist John Maynard Keynes wrote an essay to people not yet born. He made two forecasts. He was roughly right about how much richer they would be, and wrong about how little they would work.²¹ A letter to the future is worth writing only if its reader can check it. So here is what I argued, and what would show me wrong.

I argued three things that do not depend on what machines can do. A machine will never be you. A machine’s makers and users will never be free of what it does. And being intelligent will never give a machine the right to decide for people. I also argued that the gains from machines draw on a commons nobody owns. It is made of trust between strangers, people able to check and refuse, and the habit of asking what a saving is for.

Here is what I expect. It is my own guess, and nobody has scored it. By 2030, I expect cameras like the ones in this chapter to check most of the cloth that leaves the mills around Coimbatore, somewhere between half and nearly all of it, and many of them to be set to a buyer’s own limits. If by 2030 less than half of that cloth passes a camera, or almost none of the cameras carries a buyer’s limits, I was wrong about the machines. By 2035, the cluster will have fewer people under thirty who can grade cloth by eye than it has now: perhaps a modest fall, perhaps most of them gone. If a count like the one this chapter describes finds as many as now, I was wrong. What I cannot guess is whether that will matter. Suppose the mills that stopped paying inspectors keep their buyers, their returns stay level through 2030, and the cameras learn each buyer’s limits from what comes back. Then much of the eye was a record that nobody had written down, and what was lost was mainly livelihoods. That is a real loss, and this letter will not pretend otherwise. If buyers begin to leave the mills that kept no one able to check, I was right in that one place. Either way, you will know, and I cannot.

Here is what worries me. This book has described qualities that people wear into themselves by their own choices, and practices that take time, as if they were open to everyone. They may not be. Rajan has the means to spend a day each month at a table with eleven other owners; the woman who runs a dye house in Tiruppur does not have the means to pay a trainee. Such means are not luxuries, but they are conditions. Who gets those conditions, paid time to learn and shared training for firms too small to train alone, is itself a matter of politics and price. There are billions of people for whom the daily pressure of survival leaves little room for the kind of practice this book describes. If what the print edition called the consciousness dividend is real, it may be kept first by those who already have the margin to keep it, and serve them first. That would make it not a corrective to inequality but a deepening of it. I do not have an answer to this. A commons kept only by those who can afford to keep it is the same problem in a smaller room.

With that said, I hope for certain things. I hope people still know their neighbours, as the twelve came to know one another round a borrowed table once a month. I hope the question what is this for? is asked as naturally as the question how do we make this faster? I hope someone near Coimbatore still knows what a buyer in Belgium will send back, and that it is written down somewhere as well.

The Cambridge mathematician Frank Ramsey wrote in 1928 that to count later enjoyments as worth less than earlier ones is “ethically indefensible” and arises merely from “the weakness of the imagination”.²² That is an ethical claim, not a rule for the rates at which economists discount. I have tried to write as if your good counts as much as mine.

You can check me. That is what a dated letter is for.

Nithin Palal

This is the last chapter of Part VIII, the part of the book about what we are becoming. Each of its chapters looked for the person of 2035 and found something being decided now. A machine enlarges whatever is brought to it, lapses as well as attention, and what we become is decided less by what machines will be able to do than by the habits we keep, the children we form and what we agree about who answers. Nobody can hand on her attention. She can show her practice to someone who must form it for herself, and answer for what she handed on only while she can still be asked. And the gains from machines draw on a commons that nobody owns, kept only where its users agree to keep it and pay. Together they change what Part VII meant. Part VII found that an organisation can make practising possible for its people, or make it impossible. Part VIII finds that what makes practising possible does not belong to any one organisation either. The trust, the rungs and the people able to check are held in common, across firms and across generations. They are kept only by agreements that someone pays for.

### What would prove this wrong?

The letter’s forecasts can be scored only if their words are fixed. By “most of the cloth” I mean the share of bolts, counted one by one, that pass a camera before they leave the export mills of Coimbatore, Tiruppur and Karur. A buyer’s own limit is a setting written into the camera from that buyer’s specification or returns, not the vendor’s general setting. A person “can grade” if, shown a buyer’s sample, she sorts it as the buyer would.

I have found no count of either kind for 2026, and I have not made one. The camera forecast names a share in 2030, so it needs none. The forecast about graders compares 2035 with now. A reader would have to count graders under thirty with a test like the one above. The first count would become the baseline, which makes my claim weaker than it sounds.

Four questions must be kept apart: whether the skill declined, whether quality fell, whether buyers left, and whether the first caused the others. Buyers also leave over prices, currencies and freight. And there is a third outcome the letter does not name. The cameras may improve inspection while new roles and routes form around them, fewer and different, in which people still learn to see. If that happens, my worry was wrong in its form. Who paid for the change would still be a fair question.

## Ending: An Hour in the Wage Book

At the table, the younger of the two owners read out his figures. On standard cloth, in four months, his inspectors had stopped eleven bolts that the camera had passed. On a second look nine of them had gone to buyers after all, and none had come back. He wanted to move two of his four inspectors to other work and let two go, with three months’ wages each.

The woman from the dye house spoke next. The figures showed what the machines could see, she said, and nothing about what they could not. At the end of January 2011 the court had closed her father’s dye house with seven hundred others. For years every unit had saved on what it sent into the river, and the farmers had paid. “Every unit’s saving was sensible,” she said. “The river paid for all of them.”

Rajan spoke last, and first told them his interest: in September he had put a modest sum into a young company that matches buyers to mills, to which shared logs of stopped bolts would be worth something. He asked them to weigh what followed with that in mind.

Then he proposed a rule for speciality cloth only, the cloth woven or finished to a buyer’s own standard. On every shift, each member who signed would name one person who might stop any bolt the camera had passed, and write one line of reason in a log. A stop made in good faith would never be held against her, even if the buyer later took the bolt. The logs would be read at each meeting, and would stay with the twelve, never sold or passed to anyone, his young company included. Beside her would stand one trainee, paid for the hour, learning what she saw. Each member would write down the limits his buyers had shown by what they sent back, and have them set into the cameras. A member who broke the rule would be named at the next meeting. A second time, he would lose the shared reports, and a third time he would leave. A member who could not pay would not be treated like one who would not, and any disputed stop or sanction would be heard by the whole meeting. And twice a year two inspectors from other members’ floors, a different pair each time, would read each mill’s logs. They would declare any tie to that mill, and tell the meeting what the logs were missing.

He did not claim that the rule would keep a single buyer. Nor did he claim that an hour a shift would make a grader. It would be a beginning. He said only that he did not want to learn in 2030 that nobody under thirty could see a band of tight weaving.

Seven of the twelve signed. The two young owners did not; they made mostly standard cloth, and on standard cloth the logs were on their side. The towel maker from Karur had no cameras and saw no reason to. One owner asked for a month. The woman from the dye house had argued hardest for the rule, and she could not pay a trainee’s wages. Rajan offered to pay half for a year. She refused. “Then it is your rule, not ours,” she said. She would not owe the rule to the man who paid for it. If the members wanted trainees, they could pay into a common training fund, each by its output, and she would propose that next year. She did not sign.

So the seven would carry the trainees’ hours and a slower speciality line, while the five who had not signed went on selling under the region’s name. The seven said they would accept that for now and look at it again in a year. Their signatures were the start of an experiment. It might work, or fail, or be taken over by the larger firms, and the rule did not keep the commons by existing.

On his own floor, Muthu already held the seat the rule described, and had kept his log since August 2025. The trainee was the new cost. Of the four young men who set up looms under Muthu, Rajan asked two the same question. Did they want an hour a shift beside Muthu on the speciality line, learning to see a fault in running cloth? One said yes. The other said he would rather learn the controls of the newer looms, and Rajan let him. The hour came out of loom time, so the speciality line would run a little slower. Rajan paid for it from what he had saved in the spring, when he moved one Tiruppur dye house’s share of his dyeing to two others. Two of that dye house’s dyers had paid for the saving with their shifts. One rung was being kept with money another trade had lost, and nobody had asked the two dyers.

On Monday morning the young man stood beside Muthu at the speciality looms. Rajan wrote the hour into the wage book himself.

### Practice

*What the saving spends.* Take one saving that machines let your work make this year: fewer checks, fewer hires, less time on a task. Write down what the saving draws on that nobody is paid to keep. It might be people who were trained somewhere else, the trust strangers place in your trade’s name, a public record, a road or a river. Write who keeps it now, and at whose cost. Write who else draws on it, and whether you could agree a rule with them. Write who would be affected by such a rule and would not be in the room where it was made. Then write what keeping your share would cost you, in money or in hours. This takes an hour; if the hour is your employer’s, note who gave it. It will not make your firm richer, and it may cost you the saving and a rival’s goodwill. It shows what your gains stand on.

## Words to keep

- <strong><em>Aggañña Sutta</em></strong><strong>:</strong> Pāli, “What Came First” (*Dīgha Nikāya* 27), the Buddha’s account of how ranks arose among people, told against the claim that some are born better.
- **common treatment plant:** a plant, shared by many factories, that treats their waste water (in India, a common effluent treatment plant).
- **commons:** a stock that many people draw on, that nobody owns, and that each use, or each betrayal, can spend; in this chapter, trust between strangers, people able to check and refuse, and the habit of asking what a saving is for. The word covers conditions of different kinds, whose benefits reach beyond whoever keeps them.
- **design principles (Ostrom):** the conditions Elinor Ostrom found in commons that their users kept for a long time.
- **dividend:** the share of a company’s profit paid to the people who own its shares.
- **externality:** a cost or benefit that falls on people who were not party to the sale that caused it.
- <strong><em>mahāsammata</em></strong><strong>:</strong> Pāli, the first ruler in the *Aggañña Sutta*, chosen by the beings and paid with a share of rice; “popularly elected” in Sujato’s translation.
- **productivity:** more or better from the same inputs, or the same from fewer; not the same as output. Total factor productivity is output per unit of labour and capital taken together; its growth is the part of output growth that more labour and capital do not explain.
- **public good:** something many people benefit from whether or not they pay for it, which is not used up by being enjoyed and is therefore often undersupplied.
- <strong><em>sammannati</em></strong><strong>:</strong> Pāli, to agree upon or authorise; the verb behind *mahāsammata*.
- **zero liquid discharge:** treating and reusing all of a factory’s waste water, so that none is released.

## Sources

1. Daron Acemoglu, “The Simple Macroeconomics of AI,” NBER Working Paper 32487 (May 2024): “no more than a 0.66% increase in total factor productivity (TFP) over 10 years,” and, allowing for harder tasks, “less than 0.53%”; published in *Economic Policy* (2025). See Chapter 5.
2. Christian Catalini (MIT), Xiang Hui (Washington University in St Louis) and Jane Wu (UCLA), “Some Simple Economics of AGI,” arXiv:2602.20946, v1 24 February 2026, v2 25 February 2026 (working paper): “Unverified deployment becomes privately rational”. See Chapter 23.
3. “Tirupur dyeing units told to close,” *Down To Earth*, 14 February 2011: the Madras High Court “has ordered closure of more than 700 bleaching and dyeing units” on a petition by the Noyyal River Ayacutdars Protection Association, for failing “to comply with its earlier order to achieve zero liquid discharge of effluents”; “On January 28, the court also asked the state government to disconnect electricity supply”; the same article reports that the units appealed the 2006 orders, that the Supreme Court stayed the closure in 2007, and that “In October 2009” it set “a January 5, 2010 deadline”; and that the Supreme Court’s orders of October 2009 also required the units to “pay compensation for damaging the area’s ecology till April 2010”.
4. “Towards zero discharge,” *Down To Earth*, 13 May 2011: farmers in court since 2003; “the court’s 2006 order to ensure zero liquid discharge”; “At present, there are 20 common effluent treatment plants (CETPs)”; according to factory owners, “implementing ZLD has increased costs 25 to 30 per cent”.
5. Elinor Ostrom, “Beyond Markets and States: Polycentric Governance of Complex Economic Systems,” *American Economic Review* 100 (June 2010): 641–672, her Nobel prize lecture of 8 December 2009: Hardin’s users “trapped in an inexorable tragedy of overuse and destruction” (p. 649; the original sets “a pasture open to all” between dashes, so the sentence is quoted in part); irrigation in Nepal and forests (p. 641); the eight design principles, three quoted and five paraphrased in the text (p. 653, where Ostrom adapts the wording of Cox, Arnold and Villamayor Tomás, 2009); “polycentric” as “many centers of decision making that are formally independent of each other” (p. 643); “Many groups have struggled and failed” and policies that “may also fail” (p. 659); “a core goal of public policy should be to facilitate the development of institutions that bring out the best in humans” (p. 665). Garrett Hardin, “The Tragedy of the Commons,” *Science* 162 (1968): 1243–1248. Elinor Ostrom, *Governing the Commons: The Evolution of Institutions for Collective Action* (Cambridge: Cambridge University Press, 1990), ch. 3, pp. 58–102 (the institutions studied are from more than 100 to more than 1,000 years old). An earlier table of the same principles: J. M. Anderies, M. A. Janssen and Elinor Ostrom, “A Framework to Analyze the Robustness of Social-ecological Systems from an Institutional Perspective,” *Ecology and Society* 9, no. 1 (2004), art. 18, Table 3, “Design principles derived from studies of long-enduring institutions for governing sustainable resources” (“Based on Ostrom (1990)”).
6. “Elinor Ostrom, 1st woman to get Nobel economics prize,” *The Boston Globe*, obituary, 12 June 2012; she shared the 2009 prize with Oliver Williamson (“Two Americans win Nobel economics prize,” NBC News, 12 October 2009).
7. Integrated Values Surveys (World Values Survey and European Values Study), question as given by Our World in Data, “Share of people agreeing with the statement ‘most people can be trusted’”, 1984 to 2022.
8. Stephen Knack and Philip Keefer, “Does Social Capital Have an Economic Payoff? A Cross-Country Investigation,” *Quarterly Journal of Economics* 112, no. 4 (1997): 1251–1288, abstract: “This paper presents evidence that ‘social capital’ matters for measurable economic performance”; World Values Survey data for twenty-nine market economies.
9. Yann Algan and Pierre Cahuc, “Inherited Trust and Growth,” *American Economic Review* 100, no. 5 (2010): 2060–2092, abstract: “inherited trust of descendants of US immigrants is significantly influenced by the country of origin”; “the sizeable causal impact of inherited trust on worldwide growth during the twentieth century”.
10. Adam Smith, *Lectures on Jurisprudence*, report of 1766; text as in *Lectures on Justice, Police, Revenue and Arms*, ed. Edwin Cannan (Oxford: Clarendon Press, 1896), Part II, Division II, §17, “Of the Influence of Commerce on Manners”. See Chapter 15. On agreements among rivals: Competition Act, 2002 (India), s. 3(3): an agreement among enterprises “engaged in identical or similar trade” that “directly or indirectly determines purchase or sale prices” (a), “limits or controls production, supply, markets, technical development, investment or provision of services” (b), or shares the market by “type of goods or services, or number of customers in the market or any other similar way” (c), is presumed to have an appreciable adverse effect on competition; the subsection also covers a “practice carried on, or decision taken by, any association of enterprises”; s. 3(2): such an agreement “shall be void” (text as reproduced by taxheal.com, checked 31 October 2026).
11. Annette Baier, “Trust and Antitrust,” *Ethics* 96, no. 2 (1986): 231–260; see Chapter 3.
12. Anthropic, “Agentic Misalignment: How LLMs Could Be Insider Threats,” 20 June 2025, on “16 leading models from multiple developers”: “In at least some cases, models from all developers resorted to malicious insider behaviors”; “Claude Opus 4 blackmailed the user 96% of the time”, as did Gemini 2.5 Flash at the same rate; “All the behaviors described in this post occurred in controlled simulations”; “Our experiments deliberately constructed scenarios with limited options”; “We have not seen evidence of agentic misalignment in real deployments.” Dario Amodei, chief executive of Anthropic, “The Urgency of Interpretability” (darioamodei.com, April 2025): “People outside the field are often surprised and alarmed to learn that we do not understand how our own AI creations work.”
13. *Aggañña Sutta* (*Dīgha Nikāya* 27), “What Came First,” trans. Bhikkhu Sujato (SuttaCentral, 2018; dedicated to the public domain), segments 0.2, 1.2–1.3, 16.1–16.2, 17.5–17.8, 17.9–17.20 (paraphrased), 17.12, 17.21–17.23, 18.2, 18.27, 19.1 and 19.12 (paraphrased), 20.2 (paraphrased), 20.3–20.4 and 21.1; Pāli from the Mahāsaṅgīti edition on SuttaCentral: *sammanneyyāma* (20.3), *mahāsammato* and *mahājanasammato* (21.1).
14. Richard Gombrich, “The Buddha’s Book of Genesis?”, *Indo-Iranian Journal* 35, no. 2 (1992): 159–178, which reads the sutta as satire and parody of brahmin ideas of caste (as summarised in Yingjin Chen, “The Establishment of Buddhist Creation Myth Text,” *Religions* 14, no. 6 (2023), art. 706); Steven Collins, “The Discourse on What Is Primary (Aggañña-Sutta): An Annotated Translation,” *Journal of Indian Philosophy* 21, no. 4 (1993): 301–393, which has a section on “the nature of the ‘satire’” in the sutta (contents as listed by the Digital Library of Buddhist Studies, National Taiwan University).
15. World Health Organization, “Social connection linked to improved health and reduced risk of early death,” news release, 30 June 2025, on the global report of the WHO Commission on Social Connection: “1 in 6 people worldwide is affected by loneliness”; “Loneliness is described as the painful feeling that arises from a gap between desired and actual social connections”; “more than 871 000 deaths annually”.
16. Yuntao Bai et al., “Constitutional AI: Harmlessness from AI Feedback,” arXiv:2212.08073 (15 December 2022): “sample from an initial model, then generate self-critiques and revisions”; “The only human oversight is provided through a list of rules or principles”.
17. DeepSeek-AI, “DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning,” arXiv:2501.12948v1 (22 January 2025), §1.1: “DeepSeek-R1-Zero demonstrates capabilities such as self-verification, reflection, and generating long CoTs”; §2.2.4: “the model revisits and reevaluates its previous steps”. The later wording, “self-reflection, verification, and dynamic strategy adaptation”, is from v2 (4 January 2026), the version published in *Nature* 645 (2025): 633–638.
18. University of North Carolina Press, description of Seymour Drescher, *Econocide: British Slavery in the Era of Abolition* (Pittsburgh: University of Pittsburgh Press, 1977; second edition, Chapel Hill: University of North Carolina Press, 2010): “a classic analysis and refutation of Eric Williams’s 1944 thesis”; abolition “detrimental to Britain’s economic interests at a time when British slavery was actually at the height of its potential”. Eric Williams, *Capitalism and Slavery* (Chapel Hill: University of North Carolina Press, 1944).
19. UK Parliament, Living Heritage, “Parliament abolishes the slave trade”: “the Bill received Royal Assent on 25 March 1807”.
20. Full Fact, “This is what we know about the government loan to pay slave owners compensation after slavery was abolished in 1833,” 2 July 2020 (updated 6 July 2020): loans “amounting to £20 million”, though “A further £5 million was paid out directly in government stock”; government expenditure in 1833 “£48.8 million, so the £20 million was around 40% of that”; “Freed slaves did not receive any compensation”; most former slaves “still had to work without pay for a number of years”; “finally repaid in 2015”. HM Treasury, freedom of information reply FOI2018/00186 (31 January 2018): “The Government used £20 million to fund the Slavery Abolition Act 1833”; “The 4% Consolidated Loan was redeemed on 1 February 2015”. On apprenticeship: Slavery Abolition Act 1833 (3 &amp; 4 Will. IV c. 73), s. I (from 1 August 1834 those aged six or over to “become and be apprenticed Labourers”), s. V (praedial apprentices not beyond 1 August 1840, nor to labour “for more than Forty-five Hours in the whole in any One Week”) and s. VI (non-praedial apprentices not beyond 1 August 1838), text as on the Irish Statute Book (irishstatutebook.ie, checked 31 October 2026); The National Archives, “The 1833 Abolition of Slavery Act and compensation claims”: “The period of apprenticeship ended in 1838”. On resistance: The National Archives, “Spotlight On: Baptist War,” video transcript: the revolt “took place in Jamaica over Christmas 1831”, “Some sixty thousand of the enslaved took part”, and it “accelerated the passing of the Slavery Abolition Act in 1833”. On the Australian welfare-debt rule: “Crude, cruel and unlawful: Robodebt Royal Commission findings,” *Law Society Journal* (NSW), lsj.com.au, 9 August 2023, on the *Report of the Royal Commission into the Robodebt Scheme* (7 July 2023): “a crude and cruel mechanism, neither fair nor legal”. The report’s own sentence (p. xxix) is “Robodebt was a crude and cruel mechanism, neither fair nor legal, and it made many people feel like criminals”, as quoted in the Australian Government’s response to the report (November 2023), p. 3, whose footnote 1 gives the report’s page; the report itself was not read for this edition, because its website refuses automated access. Income averaging, as used, “neither produced accurate results nor complied with the income calculation provisions of the Social Security Act 1991 (Cth)” (the Analysis and Policy Observatory’s entry for the report, apo.org.au; report of 7 July 2023, updated 11 July 2023). See Chapter 6.
21. John Maynard Keynes, “Economic Possibilities for our Grandchildren” (1930), first published in *The Nation and Athenaeum*, 11 and 18 October 1930; reprinted in *Essays in Persuasion* (New York: W. W. Norton, 1963), pp. 358–373. See Chapter 4, which sets his forecasts against what happened.
22. F. P. Ramsey, “A Mathematical Theory of Saving,” *The Economic Journal* 38, no. 152 (1928): 543–559, at p. 543: discounting later enjoyments is “a practice which is ethically indefensible and arises merely from the weakness of the imagination”. See Chapter 21.

*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 “intelligence makes an economy productive; consciousness makes it wise”, that consciousness “produces a dividend of its own”, and that “The relationship between the two dividends is multiplicative”; that a camera “did not know which defects mattered to which buyer” because knowing it “required not intelligence but the accumulated consciousness of people”; that the difference between three mills “was not the tool. It was the consciousness surrounding the tool”; that the cluster’s judgement is “an economic asset” that “attracts buyers who could source fabric more cheaply elsewhere”, and that Coimbatore “commands a premium”; that “AI cannot build this kind of trust” because “trust between conscious beings requires the possibility of betrayal”, and that “A system that is incapable of betrayal is also incapable of earning trust”; that there is “a question that only consciousness can ask”, and that an AI “cannot stand outside that constitution”; that “Economic historians have shown” that abolition followed the declining profitability of the plantations; that the traditions converged “in contexts that could not have influenced one another”; that the choice is “available to anyone who is conscious enough to notice”; that the fabric of a mill without its graders would be “technically flawless and emotionally dead”; that the book “describes twelve qualities of consciousness”; that a teacher in Madhya Pradesh “has the institutional support of a state education council”; and, in the letter, that the reader’s debates “have been settled”, and “That choice is consciousness.” It also set the chapter in November and December, after the book’s present; gave the network “two years of monthly meetings”; named the start-up Rajan invested in after a real company and had it ask to feature the cluster; left the oldest mill’s owner and its supervisor with names, one of them shared with real textile companies in the region; sent her cloth to a city that in this book belongs to another of Rajan’s buyers; put the meeting in a government body’s offices; priced a shirt in “kroner”; called on “the book’s own three-tier system”, which it never defined; gave a teacher from another chapter a circle of “three hundred teachers across three states” and a state council’s validation; made a radiologist from another chapter the wiser doctor against an invented hospital “down the road”; listed three of the book’s other people in a roll-call; and closed on a room that “reached something better” and on “That is the consciousness dividend.”
- *What I now hold:* consciousness pays nobody, and nothing in this chapter measures it. The gains from machines are paid out, to customers, firms, workers, the owners of the machines and the public, by competition, contracts, law and choices. What they draw on is a commons: trust between strangers, people able to check and refuse and the places where they learn, and the habit of asking what a saving is for. It is spent by sensible savings, and parts of it, like a river, are more fragile than a public good, because use spends them. Part of a supervisor’s eye is a record nobody kept, part is a position someone grants or withholds, and part is a rung; on standard cloth the camera is right more often. A cluster has spent a commons before (the Noyyal, 2011). Users can keep a commons without dividing it or appointing a ruler, but only under conditions, and many fail (Ostrom); the remedy is an agreement among those who depend on it (the *Aggañña Sutta*), and a rule kept only by the richest member is not one the users keep. Trust can be counted in part and may follow institutions as much as build them; records build trust between strangers and can spend it; trust in a machine is trust in the people behind it. Whether a machine can question its criteria is a dated doing; who may decide what a society’s gains are for is a question of right. Abolition is contested history, and its compensation went to the owners. The inequality worry stands. Trust is spent by being let down, not by being relied on, and any seller under a shared name can spend it for all. A machine’s makers answer for releasing what they cannot fully explain. A rule among rivals that shuts out newcomers is a guild, not a commons kept. Some single delegations are catastrophic and seen; the gradient is made of the defensible ones. “Commons” is the chapter’s broad word for conditions of different kinds, from a river used up by use to a right that is no stock at all. What joins them is that their benefits reach beyond whoever keeps them. A firm can grow more productive on what it did not train, build or keep, and an economy less able to reproduce what its productivity stands on. Productivity is not output, and not every saving spends a commons; the questions are what value is created, who captures it, and to whom the costs move. Keeping a skill is not keeping the jobs, nor a duty on every worker. A platform moves trust rather than removing it, and incomplete understanding does not by itself make a release wrong. Rivals who share information keep to a stated boundary, and an open door separates a commons from a guild; buyers are inside the commons. The sutta is an analogy, and Ostrom was not against the state. The third never does not forbid delegation; legitimacy asks who is represented and who speaks for people not yet born. The rule signed by seven is an experiment: the seven carry its cost while the five who did not sign still trade on the region’s name.
- *What changed my mind:* the rule that consciousness is never the hinge and that a machine’s abilities carry a date; Elinor Ostrom’s account of how users keep a commons; the *Aggañña Sutta*’s story of the rice; the Noyyal case; Knack and Keefer, and Algan and Cahuc, on trust; Anthropic’s 2025 simulated tests and the self-critique of reasoning models; Drescher’s dispute with Williams and the record of the 1833 compensation; this book’s earlier findings on unkept records, empty positions, the missing rung and gains as a choice; and two outside readings of an earlier version of this chapter, in October 2026, which showed it calling very different things a commons.
- *What survives:* Rajan’s question, “whether, in making things better, they were also making something invisible worse”, now Rajan’s worry in the opening, under the chapter’s doubt; “No firm gains an edge from having electricity. Every firm suffers from lacking it.”; “The question is who decides what the production is for.”; “No single delegation is catastrophic. The gradient is.”; “What is the knowing for?”; the dyer’s line about what the data shows; “What kind of cluster do we want to be?”, now written on the back of Rajan’s copy of the proposal; the letter to 2035, with its inequality paragraph nearly as written (“If the consciousness dividend is real, it may accrue first to those who already have the margin to cultivate it ... I do not have an answer to this.”), and its hopes (“I hope people still know their neighbours.”); the network of twelve and the meeting, now a decision with its price; the dye-house owner’s refusal, now with her reason and her proposal for next year. The title is kept and argued with.
- *Section titles in the print edition:* 39.1 was “The Economic Case for Consciousness”. The other four keep their print titles.

*Next in the series:* [*Part IX: The Unfinished Chapter (overview)*](https://nithinpalal.com/ai/ai-proof/part-9-the-unfinished-chapter)*.*