NP Nithin Palal

Chapter 38: Building Beyond Yourself

Part VIII: The 2035 Human

· 40 min read · Discuss · Plain text

AI-Proof: What Machines Can Never Be — Volume III: Transformation, Part VIII: The 2035 Human, Chapter 38: Building Beyond Yourself

Part VIII asks what kind of person the next decade needs. After the amplified person of Chapter 37, this chapter asks what one person can hand to another when machines explain almost anything, and what outlasts her.

A guest teacher inherits a class that keeps asking how anyone found out. Her phone writes the stories, and they work. What can a teacher hand on now, which habits should not be copied, and who answers once it leaves her?

In Brief For readers who teach, mentor or train others, and wonder what is still theirs to pass on. Chapter 38 follows a science teacher in Sagar whose last year’s class carries one of her questions to its new teacher, a guest hired for a single school year. The new teacher sits at the back of her room, a phone writes stories for them both, and a committee in Bhopal turns the older teacher’s stories into a module with a rubric. The Taittirīya Upaniṣad’s parting charge to a pupil, which hands on a test rather than a copy, and Alasdair MacIntyre’s account of a practice and its internal goods help ask what one person can hand to another when machines explain almost anything, who gets it, and who answers for it once it leaves her.

Nobody can hand on her attention; she can show her practice to someone who must form it for herself, say which of her acts not to copy, and answer for what she handed on only while she can still be asked.

Opening: A Question Down the Corridor

“But how did they find out?”

The question came through the open door of the Class 10 room, in a boy’s voice that Kavitha Sharma knew. It was late July 2026. Kavitha teaches science to Class 9 at a government school in Sagar, in Madhya Pradesh, and in April she had begun her thirteenth year. The voices in that room were her last year’s class, forty-three of them, a year older and with a new teacher.

For a year she had told them stories about how someone came to know something. Who first saw it, how they checked it, who refused to believe them. She called these her teaching stories. The class had kept the question and carried it down the corridor.

The new teacher was Komal. She was a guest teacher in her twenties, with a science degree, a teaching degree and a post for one school year. The school’s second science post had been filled that way for years. After the bell she stopped Kavitha in the corridor. Your old class asks that about everything, she said. Who found out, and how. She had a syllabus to finish before the Board examination, and a class that wanted the history of every fact. “What do you do with them?”

Kavitha found that she could not say. She told Komal she could sit at the back of the Class 9 room in her free periods and watch. The head teacher, when they mentioned it, called it a good idea and gave neither of them a period for it.

That same month a committee in Bhopal, the state capital, had asked Kavitha to present her method in September, so that it might go into the state’s guidance for teachers. So in one month her practice was asked to travel in three ways. A class carried one of its questions to the next teacher. A committee wanted it in writing. A young teacher wanted to watch. A fourth way needed nobody’s permission. Komal’s phone would explain almost anything in her syllabus, in Hindi, at any hour, and tell a good story about how it was found. When a machine can explain almost anything to anyone, what can one person hand to another that outlasts her, and who answers for it once it has left her hands?

The print edition answered with consciousness. “Individual consciousness is the foundation,” it said. “Collective consciousness is the structure.” Legacy was “the slow, imperfect, irreplaceable transmission of awareness from one life to the next”. This edition cannot say that. Awareness stays where it is. What one person can show another is a practice. A practice forms only in the one who practises it, by her own repeated acts, on occasions that somebody pays for. Words and examples can help it form. They cannot stand in for doing it. Explanations and stories can come from a book or, as of 2026, from a machine, often well. What a teacher can add is smaller and harder to find. It is the sight of someone keeping her own rules when they cost her, and her word about which of her habits not to copy. It reaches first the people who happen to be near her. Once it leaves her, the people who take it up answer for what it becomes.

The chapter’s title needs one correction. “Building” suggests something its maker can finish and leave standing. What outlasts a person is taken up by others, or it is not. So building beyond yourself can only mean building the conditions in which someone else might take something up: a practice written down, an hour given, a door left open. Whether anyone takes it up is theirs to decide.

A practice outlives its maker only when someone else can enter it and do real work in it. She must be corrected, question its habits, change what no longer works and in time answer for it. That takes more than a teacher willing to teach. It takes people with time to learn, institutions that pay for that time, colleagues who let newcomers in, and successors free to become something other than copies. Machines can keep, explain and extend what a practice has learned. They can also remove the work through which its next practitioners would have learned. What survives depends less on what one person leaves than on whether the next is given the means to take it up.

38.1 What a Practice Needs From Others

This section is about one person’s practice inside a world that is moving. It is not about moats.

The old argument ran like this. Each person’s protection against machines is a wall around one person. People who share a standard can build a larger wall together, one that “No platform can generate” and “No algorithm can fake”. Both halves fail. As of 2026, a platform that matches buyers to sellers keeps a record of how each seller has behaved. Steady conduct across many dealings is exactly what such a record shows. And a group of competitors held to one standard of refusal has an older name. It is a guild. A guild protects its members and closes its door on newcomers. This book has already said that belonging used as a moat is lock-in. Not every shared standard is a guild. A standard can protect the people a trade serves, as a licence to practise medicine is meant to. It becomes a moat when its first use is to keep newcomers out.

A practice needs other people. Kavitha’s teaching stories need a timetable with room for them. They need a head teacher who lets a lesson leave its plan. They need a district office, which decided that from August 2025 a learning platform would teach her class the syllabus twice a week, on tablets shared in pairs. They need parents who do not complain when a story takes most of a period. They need a Board examination that tests what can be marked in three hours, and a school that is judged by its results. And they need next year’s teacher, who inherits whatever Kavitha leaves in the students. Each of these is someone else’s decision.

Kavitha controls none of them, and she had always known it. What she had not thought about was the last. A practice that lives in one teacher’s room is also a habit in a whole class of students, and the students move on every April. Whatever a teacher builds in a class, she builds partly in someone else’s classroom, without asking that someone. The students had not taken her practice. They had taken a habit of asking, which is something else, and it worked on whoever taught them next.

That someone was now Komal. A class that asks “how did they find out?” about every fact is slower to teach. In a Board year, every period spent on a history is a period not spent on what the examination tests. Kavitha had left something in her students, and the next teacher was paying for it in time. By the end of August Komal reckoned it had cost her about six periods, most of a chapter on heredity she would now have to rush in October. Nobody else counted them. No report the school sent anywhere recorded a period spent on a question. Some of the students, Komal suspected, asked the question partly to slow her down before a test. She was probably right about a few. A good question can be used for a poor reason, and a teacher’s habit, once it lives in a class, serves the class’s ends as well as hers. A legacy can arrive as a cost to the person who receives it. It is not only what a predecessor gives. It is also what a successor must carry, usually without the hours or the authority to manage it.

This book has learned from ubuntu, the southern African idea that a person is a person through other persons. On that view a person’s capacities, duties and sense of self are formed through relations with others. What happens between people can also close around them. A class that has learned one good question may stop asking others. A staffroom where everyone agrees can stop seeing. A practice needs other people, and it needs them to stay different from one another.

38.2 Building Communities Around a Practice

A community is a group of people who share a practice, hold each other accountable to that practice, and generate understanding together that none of them could generate alone. The word that carries it is practice, and a philosopher has given that word a precise sense.

Alasdair MacIntyre, a Scottish-born philosopher, published After Virtue in 1981. In its fourteenth chapter he defined a practice as a shared and complex human activity “through which goods internal to that form of activity are realized” by trying to meet “those standards of excellence which are appropriate to, and partially definitive of, that form of activity”.1 His example is chess. A clever child who is paid in sweets to play chess has every reason to cheat, as long as the sweets are what she wants. Sweets, money and status are external goods. They can be had in other ways. Only when she comes to want what chess itself gives, a particular kind of analytical skill and strategic imagination, does cheating begin to defeat her. Those internal goods, MacIntyre argues, can be recognised only by people who have taken part in the practice. Excellence inside a practice does not show that its purposes are just, or that its door is open.

He later said, in a published conversation, that “teaching itself is not a practice”. Teaching, on his view, is a set of skills that serve other practices.2 I read his point this way, and the reading is mine. The practice Kavitha’s stories serve is science. What they hand on is the beginning of it: how someone came to know a thing, and how they checked.

Test the definition on two communities. The first is large. Kavitha belongs to a group of science teachers from across the state, on a messaging app, with hundreds of members. In April 2026 a teacher in Bhopal asked how to teach a topic so that students understood it instead of memorising it. The replies came quickly: links, worksheets, a video, tricks for the examination. Many of the worksheets had been made by machines. Kavitha typed one of her stories, slowly, in Hindi and English. Two teachers replied. One asked whether she had written the others down.

She had not. Over the next weeks she wrote four of them, and in May she sent them to the state programme. That is how her stories first left her: as text, to strangers. Whether anyone in the group used the story she typed, she never learned.

By that definition, the group is a place to share material more than a community of practice. In MacIntyre’s terms it trades mostly in external goods: worksheets that save an evening, tricks that raise marks. Those are real goods, and teachers who are short of time are right to want them. What the group cannot pass round is what shows only inside a lesson. Nobody in it holds anybody to anything. Its attention moves together. The same worksheet comes round again and again, forwarded by teacher after teacher, until everyone teaches the topic the same way. A group’s attention can make its members alike as easily as it makes them wise. Distance is not what makes the difference. Teachers online who share recordings of their lessons and criticise them hold each other to something. What matters is whether people show their work to those who can judge it.

The machine’s part in the group was mixed and, in one case, plainly good. The most useful thing posted that month was a machine-made worksheet that one teacher had checked line by line against the textbook. She said so when she shared it, and she said what she had not checked. Her post did what a careful signature does: it told the reader where to look again. If others did the same, the group would have the start of its own standard for machine-made material.

The second community is small. It is the staffroom. Sunita has taught mathematics in the room next to Kavitha’s for almost ten years. Now Komal had a chair at the same table. They could see each other’s classes, and they could be asked why. That is the part of the definition the large group lacked. Nobody had arranged it. Sunita watched it begin and said what she thought, kindly. “She will take your evenings, Kavitha. And when the year ends, she will be gone.”

Sunita was not against it. She had seen other arrangements begin in that staffroom and end when a post did, and her warning was about hours. It was fair. The hours would not be wasted, but they would be given away. Whatever Komal learned would leave with her, to whichever school hired her next, and that school would not have paid for it either. A small community has its own danger too. Teachers who see each other every day can come to agree too easily, and agreement among people who like each other is not the same as being right. Komal was useful to the staffroom partly because she was new. She did not yet know what the others had stopped noticing. The staffroom also has ranks. When Komal later counted whom Kavitha called on, it was because Kavitha had asked her to watch. A guest teacher who told a permanent colleague to change, unasked, would be taking a risk with her next post.

38.3 The Mentorship Multiplier: What Passes From One Person to Another

In August Komal sat at the back of the Class 9 room in her free periods. In her own free periods, Kavitha sat at the back of Class 10. Neither corrected the other. Afterwards they talked in the corridor or, some evenings, on the telephone.

Once Kavitha stopped a story halfway. A girl near the window had frowned, and Kavitha asked the class what they expected to happen next. After the period Komal asked her, “How did you know to stop there?” Kavitha said she did not know. It was not modesty. She did not know.

Komal saw something else too. In the second week she showed Kavitha a page of her own notes. In five periods, Kavitha had asked the front three rows by name eleven times and the back two rows twice. That evening Kavitha looked through her own notebook. The back rows’ names were there less often as well. Nobody had sat at the back of her room for years, and the platform’s reports did not count whom she called on. A blind spot shows to someone sitting where its owner never sits. Kavitha did not promise to change. She said she would count for a month, and see. The newer teacher had seen what the older one could not. Mentoring is not only knowledge passing down. It is two people seeing from where the other does not sit, while the newer one becomes able to act alone.

A mentor does not upload information to a mentee. A mentor creates conditions in which the mentee can discover, through their own experience, what the mentor has already learned. A school or a tool can create some of those conditions too, and different people are responsible for each. The print edition claimed that mentoring was “the form that AI has the least capacity to replicate”. As of 2026 the evidence goes against that claim. Some of the conditions a mentor creates can now come from a machine, and they reach people no mentor would choose.

This call goes to the machine. Komal’s phone wrote her stories in Hindi whenever she asked, and many of them were good. They were clear, usually accurate and better shaped than Kavitha’s. In a trial at Harvard in the autumn of 2023, students in an introductory physics course took two lessons. They learned one from an AI tutor at home and the other in an active-learning class. With the tutor, their median learning gains were more than double, in less time.3 The tutor had been built to guide them step by step, and its authors said it should prepare students for class, not replace teachers. In Nigeria, in June and July 2024, secondary students spent six weeks in after-school English classes, working with a chatbot while teachers guided them. Their scores rose by about 0.3 standard deviations across English, knowledge of AI and digital skills, and by 0.23 in English itself.4 In Turkey, a trial of about a thousand students found the other side. Compared with students who had no chatbot, those given unrestricted answers did better in practice and 17 per cent worse in the exam taken without it. A version that gave hints instead of answers did no measurable harm.5 The models tested were those of 2023 and 2024. What the machine did in these trials, with teachers in the room in Nigeria, was explaining and correcting, the parts of mentoring the print edition thought safest.

The trials show something modest and useful: explaining, which most people mean when they say teaching, is no longer scarce. Read together, they say more about the conditions of learning than about which channel is better: a tutor built to guide, teachers in the room, hints instead of answers. They tested students learning a lesson. None tested a young teacher learning to read a silent class, and nobody yet knows what a machine would do there. One trial came nearer. In 2021, a free online programming course run from Stanford had about a thousand volunteer instructors. All could see automated feedback on how they responded to their students, and some, chosen at random, were reminded of it each week. Those who checked it built on their students’ ideas 13 per cent more often.3 Komal’s page of names was a count of that kind, made by a person at the back of the room. Whether a machine can learn to read a silent class is still open. This book will not settle that question by saying what machines cannot do.

Receipts, please

Two of the trials above carry most of the good news. Both were short, and in both, people built the conditions around the machine.

  • At Harvard, in the autumn of 2023, there were 194 students in one introductory physics course. Each took one lesson with the tutor and one in class, in consecutive weeks. They were tested straight after each lesson, and the report gives no later test of what they kept. The tutor’s accuracy rested on solutions written in advance.
  • In Nigeria, in June and July 2024, about 1,300 students in nine schools in Edo State were split at random. Half were offered after-school English classes, where teachers guided them in using the chatbot. On average, students spent about 13 hours in the programme over six weeks. The others continued their ordinary lessons with no extra class, though a few got into the sessions by mistake. So the gain belongs to the whole package: extra hours, teachers and the chatbot. The authors say so, though they doubt that the extra time alone explains it. They liken the English gain to a year and a half of ordinary schooling in Nigeria. That is a conversion, not a measurement. The paper follows the students no further than the end of the programme.

Emerging. A machine that explains can lift scores soon after a lesson, when people have built the conditions around it. Neither trial measured whether the gain lasts.

The phone’s stories were not always right, either. Kavitha told Komal what had happened in July 2025, before the platform came. A model had invented a compound found in neem, and a study of it in a real Indian journal. She found the error only because she went looking for the paper to show her class. After that, Komal checked the dates in her phone’s stories against the textbook before she told them.

So what could Kavitha show? She could show a practice, and it did not travel well. Soon after the platform came, a year earlier, she had begun writing in a notebook before each of its checks. She wrote the names of a few students and one word beside each, understood or not yet, and in the evening she compared her words with the results. The notebook had taught her that she read silence as confusion when it was often something else. She showed it to Komal. Komal’s Class 10 had no platform to check her guesses against. She tried three questions on a slip of paper at the end of a period. After two weeks she stopped, because reading the slips took the time she had for marking. The slips were her own way, not a copy, and dropping them when they cost too much was a judgement too. A practice is built for one room’s conditions, and it does not carry them with it.

An old text says what a teacher can honestly hand on. Near the end of the first part of the Taittirīya Upaniṣad stands a teacher’s charge to his pupil, given once the Veda has been taught. It is traditionally read as words for the end of his studies. In Swami Gambhirananda’s translation6 it begins: “Speak the truth. Practise righteousness. Make no mistake about study.” Further on comes the line this chapter needs. “These actions of ours that are commendable are to be followed by you, but not the others.” In Sanskrit it is yāny asmākaṃ sucaritāni, tāni tvayopāsyāni, no itarāṇi. Max Müller’s older translation keeps the same edge: “Whatever good works have been performed by us, those should be observed by thee”, and then, “Not others.”7 The teacher refuses to be copied whole. And a pupil who is unsure is not sent back to the teacher. The charge goes on: “Then, should you have any doubt with regard to duties or customs, you should behave in those matters just as Brahmanas do,” and it describes them as people “who may happen to be there and who are able deliberators”. On my reading, he is sent past the teacher, to other careful people. The charge says what such people are like: “who are not directed by others, who are not cruel”. Rank is there too, since they must be brahmins. But among the marks it adds are independence and gentleness. When Komal was unsure, she sometimes asked Sunita, who did not teach science at all.

The passage is a Vedic teacher’s charge about duty and ritual, not a theory of mentoring, and I borrow only its move. The teacher hands on a test, not a copy. In the last week of August, Kavitha did something like it. She told Komal three habits of hers not to copy. She had read silence as confusion for years before her notebook showed her. She let stories run past the bell, and the next period belonged to someone else. And that month she had paid a laboratory fee herself, to check a boy’s machine-written report on the water from the school’s hand pump. “Do not do that,” she said. “It should not have been mine to pay, and it is even less yours.” Generosity that covers for what an institution will not fund is admirable in a person and dangerous as a standard. The institution learns to count on it.

What she did not say, Komal watched. Every note to parents that a machine drafted for Kavitha, Kavitha read before it went out. Before the platform came, she had made a rule: nothing would leave with her name on it that she had not read and could defend. Komal saw her keep it on ordinary days, when nobody would have known. A machine keeps rules too, often more steadily than a tired teacher. But its rules were set by others, and it does not bear the cost of keeping them. What Komal watched was a person keeping her own rule when it cost her an evening. A learner takes in what a mentor does under pressure, whom she credits and which rules she bends, as well as what she says. That is why the habits Kavitha named mattered, and why the ones she did not name mattered too.

MacIntyre puts the newcomer’s position another way. To enter a practice, he argues, is first to accept the authority of its standards, and the poorness of one’s own performance as they judge it.1 The two lamps agree on one thing that matters here. Both put a standard above the teacher: the blameless acts in the Upaniṣad, the practice’s standards in MacIntyre. They part on who judges, and when. MacIntyre’s newcomer submits first and judges later, inside a practice whose standards change as its people argue about them. The Upaniṣad’s pupil judges at the door, by a code fixed long before him and by the careful people present. Komal, at the back of the room, needed both. She needed Kavitha’s standards before she could judge them, and Kavitha’s word about which of her acts to leave behind.

The same passage also says, “Let your teacher be a god unto you.” The limit follows almost at once: “The works that are not blameworthy are to be resorted to, but not the others.” On my reading, honouring a teacher does not mean copying all he does. The charge also asks the pupil to bring his teacher wealth before he leaves. And it was spoken in a school whose door was shut to most people. By the rules that governed it, Vedic study of this kind was for boys of the three twice-born classes.8 That door is part of the text, and this book does not keep it. The Mahābhārata remembers Droṇa, the teacher of the Kuru princes, who would not teach Ekalavya, a prince of the Niṣādas, a forest people, and later asked for his thumb as his fee.9 Mentorship goes first to those already let in.

It still does. Komal got Kavitha because she happened to be posted to Kavitha’s school. A guest teacher of science in a village school in the same district has nobody down the corridor. She has a phone. Komal’s own terms were those of her kind of post. In 2024–25 about 79,000 guest teachers worked for Madhya Pradesh’s school education department.10 Their pay was monthly, between ₹10,000 and ₹18,000 depending on the level of the school, after a doubling announced in September 2023.11 From July 2025, under an order of that June, a day that the attendance app on their own phones did not record was a day without pay.10 Kavitha’s post is permanent. Komal’s hours at the back of a room were her free periods, and the evenings were both women’s own. Nobody paid for any of it. A mentor helps someone grow. A sponsor opens doors. Kavitha could teach Komal, but she could not give her a post; that was not hers to decide.

Now the old chapter’s claim can be judged. “This is the mentorship multiplier. It does not produce copies. It produces collaborators.” That is true only where the receiver practises, and only for what is handed on well. A multiplier multiplies whatever it is given. Blind spots travel as easily as attention, and so does the advantage of being near. Had Komal copied Kavitha whole, she would have copied the back rows too.

So what passed between them in August? Some of it can be named. Komal learned to check her phone’s dates. Kavitha learned whom she did not ask. Each learned it by sitting where the other did not. None of it was attention, handed over like a book. Each had to do the thing herself, again, in her own room, before it was hers. The capacity to see clearly and the capacity to help others see are not the same thing. The parts of a practice also travel differently. Explaining, a task, travels easily and is already partly a machine’s. The eye that knows when to stop a story travels only by forming another person, at someone’s cost in hours. Responsibility stays with whoever can still be asked, and moves as the practice moves.

38.4 Institutional AI-Proofing: Universities, NGOs, and Industry Bodies

The title keeps a word that needs a limit. No institution is proof against machines. What an institution decides is which seats it keeps for a person, which rungs it keeps for beginners, and who pays for both. Institutions should not keep people in a task for its own sake. They should keep the human capacities, rights and duties their purposes need. A rung, in this book, is the step where a newcomer learns by doing real work, with room to fail and someone to correct her.

Kavitha’s institutions decided all of that for her. The head teacher called the arrangement a good idea and gave it no period. The district fills the school’s second science post one year at a time. The district office chose the platform. And the committee in Bhopal did what committees do with a practice. In the second week of October its draft guidance reached Sagar, with her teaching stories turned into a module called “Teaching Through Stories”, a rubric of five points each for a story’s completeness, its link to the syllabus and its use of the platform, and a line crediting her. She wrote back through the district office, asking that two lines change or that her name come off them.

One detail shows what travels. The stories Komal’s phone wrote always reached an ending, because that was what she asked it for. Kavitha’s best lessons had ended on a question. By the rubric’s first line, Komal’s phone would score higher than Kavitha’s stories.

MacIntyre’s point about judges cuts here. Suppose internal goods can be recognised only by people who have taken part in a practice. Then a rubric must score what can be seen from outside a lesson: whether a story is complete, whether it is linked to the syllabus, whether a tool was used. Those are marks a stranger can check. They are not nothing. A state that must judge thousands of classrooms it will never sit in needs such marks. But they are not what Kavitha’s stories were for. Nobody has measured what her stories do, either. A question carried down a corridor is the only evidence this chapter has. A state is right to want more before it asks thousands of teachers to tell such stories. A practice’s standards change as its people argue about them. A rubric fixes them on a date. A module can keep a method and lose its purpose: stories told because a rubric scores them are not stories told so that children ask how anyone found out. Kavitha’s rule about her name is one defence against that loss.

MacIntyre saw why this happens, and why it cannot simply be refused. A practice, he argued, cannot survive for long without institutions to carry it. Institutions deal in money, power and status, and so the institution that carries a practice is always a danger to it.1 A state cannot hand on one teacher’s practice to tens of thousands of teachers. It can hand on a document, and the conditions in which others might build practices of their own. That is the true half of the module. What the state puts in the document, and what it pays for besides, is its own decision, and the state answers for both. It is accountable for how it chose her stories, what evidence it used, what it measured, whom it asked, and how a teacher can challenge or revise the module. One teacher’s stories are a case to test, not a finished model.

What an institution can do and no mentor can is pay for the hour. It can keep a post past one year. It can protect the rung: a first attempt, an explanation, a widening share of the work, and feedback, which this book has called the institution’s part in building depth. A pharmaceutical company in this book gave its quality analysts two hours a week of working time, and let its reports to head office show the cost, to write their own reasons before reading the machine’s. Nobody gave Komal and Kavitha an hour. Machines raise the stakes. They can make an experienced teacher faster while removing the work through which a beginner once learned: the first lesson plans, the first marking, the first stumbling explanations. A young teacher who learns to run a platform may not learn to read a room. As the last chapter argued, the better a machine does a task, the less often people tend to do it. Then it gets harder to tell when the machine is wrong. Where old rungs go, someone has to build new ones on purpose.

What would it take for their arrangement to be the school’s, and not a favour two women did each other? A period in each week’s timetable, given by the head teacher. A post that does not end with the year, decided by the district. Some way for the hours to count in the annual appraisal, which the district had already used to note who had not finished its certification course in AI. None of it needs a new machine. Each is someone’s decision, and each has a cost that someone else would carry: a period taken from a subject, a post funded for longer. A school that says mentoring matters and gives it no hour has decided that it does not matter enough to pay for.

Universities are the oldest case. The University of Bologna dates its founding, by convention, to 1088.12 Universities are credentialling institutions, research ecosystems, peer networks, laboratories for physical experimentation, and engines of socialisation. They have lasted partly because they move last. In November 2012 the New York Times called it “The Year of the MOOC”, the massive open online course.13 Six years later two researchers, studying Harvard’s and MIT’s open courses, found that low completion rates had not improved. The growth in participation had been “concentrated almost entirely in the world’s most affluent countries”.14 The universities’ slowness looked wise. It had kept standards and peer review, and it had also kept out people who could not afford to come in. The figures showed that enrolling is not finishing. They did not measure learning, and they did not show that the universities were right. Clayton Christensen and Henry Eyring had argued in 2011 that a university’s habits are set by its founders and its history, and work like inherited DNA.15 That is part of why universities move last, and part of why they last. An institution can last for centuries and still fail the people it was meant to serve. Whether moving last is wise again, now that a machine can tutor a student through a lesson, nobody yet knows.

The same holds for the charity that trains teachers and the trade association that trains inspectors. Each decides whether its juniors still do the work they learn from, and who pays for their hours. A charity can train teachers while relying on staff hired a year at a time. A trade body can keep standards high and its door narrow. Each is responsible for whom it lets in, as well as for what it keeps.

One projection belongs here. It is mine, and it is offered as a projection, not as evidence. By 2030, I expect that somewhere between a quarter and nearly all of the students in schools like Kavitha’s will be able to ask a machine about nearly anything in the Class 10 science syllabus and get a patient, mostly correct explanation in Hindi. My central guess is most of them. At the low end are only those with a phone and data of their own, the only ones who can ask at any hour. In 2024, almost 90 per cent of 14-to-16-year-olds in rural India had a smartphone at home. Of all of them, 82.2 per cent could use one, and of those, 57 per cent had used it for something educational in the week before they were asked. Of those who could use one, only 27 per cent of 14-year-olds and 37.8 per cent of 16-year-olds had a phone of their own.16 At the high end is nearly every student, through phones at home and devices at school. What would show me wrong is a survey around 2030 finding that fewer than half of rural 14-to-16-year-olds can use a phone for their studies. So would tests finding such explanations in Hindi wrong too often to set as homework. Trials in Indian government schools that find no lasting gain from such tutoring without a teacher’s structure would not show the projection wrong; they would show that it matters less. Either way, the question of who answers for what is handed on stays where it is. Only the practical reason to pair Komal with Kavitha for explanations would change. That reason is not small. Access is not use, and use is not learning. A phone at home may be a parent’s, data costs money, and a mostly correct explanation still needs someone who can tell when it is wrong. How widely and how evenly such help arrives would still change much of what a school must do.

38.5 The Family and the Next Generation

Nobody has a playbook for this. Any parent or educator who claims certainty is selling something.

In September, at the meeting with parents after the quarterly examinations, a father asked Kavitha whether his daughter should use a chatbot for her homework. He had left school early, and nobody at home could help her with science. For her, the chatbot might be the first patient tutor she had ever had. Kavitha told him what was known. In one large trial, students who could ask for answers did better in practice and worse in the examination, and students who got only hints did not suffer. She told him what was not known: what years of such help would do to a girl of fourteen. Nobody knew. She did not tell him what to do. It was his household’s decision, and she said so. Deciding well takes more than the right to decide: knowing the machine’s limits, time to sit beside her, data he can pay for. Companies, schools and the state set much of that; he does not. What he could watch was whether she could explain an answer in her own words, and spot a wrong one.

The print edition went further. It called the wish to remove friction from children’s lives “the most dangerous instinct a parent can indulge”. The trial points somewhere narrower. The harm came from handing over the answer, not from the help. That is a smaller claim, and a parent can use it. A parent can ask a child to try each problem first, and to ask the machine for hints rather than solutions.

A family carries the next generation in another way too. The evenings Kavitha spent on the telephone with Komal came out of a household, where her husband and her mother made room for them. Raising children among these machines is a larger question than one meeting with parents could hold. In schools like hers, the next generation of teachers is often hired a year at a time. One of them was sitting at the back of her room. A chatbot can bring a girl in Sagar the explanations a richer child once got from a tutor. The tutor’s contacts, the advice on which stream to choose after Class 10 and the safe place to fail do not come with it. Spreading explanations is not spreading opportunity.

38.6 Work That Outlives Its Maker

No legacy is AI-proof, or proof against anything. Most are small.

Late in September, after the quarterly examinations, Kavitha spent two days in Bhopal, presenting her stories to the committee. Komal took her Class 9 periods, and the platform taught its two periods as usual. In one period Komal told a story her phone had written the night before. In 1909 two researchers in Ernest Rutherford’s laboratory in Manchester fired tiny charged particles at a sheet of gold thinner than paper, and a few of them bounced back.17 When Kavitha returned, a girl told her about it. Then she asked how anyone could know something was there when nobody could see it. It was the kind of question Kavitha’s stories were meant to draw. A machine’s story had drawn it. That does not show that her stories were unnecessary. It shows that what a story does in a class does not depend on who wrote it. What depended on a person was what came after. That is where a teacher’s practice now lives: not in the story, but in what she does with the question it draws. Kavitha replied to the girl the next day. No rubric would ever see that reply, and Kavitha could be asked to explain it.

What will outlast Kavitha? A module may carry her name, or it may not, depending on whether a committee changes two lines. Four stories sit in a state file. One day they may also sit, without her name, among the texts a model learns from; many people’s judgement already trains machines that never carry their names.18 A question sits in some of the several hundred students she has taught in thirteen years, and their next teacher pays for it in periods. And something may sit in Komal, which nobody can trace. These are different kinds of continuing: a habit in students, a practice in a colleague, a document in a state file, a trace in a machine’s training.

Her rule about her name had begun with notes to parents. It turned out to be a rule about legacy too. Her stories could outlast her in two opposite ways. On the module, her name might travel with lines she could not defend. In a model’s training, her work might travel with no name at all. The first gives a person credit for what she cannot stand behind. The second uses what she made without her name or her leave. Her rule asks for something between them: a name only on what she can defend, and a way to take it off. Chapter 20 kept three senses of “yours” apart: what the law lets you stop others doing, where a work came from, and who answers for it. A name on a module speaks to the second sense. Whether her stories may train a model is a question of the first sense. Her rule is mostly about the third. That is all a maker can ask of what outlives her, and it is more than most makers get.

The Upaniṣad’s charge has an image for this. “Having offered the desirable wealth to the teacher,” it says, “do not cut off the line of progeny.” In Müller’s words, “do not cut off the line of children!” The verse is about a household’s children, not pupils, and I borrow only its image: a thread that continues, not a monument that stands. MacIntyre’s definition ends in the same place. Through a practice, people’s powers to achieve excellence, and their idea of the ends and goods involved, “are systematically extended”.1 Extended, not preserved. Whether Komal extends Kavitha’s practice or replaces it with her own is not Kavitha’s to judge. Continuity is not copying. A practice lives by being changed by those who take it up, and that sometimes means dropping parts of it. “Not the others” runs both ways. Komal owes Kavitha’s practice nothing she cannot defend, once she has tried it honestly.

The deepest legacies are not monuments. They are transmissions. They are also smaller than the word suggests. A question carried down a corridor. A line at the foot of a module. A habit somebody else pays for. None of this makes a life worth less for leaving nothing anyone can trace. Most lives leave little that can be named. This book has refused to measure people by being needed, and it will not measure them by being remembered.

And who answers for them? Kavitha answers for what she handed on while she can still be asked about it. To own an act, this book has argued, is to answer for the part of it that was yours, not to claim to be its only cause. Once the committee prints its rubric, the committee answers for the rubric. Once Komal tells a story to her class, Komal answers for the telling. When Kavitha can no longer be asked, the responsibility does not vanish. It rests with those who adopted, changed or used what she left. What Kavitha owes the people who take up her practice is what the old teacher gave his pupil. It is a plain account of which of her acts to keep, and which to leave.

Ending: At the Back of Komal’s Room

In the first week of October, Kavitha sat at the back of Komal’s Class 10 in one of her free periods. A year earlier she had sat at the back of her own room, watching a platform teach her class. Now she watched a person teach the class she had taught.

Komal was teaching heredity. A boy asked how they found out. Komal told them about a friar in Brno who grew peas in his monastery’s garden for eight years and counted what he saw. His paper was largely ignored until it was rediscovered in 1900.19 She had asked her phone for the story that morning and checked the dates against the textbook. She told it well. Then she did what Kavitha would not have done. She finished the story and gave them the answer. Then she turned to a girl who had not spoken all period and asked her a question directly. The girl answered it.

Kavitha could not tell whether that was worse than what she would have done, or better. She could not tell whether any of it had come from her, or from the phone, or from Komal herself. Komal had been a teacher before she ever sat at the back of Kavitha’s room. Her notebook was on her knee. She did not write in it.

Practice

What they took from you. Ask someone newer, who has worked beside you for a while, to name two or three things they have picked up from you. Ask before you say what you hoped they would pick up. Compare their list with your hope. For each thing they name, ask yourself whether you could defend it to them. Where you cannot, tell them so and say why, as the old teacher told his pupil to keep his good acts and leave the others. Note whose hours the conversation takes, yours and theirs, and whether either of you was given them. It costs one conversation and some pride. It promises nothing about work, a job or a legacy. It shows you what is actually being handed on.

Words to keep

  • ācāryadevo bhava: Sanskrit, “let your teacher be a god to you”; part of the Taittirīya’s parting charge, quoted here with the exclusions of its school named and refused.
  • guest teacher: in Madhya Pradesh, a teacher engaged for one school year at a time and paid by the month, without a permanent post.
  • internal goods: MacIntyre’s term for the goods that can be had only by taking part in a practice, as against external goods such as money, status or prizes.
  • no itarāṇi: Sanskrit, “not the others”; the teacher’s limit on what the pupil should copy.
  • practice: in MacIntyre’s sense, a shared, complex human activity with its own standards of excellence, through which goods internal to it are achieved, and people’s powers and their idea of its goods are extended.
  • rung: the step where a newcomer learns by doing real work, with room to fail and someone to correct her.
  • sponsor: someone who opens doors for another person, by recommending her, introducing her or giving her a chance to show what she can do; a mentor helps her grow.
  • sucaritāni: Sanskrit, “good conduct”, the commendable acts of the teacher that the pupil should follow.
  • Taittirīya Upaniṣad: one of the principal Upaniṣads, of the Black Yajurveda; the eleventh section of its first part is the teacher’s charge to a pupil.
  • teaching stories: Kavitha’s name for stories about how someone came to know something, chosen for the moment a class has stopped understanding.
  • ubuntu: a southern African idea, often summed up as “a person is a person through other persons”.

Sources

  1. Alasdair MacIntyre, After Virtue: A Study in Moral Theory, 2nd edition (Notre Dame, IN: University of Notre Dame Press, 1984; first published London: Duckworth, 1981), ch. 14, “The Nature of the Virtues”: the definition of a practice, p. 187 (quoted); the chess example and the judges of internal goods, pp. 188–189 (paraphrased); entering a practice and accepting the authority of its standards, p. 190 (paraphrased); practices and institutions, p. 194 (paraphrased).
  2. Alasdair MacIntyre and Joseph Dunne, “Alasdair MacIntyre on Education: In Dialogue with Joseph Dunne,” Journal of Philosophy of Education 36, no. 1 (2002), p. 5, as quoted in Pádraig Hogan, “Teaching and Learning as a Way of Life,” Journal of Philosophy of Education 37, no. 2 (2003). Copies of the full sentence differ after these words, so the rest is paraphrased. Page as cited by Hogan; the printed page itself not seen.
  3. Gregory Kestin, Kelly Miller, Anna Klales, Timothy Milbourne and Gregorio Ponti, “AI tutoring outperforms in-class active learning: an RCT introducing a novel research-based design in an authentic educational setting,” Scientific Reports 15 (2025): 17458: 194 students in Harvard’s introductory physics course for life-science students, autumn 2023; a crossover design of two lessons; median learning gains for the AI group more than double, with a median 49 minutes on task; the tutor scaffolded with pre-written step-by-step solutions; the authors propose it for use before class, not as a replacement for in-person teaching. On timing (for “Receipts, please”, 38.3): “Following each lesson, students completed post-tests to measure content mastery”; no delayed test is reported; “accuracy of our AI tutor relied on pre-written answers” (Discussion; full text at PMC12179260, checked 31 October 2026). On automated feedback to instructors: Dorottya Demszky, Jing Liu, Heather C. Hill, Dan Jurafsky and Chris Piech, “Can automated feedback improve teachers’ uptake of student ideas? Evidence from a randomized controlled trial in a large-scale online course,” Educational Evaluation and Policy Analysis 46, no. 3 (2024): 483–505 (online 8 May 2023): Code in Place, Stanford’s free five-week online programming course, spring 2021; “N = 1,136 instructors”; all instructors could open the feedback through the course website, and those assigned at random to treatment were sent a weekly email reminder about it; the feedback “improves instructors’ uptake of student contributions by 13%”, an estimate for those who checked it; only suggestive evidence on students’ satisfaction and assignment completion.
  4. Martín E. De Simone, Federico Tiberti, Maria Barron Rodriguez, Federico Manolio, Wuraola Mosuro and Eliot Jolomi Dikoru, “From Chalkboards to Chatbots: Evaluating the Impact of Generative AI on Learning Outcomes in Nigeria,” World Bank Policy Research Working Paper 11125 (2025): Microsoft Copilot (GPT-4) for first-year senior secondary students, six weeks, 0.31 standard deviations on an assessment covering English, knowledge of AI and digital skills, and 0.23 in English; World Bank Education blog, 9 January 2025: an after-school programme in Edo State, June to July 2024, “about 0.3 standard deviations”; World Bank Education blog, 18 September 2024 (“From chalkboards to chatbots in Nigeria: 7 lessons”): after-school English classes in computer labs twice a week, in which “the teachers guided students in using the LLM”. On the design (for “Receipts, please”, 38.3): the working paper (dated May 2025; documents.worldbank.org, checked 31 October 2026): “Initially, 657 students were assigned to the treatment group and 671 to the control group”, across nine schools; the control group “did not receive any intervention but continued their regular learning in the classroom”, though “some control group students inadvertently gained access to the after-school sessions” (section 3.1); the effect includes “additional instructional time and the interaction with the chatbot with teacher support” (section 3.1); “We have reasons to believe that the effects are not driven solely by the additional time with teachers” (section 4.3); “Our ITT effect of 0.238 standard deviation in English” is likened to 1.5 years of “business-as-usual” schooling in Nigeria (section 4.1); the standardised assessment came at the end of the six weeks. Martín E. De Simone and others, “Addressing the learning crisis with generative AI: lessons from Edo State in Nigeria,” World Bank, Let’s Talk Development blog, 25 June 2025: “students spent an average of 13 hours total over six weeks”; “We are careful to interpret the results as reflecting the full package”.
  5. Hamsa Bastani et al., “Generative AI without Guardrails Can Harm Learning: Evidence from High School Mathematics,” Proceedings of the National Academy of Sciences 122 (2025): e2422633122: about 1,000 students in Turkey; unrestricted access raised practice scores by 48 per cent and lowered unaided exam scores by 17 per cent, both against a control group without access; a hint-giving tutor raised practice by 127 per cent with no measurable exam harm (the harm “largely mitigated”). See Chapter 4.
  6. Taittirīya Upaniṣad I.11.1–4 (Śikṣāvallī, the eleventh anuvāka), trans. Swami Gambhirananda, Eight Upaniṣads, vol. 1 (Calcutta: Advaita Ashrama, 1957), as given at shastras.com (I-xi-1 to I-xi-4), checked 23 October 2026, and in two further copies of the same translation (hinduonline.co; sites.google.com/site/upanishadscripts), which print “Practise”, checked 29 October 2026; Sanskrit text and transliteration from shlokam.org, Taittirīya 1.11.
  7. The Upanishads, Part II, trans. F. Max Müller, Sacred Books of the East 15 (Oxford: Clarendon Press, 1884), “Taittirîyaka-Upanishad” I.11 (sacred-texts.com). Müller’s sentence on the teacher’s good works carries a dash, so it is quoted in two fragments.
  8. Encyclopaedia Britannica, “upanayana” (accessed 29 October 2026): “Hindu ritual of initiation, restricted to the three upper varnas, or social classes,” marking a boy’s entrance upon the life of a student (brahmacārin).
  9. The Mahabharata of Krishna-Dwaipayana Vyasa, trans. Kisari Mohan Ganguli, Adi Parva, Section CXXXIV (Critical Edition 1.123); see Chapter 20.
  10. “Bhopal: Guest Teachers Protest Mandatory E-Attendance,” Free Press Journal, 29 June 2025: about 79,000 guest teachers in the school education department in the previous academic year; monthly pay of ₹10,000 to ₹18,000 by grade; under an order of 26 June 2025, no honorarium for a day whose attendance is not recorded on the app; guest teachers use their own phones; “No E-Attendance, No Salary: Government Warns Of Legal Action Against Guest Teachers’ Protests,” Free Press Journal, 17 July 2025: app attendance from the start of the session on 1 July; honorarium lost for unrecorded days from 18 July.
  11. “CM Chouhan doubles honorarium, recruitment quota for guest teachers in poll-bound Madhya Pradesh,” The Week, 2 September 2023: monthly honorarium raised from ₹9,000 to ₹18,000 (grade 1), ₹7,000 to ₹14,000 (grade 2) and ₹5,000 to ₹10,000 (grade 3).
  12. University of Bologna, “Our History” (unibo.it, accessed 23 October 2026): “1088, conventionally referred to as the year in which the Studium of Bologna was founded”.
  13. Laura Pappano, “The Year of the MOOC,” New York Times, Education Life, November 2012 (dated 4 November 2012 in its URL).
  14. Justin Reich and José A. Ruipérez-Valiente, “The MOOC pivot,” Science 363, no. 6423 (11 January 2019): 130–131: low completion rates had not improved over six years; growth “concentrated almost entirely in the world’s most affluent countries”.
  15. Clayton M. Christensen and Henry J. Eyring, The Innovative University: Changing the DNA of Higher Education from the Inside Out (San Francisco: Jossey-Bass, 2011); their idea of an institutional “DNA” shaped by founders and history, paraphrased.
  16. Pratham, Annual Status of Education Report (Rural) 2024 (released January 2025), as summarised by the Ministry of Education, “Leap in Rural School Enrollment” (PIB, 4 February 2025): “Almost 90% of both girls and boys report having a smartphone at home”; “82.2% of all children in the 14-16 age group reported knowing how to use a smartphone”; of these, “57% reported using it for an educational activity in the preceding week”; “Of the children who could use a smartphone, 27% of 14-year-olds and 37.8% of 16-year-olds reported having their own phone”.
  17. H. Geiger and E. Marsden, “On a Diffuse Reflection of the α-Particles,” Proceedings of the Royal Society of London A 82 (1909): 495–500; E. Rutherford, “The Scattering of α and β Particles by Matter and the Structure of the Atom,” Philosophical Magazine, 6th series, 21 (1911): 669–688.
  18. Mary L. Gray and Siddharth Suri, Ghost Work: How to Stop Silicon Valley from Building a New Global Underclass (Boston: Houghton Mifflin Harcourt, 2019); see Chapter 20.
  19. Gregor Mendel, “Versuche über Pflanzen-Hybriden,” Verhandlungen des naturforschenden Vereines in Brünn 4 (1866): 3–47, read on 8 February and 8 March 1865; experiments 1856 to 1863 at St Thomas’s Abbey, Brno (“after eight years’ pursuit”, in Mendel’s own words, as translated at MendelWeb); the work was rediscovered independently in 1900 by Hugo de Vries and Carl Correns.

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 chapter is about what happens “when individual consciousness meets other consciousness”; that “Individual consciousness is the foundation. Collective consciousness is the structure.”, with legacy as “the slow, imperfect, irreplaceable transmission of awareness from one life to the next”; that individual moats should become “collective moats”, that “Community is a moat. Belonging cannot be automated. Both claims stand.”, and that a shared standard is something “No platform can generate” and “No algorithm can fake”; that “machines do not know they are part of a system. Humans can.”; that a group’s understanding is beyond AI “because AI does not engage with a problem. It processes inputs.”; that mentoring is “the form that AI has the least capacity to replicate”; that an institution’s purpose must move “from content delivery to consciousness development”, since consciousness “cannot be distributed like a software update”; that the next decade will turn on “the competition between conscious ecosystems and unconscious ones”; that removing friction is “the most dangerous instinct a parent can indulge”; that a parent raises someone “who will carry consciousness forward”; and that people will remain “not merely relevant but essential”. It set the chapter at a conference in Singapore in March, attended by the book’s five main characters, which no date in the book’s present can hold, and closed on a coincidence of recognition between two of them. It gave a test lead a nine-year-old daughter and a junior taught at her elbow; gave Kavitha forty-seven students, a “methodology” piloted “across three states” and adopted by a state council, results “performing better on assessments” and a mentoring pilot approved six weeks after a meeting; gave a designer a studio of six and a conversation that no longer happens; gave an operations head a notebook copied from a magazine article about another character’s private notebook; named a real company as a cloth exporter’s investment and gave his network fourteen members and a foundry owner in Tiruppur; dated universities to “five hundred years”; and listed eleven “qualities of consciousness”.
  • Section titles in the print edition: 38.1 “From Individual Moats to Collective Moats”, 38.2 “Building Conscious Communities”, 38.3 “The Mentorship Multiplier: Consciousness Transmitted”, 38.5 “The Family and Next-Generation Playbook” and 38.6 “Your AI-Proof Legacy”; 38.4 keeps its print title.
  • What I now hold: nothing of consciousness is handed on. What one person can show another is a practice, in MacIntyre’s sense, which forms only in the one who practises it, on occasions someone pays for. As of 2026, machines explain and correct well in trials, and they reach people no mentor would choose. What a teacher can add is the sight of someone keeping her own rules when they cost her, and her word about which of her acts not to copy (the Taittirīya’s “not the others”). Mentorship goes first to those near and already admitted, and its hours are usually unpaid. A practice needs others, and their attention can also make everyone alike. An institution can carry a document, not a practice; it decides which seats and rungs it keeps and who pays for them, and it answers for both. Universities last partly because they move last, which has been both wise and slow. Legacies are mostly small and untraceable, and the next person may pay for them. The maker answers for what she handed on while she can still be asked; those who take it up answer for what it becomes. A shared standard is a moat only when its first use is to keep newcomers out. A practice outlives its maker only when others can enter it, do real work, be corrected, question it and in time answer for it; machines can remove the work through which beginners learn, so new rungs must be built on purpose. Students carry a habit of asking, not a practice, and a legacy is also what a successor must carry. Mentoring runs both ways; a mentor is not a sponsor; generosity that covers for an institution is not a standard. A rubric fixes on a date the standards a practice keeps arguing about, and a committee answers for how it chose, what it measured and how a teacher can challenge its module. Access is not use, and use is not learning; spreading explanations is not spreading opportunity. Credit, the law’s control and answering are different senses of “yours” (Chapter 20). Continuity is not copying, and a successor may refuse what she cannot defend. A life is not worth less for leaving nothing anyone can trace.
  • What changed my mind: the Prologue’s rule against defining people by what machines cannot yet do; the tutoring trials of 2023 and 2024 at Harvard, in Nigeria and in Turkey; MacIntyre’s account of practices, internal goods and institutions, and his later view that teaching is not itself a practice; the Taittirīya’s charge, which limits the teacher’s authority and names the school’s closed door; the terms of guest teachers in Madhya Pradesh; Reich and Ruipérez-Valiente on the MOOC; the book’s own canon, whose dates could not hold the conference; the audit; and two outside readings of an earlier version of this chapter, in October 2026.
  • What survives: the chapter’s three subjects (mentorship, institutions, legacy) and the collective; the definition of a community as “a group of people who share a practice, hold each other accountable to that practice, and generate understanding together that none of them could generate alone”; “A mentor does not upload information to a mentee. A mentor creates conditions in which the mentee can discover, through their own experience, what the mentor has already learned.”; “This is the mentorship multiplier. It does not produce copies. It produces collaborators.”, with its condition; “The capacity to see clearly and the capacity to help others see are not the same thing.”; universities as “credentialling institutions, research ecosystems, peer networks, laboratories for physical experimentation, and engines of socialisation”; “Any parent or educator who claims certainty is selling something.”; “the deepest legacies are not monuments. They are transmissions.”; Kavitha’s teaching stories, now handed to a teacher hired for a single year.

↑ 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.