# AI-Proof: What Machines Can Never Be — Volume III: Transformation, Part VI: The Inner Revolution, Chapter 32: The Examined Life in the Age of Machines

By Nithin Palal · 21 November 2026 · 43 min read
Section: Artificial Intelligence · Series: AI-Proof: What Machines Can Never Be
Source: https://nithinpalal.com/ai/ai-proof/chapter-32-the-examined-life-in-the-age-of-machines

> A man trained in Zen is paid to sit with founders; a model now asks his questions too. With 300 Likes, a computer matched people’s own personality scores better than spouses did. An examined life is an account, and it can become a show.
*Part VI looks at what happens inside a person when machines take over work. After fear, fragility and a builder’s debt, this last chapter of Part VI asks what is left to examine when machines can describe us.*

← Previous: [Chapter 31: The Moral Responsibility of Builders](https://nithinpalal.com/ai/ai-proof/chapter-31-the-moral-responsibility-of-builders) · Series hub · Next: [Part VII: The Conscious Blueprint (overview)](https://nithinpalal.com/ai/ai-proof/part-7-the-conscious-blueprint) →

**A man trained in Zen is paid to sit with founders; a model now asks his questions too. With 300 Likes, a computer matched people’s own personality scores better than spouses did. An examined life is an account, and it can become a show.**

**In Brief: For readers who wonder what is left to examine when a machine can describe them, question them and pause before it answers. Chapter 32 asks what is left of the examined life when machines can describe us, question us and pause before they answer. A man trained in a Zen monastery in Kyoto is paid to sit with founders and is not sure it is honest. A test lead in Bengaluru buys a notebook, and pauses for four seconds in a meeting about her own machine’s mistake. Two voices frame it. Plato’s Socrates examined himself and others in public and took no fee. Lal Ded, a Kashmiri poet of the fourteenth century, kept herself untouched by blame and praise. One asks for reasons given to people with a right to ask; the other refuses to perform for a crowd.**

**As of October 2026, a machine can describe you, question you and pause before it answers; to examine a life is to give an account of how you live to people who may ask why, and to change what you cannot defend, and the examining becomes a performance once it is done for show or for gain.**

## Opening: Morning in Higashiyama

Most mornings, before the street outside is awake, Sonoda Tetsuo sits on a cushion and watches the light come into a small garden. His house is in Higashiyama, on the eastern edge of Kyoto, below the hills. He is sixty-one. He has sat like this, in one room or another, since he was twenty-three.

At twenty-three he entered a training monastery of the Rinzai school of Zen in Kyoto, and he stayed seven years. He left at thirty, in 1995, and spent most of the next twenty years building software products in Tokyo and San Francisco. For about a decade now he has done something else. Technology companies pay him, and pay him well, to sit with their founders and ask them questions nobody has asked them.

He is very good at this. He is also not entirely sure it is honest.

The doubt is older than the business. At their last meeting his teacher told him that his practice had become beautiful. It was not praise. Beautiful practice, his teacher said, is practice that has learned to perform itself. You sit well. You breathe correctly. You answer the koans, the teaching riddles of Zen, with intelligence. But you are watching yourself sit. You have turned awareness into an accomplishment, and accomplishment is the opposite of what we are doing here.

He left the monastery that month. He told himself that he was choosing the world over the cloister, and that the practice would go with him. Both were true. Also true, and for thirty-one years neither accepted nor denied, is that his teacher may have been right. What his clients pay for may be the performance his teacher saw: the stillness, the attention, the questions that seem to come out of silence rather than strategy. He does not know. Depending on the day, the not-knowing is either the most honest thing about him or a convenient way of avoiding an uncomfortable answer.

His clients do not know he does not have the answer.

This book has met a doubt like his before. U Nandaka, a monk who teaches meditation outside Yangon, admits to years he wasted mistaking a quiet mind for awareness. Sonoda’s doubt has one thing the monk’s lacks. He is paid for what he doubts.

As of October 2026, three things that people like Sonoda once offered can be had from a machine. A model that reads the traces a person leaves online can describe her personality. A model can ask a founder searching questions at any hour, for almost nothing. A reasoning model pauses before it answers, and on some kinds of task the extra working makes the answer better. So this chapter asks: **When a machine can describe you, question you and pause before it answers, what is an examined life, and what keeps it from becoming a performance?**

The print edition answered with a strategy. “In the age of artificial intelligence,” it said, “self-knowledge is a competitive advantage.” The willingness to live with Sonoda’s question was “the foundation of everything the AI age demands from a human being.” This edition withdraws both. Describing yourself is a task, and friends, records and machines do parts of it better than you can. The examined life, in the sense of the man who named it, is not a description. It is an account of how you live, given to people who may ask why, and changed where it cannot be defended. It promises no advantage, no calm and no job. Nothing can guarantee that it stays honest. What helps is giving the account to people who will see you tomorrow, and refusing to let the practice become a thing to be admired.

The chapter keeps its title, which needs one correction. The examined life was not made for an age, and no age demands it. Socrates chose it, and in 399 BCE an Athenian jury condemned him to death. What machines change is what the examining can be helped by: a profile, a question, a summary of a month of notes. They also change what it can be mistaken for: a pause, a daily streak (an app’s count of days in a row), an insight a coach supplied.

## 32.1 Self-Knowledge Is Not a Strategy

The print edition’s argument went like this. Judgement depends on knowing your biases well enough to correct for them. Taste depends on telling what you value from what you were trained to value. Trust depends on being honest about a self you have examined. So every quality a person brings beyond output depends on self-knowledge, and the people who know themselves will do better when machines match their output.

The first step fails on the evidence. Knowing that biases exist does not show you your own. In 2002 the psychologists Emily Pronin, Daniel Lin and Lee Ross reported three studies in which people saw the working of biases “much more in others than in themselves”.¹ They called this the bias blind spot. In 1977 Richard Nisbett and Timothy Wilson showed that people often give confident reasons for their choices that had little to do with what actually moved them.² Earlier in this book, Renata Oliveira, a cardiac surgeon in São Paulo, saw her own pride only after it had cost a patient a week of needless pain. Seeing a pattern in yourself does not make your next decision wise.

Other people sometimes see us more clearly than we see ourselves. In 2010 the psychologist Simine Vazire reported that friends judged traits such as intelligence and creativity more accurately than the people themselves did. People were better judges of their own inner traits, such as anxiety.³ Each side has a view the other lacks.

So do machines. In 2015 three researchers at Cambridge and Stanford, Wu Youyou, Michal Kosinski and David Stillwell, ran a contest.⁴ They compared how well the people who knew someone judged her personality with how well a computer did. The computer used only the pages a person had “liked” on Facebook. Their sample was 86,220 volunteers. The yardstick was each person’s own answers to a personality questionnaire. With 10 Likes, the computer’s judgements matched those answers better than a work colleague’s did. With 70 it beat a friend or room-mate, with 150 a family member, and with 300 a spouse. So the computer did not know people better than their spouses in every sense. Its judgements agreed more closely with what people said about themselves, and a yardstick made of people’s own answers has the weakness described above. The researchers also tested the judgements against life outcomes, such as substance use, political attitudes and physical health. For some of these, the computer’s judgements predicted better than the people’s own answers did. In a study published in 2024, two general-purpose language models released in March 2023 inferred personality traits from people’s Facebook posts without being trained for the task. Their estimates matched people’s own scores about as well as programs built for the purpose.⁵ The match was modest, a correlation of about 0.29 on a scale where 1 would be perfect. That is far from a reliable reading of any one person. The point is narrower: guessing some traits from what people write no longer needs a program built for the task. By 2024, describing what a person is like had become partly a machine’s work.

If the examined life were a more accurate description of yourself, the machines would have made it cheap. It was never that. The man who gave it its name was not describing himself. He was giving an account.

Plato’s *Laches* is a short dialogue about courage. In it an Athenian general, Nicias, warns a friend what happens to anyone who talks with Socrates for long. In Benjamin Jowett’s translation⁶, the talk goes round and round, “until at last he finds that he has to give an account both of his present and past life”. “Socrates will not let him go until he has completely and thoroughly sifted him.” Nicias does not complain. He says there is no harm in being reminded of a wrong, and that “he who does not fly from reproof will be sure to take more heed of his after-life”. This is Plato’s scene, written by an admirer, not a record of a real conversation. The words Jowett renders “give an account” come from the Greek idiom *logon didonai*, and a *logos* is both an account and a reason.

A description says what you are like. An account says what you have done and why, to someone who can question it. A profile can be accurate without anyone standing behind it. An account can be questioned, defended and changed, and it is yours to give. At his trial, as Plato tells it in the *Apology*, Socrates described his own life in the same terms. In Harold North Fowler’s translation⁷, he spent his days “examining myself and others”. The examining was done aloud, among people who could reply.

An account is also more than an explanation. To explain an act is to make it understandable: I was afraid, I was told to, the system recommended it. Each may be true, and none of them says whether the act was right. To justify an act is to give reasons for it that can survive questioning. To answer for it is to accept that the questioning may leave something to repair. A person can be eloquent about her motives and honest about her fears, and still answer for nothing. The examined life is not a life that has learned to explain itself well.

Nor is the person who gives the account its only judge. To examine a life is to ask why I acted, what I actually did and what it did to others. The first question can be answered alone; the other two cannot. Someone may prize honesty and lead a team in which nobody can safely tell her the truth. Her motives are hers to report. What her conduct did to them is partly theirs to say. Chapter 31 found the same of a builder and the people her system judged. The person who tells the story of her life is not the only witness to it.

Not everyone who can ask has a right to an answer. A person owes an account to those her acts have harmed, to those whose trust she has asked for, and to those over whom she holds power. She does not owe every thought to an employer, a platform, a stranger or a model that has inferred her traits. Being answerable is not the same as being exposed, and the examined life asks only for the first. The power to observe a person is not a right to govern her. Nor is refusing an audience with no right to ask the same as escaping one that has it. A life watched all the time may even find it harder to examine itself, because every private doubt begins to expect an audience.

### Plot twist

Giving an account to other people does not by itself make anyone more honest. In 1989 the psychologists Philip Tetlock, Linda Skitka and Richard Boettger asked 325 people to write their thoughts on contested policies, such as capital punishment. Some were told they would have to explain and justify their opinions to another participant. Those who believed they knew that person’s views moved towards them. Those already committed to a position spent their effort defending it. Those who did not know the partner’s views, and had not yet taken a stand, thought in a more self-critical and many-sided way.¹

Ten years later Jennifer Lerner and Tetlock reviewed the wider research and reported the same tendencies. An audience whose views are known tends to draw less effort and more biased judgements. The review also found that the effects depend on the task and on the person.¹ These were short tasks about opinions, not lives. Still, they change what this chapter can ask of an account. It is not enough to owe it to the right people. It helps if the hearer keeps her own view back until you have spoken, and if you give the account before your mind is made up. An account shaped to please its hearer is a performance for an audience of one.

There is an obvious reply to all this. A machine can write accounts too. A model given a month of a person’s messages and calendar can already write her a fluent account of her choices. It can say what she did, what she seems to have wanted and where she contradicted herself. Some of it will be right, and some of it may be more honest than she would have been. But an account written for you and accepted unread is a ghost-written examination. It describes a life; nobody has yet given it. What makes an account yours is not who drafted the sentences. It is that you have read them, can be questioned on them, and stand behind them, and that you change what they show you cannot defend. Standing behind an account means more than agreeing with it. It means being ready to defend it, to bear what follows if it is false, and to change course where it fails.

This is the book’s own word for what a person does with a decision, applied to a life. To answer for something is to explain it, undo it where it can be undone and put right what can be put right. An examined life is a life whose account its owner can give, and change. As that reply grants, a machine can help to prepare such an account. It can find the patterns, ask the questions and draft the sentences, even a persuasive defence. It can be a good clerk and prepare the brief. The account is still yours to give, because the acts were yours, and so is the obligation the account answers. Sentences can be shared out; that obligation cannot. That is a matter of meaning, not of what machines can do this year: the life is yours, and a machine will never be you.

The strategy reading fails twice. As a fact, knowing yourself does not reliably mend your judgement. As a motive, self-knowledge sought for an advantage is a way of working on yourself as a product. Section 32.5 will call that self-optimisation. Sonoda’s teacher had a shorter name for it.

## 32.2 The Daily Practice of Attention

The daily practice is not a practice of consciousness. Consciousness, in this book, is experience known as your own. Nobody practises it. What can be practised is attention.

Sonoda sits every morning. He has sat through grief, and in hotel rooms where the cushion was a folded towel on a bathroom floor. He has sat on mornings when the whole exercise felt like habit pretending to be discipline.

He learned early that the vocabulary of Zen got in the way. The founders did not need a tradition; they needed attention. So he gives them the practice in plain words. Sit. Notice what arises. Do not follow it. Do not fix it. What you find is your business. Over months, not weeks, he tells them (he is blunt about this), people begin to see some of the machinery under their decisions. They notice the stories they tell about themselves, the fears that pass for reasons, and the wants that present themselves as principles.

He knows what this costs. His teacher would have said that a practice stripped of its tradition loses what held it. The tradition is not decoration. It is the container. Sonoda is not sure his teacher is wrong. Sometimes he wonders whether his clients receive attention or only the feeling of it: a high-end simulation of the real thing, impressive and empty.

The container holds a purpose larger than the sitter’s comfort, and a teacher with the standing to say you are wrong. It also holds names for the practice’s own counterfeits. Japanese Zen has a name for a fault close to the one his teacher saw: *yako-zen*, “wild-fox Zen”. Japanese dictionaries define it as the conceit of a practitioner who thinks he is awakened when he is not.⁸ The name comes from an old koan in *The Gateless Gate*, a Chinese collection of the thirteenth century. In it an old man says that long ago, as a teacher, he told a student that a person of deep practice no longer falls under cause and effect. For that answer he lived five hundred lives as a fox.

A tradition is no guarantee either. It can shelter a teacher whom nobody may question, and its authority is not right because it is old. What matters is what it keeps for correcting itself. Sonoda gives his founders the technique without the container. The one person in the room who might tell them they are wrong is him.

The founders who work with him say that they make fewer decisions in a panic. That is what they report, and a report is all it is. Sonoda does not claim to know whether his practice causes it, or whether any steady habit of attending to oneself would do the same. He is right not to claim it. In 2014 a team led by Madhav Goyal at Johns Hopkins reviewed the trials of meditation programmes.⁸ They found moderate evidence that mindfulness programmes improved anxiety, depression and pain. For positive mood, attention, sleep and several other outcomes the evidence showed no effect, or was too thin to say. And they “found no evidence that meditation programs were better than any active treatment”, such as drugs, exercise or other therapies. The practice also has a cost some people pay. A review of 83 studies of meditation practices and meditation-based therapies, published in 2020, put the share of participants who had an unwanted effect at about 8 per cent. The share was about 4 per cent in experimental studies and about 33 per cent in observational ones, so the figure depends heavily on how a study looks. The effects the studies reported most often were anxiety and depression. Unwanted effects can occur even in people with no history of mental illness.⁹ The review does not show that every tradition, or every way of sitting, carries the same risk. None of this is advice. It says what the evidence can bear: a practice may help, it is no cure, and it promises no calm.

Then there is the money. Companies pay Sonoda so that their founders will decide better. The founder’s calm is bought because it is expected to serve the company. Sonoda takes the fee, and his doubt is partly about the fee. Socrates refused one. At his trial he told the jury⁷ that his accusers could not “produce a witness to testify that I ever exacted or asked pay of anyone”. He added: “I think I have a sufficient witness that I speak the truth, namely, my poverty.” In Plato’s telling this is part of his defence, not a full record of how he lived. Teachers must live, and a fee does not make a practice false; this book is sold too. But a practice paid for by those who want a certain result is pulled towards producing it, and towards looking as if it does. A questioner whose fee depends on the people he questions may also learn to spare them the hardest question. The examining becomes a performance when the result the payer wants becomes its reason.

Honesty about the product does not take it out of the market. Even a refusal to promise results can be bought, by clients rich enough, as a sign of depth. So what a seller says about his hour cannot settle whether it is a performance. What he does comes closer to settling it: whether he tells a client who still pays that the hours have stopped helping.

A voice from another tradition says the same about the outward signs. Lal Ded, also called Lallā, was a Kashmiri poet of the fourteenth century who left her household and wandered as a beggar. Her sayings are short verses in Kashmiri called *vākh*. George Grierson and Lionel Barnett translated a collection of them in 1920. Their translation is a century old and reads her through their own learning, and other collections number her verses differently; I cite theirs. One verse describes the place her search led, and who may enter it: “Not by vow of silence, not by mystic attitudes, is there entry there.”¹⁰ By *mystic attitudes* the translators seem to mean the postures of the practised meditator; that reading is mine. The editors’ note says that “there” means the Supreme, which “is not even Siva and his Sakti, or energic power, for these have form and name”. The note also softens the verse. Vows of silence and the like “do not lead directly to Him”, and “The utmost they can do is to lead the mind to that knowledge of the Supreme which brings it into union with Him.” Her subject is that Supreme, not a founder’s decisions. I borrow only the refusal, in that measure: the forms of the practice do not, by themselves, admit anyone. A practice that has learned to perform itself has kept the forms. It may be the beautiful practice that Sonoda’s teacher warned him about.

The machine has a side here, and it should be said plainly. In 2022 four researchers, led by Nicky Terblanche, reported two randomised trials of coaching over ten months.¹¹ In one, students were coached by people. In the other they were coached by a chatbot called Coach Vici, a scripted program with no large language model behind it. By the end, the chatbot “was as effective as human coaches” at helping clients reach the goals they had set. The samples were modest, the clients were students, and the measure was their own report of their goals. It says nothing about founders, about hard decisions or about whether anyone chose more wisely. The authors granted that the program lacked the subtlety of a good human coach. It matched the human coaches all the same, on the one thing measured.

In July 2026 one of Sonoda’s clients, a founder, showed him a page on her phone. She had asked a model to question her about a decision she was avoiding. The questions on the page were his questions, nearly word for word, and one of them was better. He told her so. She asked him, not unkindly, what she was paying him for. He said he did not know. It might be an hour with someone who would be there next month. It might be the performance. He would not pretend it was the questions.

The first of his possible answers deserves its due. A model will question anyone at any hour, and forget nothing it is allowed to keep. What it does not do is sit across the table next month; and if its questions lead somewhere wrong, it is not the one who answers for where they led. There is a second difference, and it cuts both ways. The person a model questions holds the off switch. She writes the prompt, ends the session when it stings and deletes what she likes. A model may also be built to please her. A 2026 study, reported in Chapter 12, found eleven models affirming users’ actions about 49 per cent more often than people did.¹¹ A model can be told to be blunt, and some are. But a tab can be closed, and a person who will see you tomorrow cannot. Nor is feeling understood proof of being understood; a fluent reply can create the feeling from very little. None of this shows that a person questions better. People flatter too, and Sonoda’s fee depends on the people he questions. An investment adviser in Pune, earlier in this book, offered her clients a reason they could question and her presence next month to stand behind it. Sonoda can offer that too. But it is not what his letters have promised for nine years.

His fee buys four things, and the letters have bundled them. There are the questions, which a model now asks for almost nothing. There is another person’s time and attention, which a model can imitate in part. There is a relationship carried from one meeting to the next, which needs someone who stays. And there is a claim about results: the letters have promised clearer decisions. The first three he can offer and describe. The fourth needs evidence, and he has none. A founder may value the hour, feel heard and still decide no better; that she values it is no proof that it works. The 2022 trials cannot settle this either way, since they measured students’ reports of their own goals. The hour is real. Whether it is worth what the companies pay is their decision. Up to that July, he had never told them how little he could prove.

## 32.3 The Technologies of Awareness

The section keeps its title in the word’s plain sense. A technology is a technique, a way of doing something on purpose. Sitting still is one. Writing is another. A machine is now a third.

Priya Chandran works at a software company in Koramangala, Bengaluru. When this book began she was a test lead. Her team built the platform that now tests the company’s code, and since October 2025 she has led the effort to carry it across every product the company makes. She lives in a flat nearby with her mother and her cat, Turing. In April 2026 the company’s chief product officer approved, for one year, a small study she had proposed. A few volunteers would be tested each quarter on what they could do without the platform.

Later that month, after the approval, she bought a paper notebook from a stationery shop near the flat. She could not have said why, and any reason she gave might not have been the one that moved her. One link is plain from outside, though she never drew it. Her study asked what people could still do without the platform. The notebook asked what she noticed with no machine at all. At the top of the first page she wrote one question: *What did I notice today?* She wrote most evenings after dinner. Some nights it was a page; some nights it was three sentences. The first entries were small. The colour of the sky over the flyover on the way home. A developer who apologised twice for a question that was a good one. Her mother’s silence at dinner, which she had not noticed until she wrote it down.

Writing is a different technique from sitting. Sitting lets what arises pass without words. Writing turns it into sentences, and sentences stay. They can be read back the next week and questioned, which makes a notebook a little like an account. But an account is given to someone who can reply, and nobody would read this one. Whether a page read by nobody is closer to Socrates or to the cushion is a question this chapter comes back to.

The book has met other notebooks. Priya once kept a page of her week’s decisions so that she could explain each of them. A teacher in Sagar kept a notebook to test her eye against a machine’s. A designer in Mumbai kept a record of where each piece of his work came from, so that he could show it to others. Priya’s new notebook had no such use. It was for noticing, and for nobody. Nobody else read it.

Three words are often confused here, and the confusion matters. The first is monitoring: checking your own processing, noticing that a step is wrong. Reasoning models now do this: they check their own steps and change course. The second is mindfulness: attending, on purpose, to what is present, without judging it. The third is consciousness, which in this book means experience known as your own. The print edition defined consciousness here as something larger than both, a “recursive capacity” to “find meaning in the monitoring”. That definition is withdrawn. It turned consciousness back into a skill, and a skill can be matched.

Looking inward has a known cost. In 1999 the psychologists Paul Trapnell and Jennifer Campbell studied people who score high on attending to themselves.¹² Such people showed what the authors called a “self-absorption paradox”: “more accurate and extensive self-knowledge yet higher levels of psychological distress”. Their explanation was that one scale had been measuring two different habits, with different motives. They called one reflection. The other was rumination: going round and round the same thoughts. Both happen in notebooks. Nothing on the page tells you which one you are doing.

Looking inward is not yet examining. Introspection asks what I am feeling or thinking. Examination asks what my choices show about how I live, and whether I could defend them to the people they touch. A person can know her moods in fine detail and never look at how she treats the people who report to her. Another can be no great introspector and still hear a complaint, own a harm and change. A private notebook can serve examination in three ways. It keeps what would otherwise be lost. It shows patterns too slow to see in a day. And it makes a record against which a later, tidier story can be tested. It can also become a private theatre, where the same flattering story, or the same cruel one, is rehearsed every night. The page cannot question its author. Privacy can make honest noticing possible; it takes someone else to question what was noticed. So a notebook kept for nobody is not the opposite of an account. Often it is the first half of one.

A machine could tell Priya some of what her pages hold. Journaling apps with a language model behind them already summarise a month of entries and name the patterns in them. A model given the notebook might find its patterns faster than she would. She kept it on paper. That was a choice about who reads it, and a defensible one, since what a person writes about her days is hers to show. It was not proof that paper sees more.

Being described has a quieter risk. Suppose a model tells someone that she avoids conflict, or that fear of rejection drives her choices. Some of it may be right. Once she accepts it, she may read each new day through its words and act in ways that fit them. Her new acts leave new traces, and the next description agrees with the last. The danger is not only that the machine gets her wrong. It is that she stops questioning it, and starts to live as the performance of a profile. A description can be used without being obeyed. Nor does an examined life owe anyone a finished theory of itself. People change, contradict themselves, and sometimes never learn why they acted. Refusing to examine is one thing. Refusing an early explanation is another, and an examined life is allowed it.

## 32.4 The Relationship Between Stillness and Speed

Late in May 2026, Priya sat in on a team’s triage meeting. Triage is the sorting of bugs by how serious they are, so that the worst are fixed first. Her platform suggested a grade for each bug it found. It had learned to grade by studying how the teams had graded bugs in the past. When a team disputed a grade, the dispute came to Priya, because what the platform taught was hers to decide.

The disputed bug was on a payment screen. When a customer confirmed a payment, a screen reader skipped the amount. A screen reader is software that reads a screen aloud, for people who cannot see it well. The platform had graded the bug minor. A senior engineer agreed. Three customers had reported it in a quarter, among millions of payments. A junior engineer, two years out of college, said it was critical. Both had their data. Both were sure.

Once, Priya would have weighed the two cases, ruled and moved on. That is what a lead does. This time she paused, for perhaps four seconds. Then she asked a question instead of ruling. Is this about this bug, or is this about something larger?

It was about something larger. The junior engineer had spent an afternoon in the platform’s own records. She had found five bugs like this one in three months, each of them stopping a screen reader on a payment or a login, and each graded minor. The platform had done exactly what it was built to do. It had learned from the teams’ past grades, and the teams had graded by how many customers complained. The people who rely on a screen reader are those least able to complain through a screen that does not read. It was the shape Chapter 31 met in hiring: a record can learn only from what reaches it. On that one bug, the senior engineer was right about how few people had reported it. And the record that showed the pattern was the machine’s own. Nobody had looked at it that way until she did.

The grade measured complaints, not harm. Someone had chosen that measure long before, and had chosen it without asking for whom it would fail. That week, with the team’s lead, Priya changed the rule. A bug that stops a screen reader from reading a payment or a login could no longer be graded minor because few people had reported it. She wrote a note to the chief product officer under her own name. It said what the platform had learned, from whose decisions, and for how long. The five bugs were already in releases customers were using. Whether to go back to those releases, and whether to tell the customers who had been affected, was the officer’s decision. At the end of May it had not been made. Changing a rule stops the next harm; it repairs nothing that has already happened. The customers who paid for the old grade had never been asked about it. The junior engineer’s afternoon had come out of her own evening.

The senior engineer did not object to the new rule. He objected to one thing, and he was right. A rule written after five bugs may over-grade the sixth. Some faults that trouble a screen reader are cosmetic, and the rule would send them to the top of the list with the rest. Nobody had yet counted how many such bugs there were. A rule that ends one harm can start a smaller one, and the team, not Priya, would pay for it in hours.

The fix for a bad measure is seldom one fixed rule. A better grade would ask several things. It would ask whether the screen reader is blocked or only slowed, and whether the fault touches a payment, a login or only a heading. It would ask whether there is a way round it, and whether the people it fails can tell anyone. Even then, a rule written after five bugs can become the next measure nobody questions. What keeps it honest is that someone, here the senior engineer, can go on objecting.

Nearly a year earlier, on the same platform, one team had spent six weeks pressing “acknowledged” on each finding the moment the screen allowed. That was four seconds after it appeared, and nobody read it first: a pause of the same length.

That is the point of this section. Slowness is not awareness by itself. A pause is a stretch of time, and it holds whatever is put into it. The click held nothing. Earlier in this book, U Nandaka warned that a still mind can be asleep, and the click proves him right. Another chapter of this book set a rule for acts: be fast where an act can still be undone and slow where it cannot. That rule is about the work. This section is about the four seconds inside a person.

What filled Priya’s four seconds was not a look inward. She did not consult her feelings or check her notebook. Her attention went to someone else, and to what the junior engineer had not yet managed to say. In a letter of 1942, Simone Weil called attention the rarest and purest form of generosity.¹³ The pause made room for a question, and the question let another person’s evidence be heard. A reasoning model pauses too: it does more computation before it answers, and on some tasks the answer improves.¹³ That is more work, not more attention. A person’s pause can change whose answer, or whose question, is heard. That is all a pause can do. It cannot tell you what to put into it.

Attention is the first of four steps, and the easiest to praise. Something overlooked becomes visible. Someone sees that it matters. Someone decides what ought to happen. Then someone acts, and repairs what can be repaired. The junior engineer took the first step. Priya’s pause and question let it be heard, and the team took the second. The rule change was a judgement and a start. The repair, for customers already using the five releases, waited on someone else’s decision. And the company’s new attention had been paid for out of the junior engineer’s evening. A company that praises such a finding and leaves the harm has turned one person’s unpaid attention into its own gain.

The print edition told this story as a triumph: the model was audited, and a whole category of errors was found and fixed. As told here, it is not a triumph. The pause was short, and it found a fault in a machine Priya was responsible for. Five bugs were still unrepaired in releases customers were using.

The print edition gave this section a second case: a founder in Singapore whose fast decisions had been shaped by a fear of looking indecisive to her investors. Investors’ pressure is real, and it belongs to the investors as much as to her. But it came with a coach who led her to diagnose herself. This book refuses that shape, and the case is cut.

## 32.5 Building a Life That Requires a Human to Live It

The section’s title is true in one sense and false in another. It is true by meaning: as an earlier chapter said, no one else, human or machine, can live a person’s life for her. This chapter’s claim is narrower. Nobody else can give its account. It is false if it is read as a list of tasks machines cannot do. Any such list shrinks every year. A place in a life chosen because machines cannot yet reach it is, as an earlier chapter put it, located by a date. The title is kept in its first sense only.

Here a line can be drawn between two things. On one side is self-optimisation: the logic of the algorithm applied to the self. The quantified routine. The tracked habits. The engineered morning. The steady improvement of one’s own numbers, as though one were a system to be tuned rather than a person to be understood. Self-optimisation can produce real improvements in health, in work and in discipline. It is also easy to read from outside. A machine can model it, and a machine coach can now run much of it; one scripted coach matched human coaches on goals in 2022. This book has no quarrel with that. The quarrel in 32.1 was with tuning yourself for an advantage and calling it the examined life.

On the other side is self-knowledge, which does not optimise. The print edition also said that it “illuminates”, and this edition withdraws that word. What the examined life does is not shed light on a fixed self. It gives an account of a life that is still being lived, and changes it. You are brave in some places and cowardly in others, generous in some relations and withholding in others. An examination does not remove these contradictions. It makes them something you can be asked about, and can answer for. The difference lies between two orientations toward the self, not between two types of people. One asks how to perform better. The other asks whether this is how to live, and whether you could say why to someone who has the right to ask.

The same practice can serve either orientation. Someone who tracks her blood pressure may be caring for a body she lives in. Someone who examines her motives may want to be admired for it. So the question is not whether anything improves. It is what the improvement serves, and whether that aim survives the question. Optimising asks which measure shows progress. Examining asks what the measure leaves out, and whether the target is worth reaching. A life can become more efficient without becoming more defensible.

Socrates lived by the second question. As Plato tells it,⁷ the jury found him guilty, and he was asked to propose his own penalty. He put the question to himself: “Perhaps someone might say, Socrates, can you not go away from us and live quietly, without talking?” He replied that “such conduct would be disobedience to the god and that therefore I cannot keep quiet”. Then he said the sentence his life is remembered by, knowing the jury would believe it even less: that “to talk every day about virtue and the other things about which you hear me talking and examining myself and others is the greatest good to man, and that the unexamined life is not worth living”. He chose the examining over his life, and the jury chose his death. That was an act with its price.

Two marks belong beside that sentence. First, this is Plato’s account of the defence, written after Socrates died in 399 BCE, and Socrates gave a god’s command as his reason. I borrow the shape of his life, not his theology. Second, “not worth living” is Socrates speaking of the life he would choose for himself. It is not a measure of anyone’s worth. A child, a person with dementia, a person too ill or too tired to examine anything: their lives are worth no less. This book never measures a person by whether she examines her life. Some people will never be able to give an account of this kind, and others must speak up for their interests. Their claim to be treated justly is no smaller.

The conditions for examining are not shared out evenly either. Time, privacy, enough safety to admit a mistake, and someone who will ask you why: some people have all of these, and many have few. A person working two jobs, caring for a parent or afraid of her employer may examine her life as truly, with less room to do it in. The examined life is open to anyone and easier for some. It must not become one more duty laid on those already carrying the most.

Lal Ded would put Socrates’ question differently. One of her verses, in Grierson and Barnett’s translation, reads:¹⁴

> “Let him bind abuse upon me, let him orate blame against me, Let each one say to me what pleaseth each. Yea, let him worship me with the offering of his own soul for the flowers. Still keep I myself untouched and undefiled by all these; so who getteth what therefrom?”

The editors’ note says: “She is callous to the blame or praise of the world.” Here too I borrow only one thing from a mystic’s verses: her refusal of the audience.

The two seem to disagree about whom an examined life is for. For Socrates it was for others as much as for himself. He examined in public, aloud, among people who could reply, and at his trial he answered his accusers at length. Lal Ded, as another verse has it, went out searching day and night and found what she sought “in mine own house”.¹⁵ On my reading she found her teacher at home; the editors read the house as her own soul. Either way, and this too is my reading, the watchers did not decide her worth. Praise and blame did not touch her. They agree on one thing: the outward signs of a practice, its forms and its rewards, are not the thing itself. Socrates took no fee; Lal Ded denied that silence and postures could, by themselves, admit anyone. Where they part is narrower than it looks.

Being untouched by praise and blame is not the same as owing no one an answer. Nothing in the verses I cite says that she refused to answer for a harm she had done. What she refused was the crowd’s verdict as the measure of herself. Socrates held that a life must stay open to challenge; Lal Ded, that the challengers’ approval must not become its worth. Put shortly, you owe other people your reasons, and you do not owe a crowd a performance of your virtue. One governs the account and the other the motive. They still part on what the examining is finally for. For Socrates it was the best thing a human being could do, every day, with others. For Lal Ded it served a search for the Supreme, which no city could judge. I borrow a question from each, not their ends.

Socrates’ way has its own danger. In his defence at the trial he says that questioning one man made him “hateful to him and to many of those present”.⁷ Questions can shame without helping, and questions without end can wear a person into agreement. A model can produce them without end. Openness to challenge has to run both ways: the one who questions must be open to correction too.

Sonoda lives between the two. His sitting needs Lal Ded’s freedom from the watcher’s verdict, or it becomes the beautiful practice his teacher warned him about. His work needs Socrates’ account, because his clients are owed an honest account of what they pay for. Another of her verses says that “Integrity and high repute are but water carried in a basket”:¹⁶ only someone who could grasp the wind in his fist would keep them. She meant both, and keeping both at once is Sonoda’s task: the honesty of his doubt, and the good name that brings him clients. Priya’s notebook, kept for nobody, is on Lal Ded’s side of the line. Her question in the meeting, asked aloud to another person and open to reply, was on Socrates’.

There is a harder question under this one. Machines may one day require nothing of us at all: no work, no decisions, no account. Is that the danger, or the goal? For Lal Ded, to be required by nobody’s praise or blame was close to freedom. For Socrates, examining was the best thing a human being could do every day, with others, for as long as he lived. On Lal Ded’s terms, a machine that asked nothing of us might leave us free to search without an audience. On Socrates’ terms, it might leave nobody to ask us why. This chapter does not decide between them. It says only what each would ask of a person. Lal Ded would ask whether the practice still works with nobody watching. Socrates would ask whether you can give the account to someone who may say you are wrong.

The print edition ended here with a promise. Build a career on judgement, it said, and “it compounds”; a business on trust, and “it endures”; work that demands taste, and “it resists automation”. None of the three is a promise this book can keep, and they are withdrawn. What can be kept is smaller. The examined life does not get solved. It continues, honestly and imperfectly, for as long as you are alive to examine it.

This is the last chapter of Part VI, the part of the book about what happens inside a person when machines take over work. Each of its chapters found that something it was asked to supply from within cannot be supplied from within alone. The fear of becoming obsolete is often right about what can be taken and wrong about what a person is, and seeing the difference neither ends the fear nor pays the rent. Some people come out of a shock stronger, but only with slack they seldom made alone, and whoever imposed the shock answers for it. A builder’s good intentions settle her motive, not her debt to people her system judged and she never saw. And an examined life is an account given to others, not a self found by looking. Together they change what Part V meant. Part V found that the one who answers is not a remainder found at the end: the seat is designed in advance, held for a stated time and paid for. Part VI finds that beneath the work lies no sturdier self waiting to be uncovered, only a worth that never rested on being useful, and that does not pay the rent. That leaves the person in the seat. She is someone who keeps giving an account, among people who can ask her why. No inner work she does replaces what the institution around her owes.

## Ending: The Second Paragraph

Every October Sonoda sends his clients a letter before they decide whether to renew. For nine years its second paragraph had said that the sessions help founders make clearer decisions.

At the end of September 2026 he rewrote it. The new paragraph said what he could stand behind. He would sit with the founder for an hour, ask questions and listen. A machine might ask the same questions, and some would be as good as his or better. He could not promise clearer decisions, and the studies he had read did not let him. And he would tell them when he thought the sitting had become a performance, if he could see it. He did not know whether the new paragraph was honesty, or a better performance of it.

He sent the letter in the first week of October, before he knew who would renew. The next morning he sat as usual and watched the light come into the garden.

### Practice

*An account to one person.* Choose one thing you do most days that nobody has asked you about for a long time. It might be a habit of work, a practice you keep, or a decision you now leave to a machine. Write two sentences: what you do, and why. Then ask one person whose days it touches to ask you about it and to tell you what it does to them. Ask them to keep asking why until your answers are ones you have not given before. They may say no. If they do, or if there is nobody with whom this feels safe, write the two sentences as a letter to that person. Read it a week later as if you were them. Do not defend. Listen to what you say. The next day, write down which answer you could stand behind, and one thing you will change or keep, on purpose. It costs an hour of yours and an hour of theirs, and theirs is a gift; say so. A machine can ask the questions too, and may ask them well. This exercise asks for someone who will see you tomorrow. It will not make you calmer, and it is not therapy; if it reaches something painful, stop. If you are struggling to cope with a low mood, or what you try yourself is not helping, speak to a doctor.¹⁷ No practice is owed. It shows whether you can give an account of one part of how you live.

## Words to keep

- **account (in this chapter):** what you have done and why, given to someone who can question it, and changed where it cannot be defended; distinct from a description of what you are like.
- **answerability:** owing an account of one’s conduct to those with a claim to it, such as people one has harmed, people whose trust one asked for and people under one’s authority; distinct from exposure, being visible to anyone who can look.
- **bias blind spot:** the tendency to see biases at work in other people’s judgements much more than in one’s own (Pronin, Lin and Ross).
- **examined life:** Socrates’ phrase, in Plato’s *Apology*, for a life spent examining oneself and others; in this chapter, a life whose account its owner can give and change.
- **introspection:** looking at one’s own thoughts and feelings; one part of examining a life, not the whole of it.
- **justification:** reasons for an act that can survive questioning; distinct from an explanation, which only makes the act understandable.
- **koan:** a teaching riddle in Zen, set by a teacher to a student.
- **mindfulness:** attending, on purpose, to what is present, without judging it.
- **monitoring:** checking one’s own processing, noticing that a step is wrong; as of 2026, reasoning models do some of it.
- **Rinzai:** one of the main schools of Zen Buddhism in Japan, known for its use of koans.
- **rumination and reflection:** two habits of attending to oneself, one a going round the same thoughts, the other a looking that learns (Trapnell and Campbell).
- **screen reader:** software that reads the text on a screen aloud, for people who cannot see it well.
- **self-optimisation:** working on oneself as a system to be tuned, by tracking and improving one’s own measures.
- **sycophancy (in language models):** a tendency, learned partly from people’s approval, to agree with the user rather than with the truth.
- <strong><em>vākh</em></strong><strong>:</strong> Kashmiri, a short saying in verse; the form of Lal Ded’s poems.
- <strong><em>yako-zen</em></strong><strong>:</strong> Japanese, “wild-fox Zen”: the conceit of a Zen practitioner who thinks he is awakened when he is not.

## Sources

1. Emily Pronin, Daniel Y. Lin and Lee Ross, “The Bias Blind Spot: Perceptions of Bias in Self Versus Others,” *Personality and Social Psychology Bulletin* 28, no. 3 (2002): 369–381, abstract (three studies; biases seen “much more in others than in themselves”). Accountability and its audience (added with “Plot twist” in 32.1): Philip E. Tetlock, Linda Skitka and Richard Boettger, “Social and Cognitive Strategies for Coping With Accountability: Conformity, Complexity, and Bolstering,” *Journal of Personality and Social Psychology* 57, no. 4 (1989): 632–640 (325 subjects, told they would “explain and justify your opinions to another subject”; the abstract’s three strategies are the “low-effort acceptability heuristic”, “preemptive self-criticism” and “defensive bolstering”) (CORE copy). Jennifer S. Lerner and Philip E. Tetlock, “Accounting for the Effects of Accountability,” *Psychological Bulletin* 125, no. 2 (1999): 255–275 (PubMed 10087938), as reported in Carnegie Mellon University’s release of June 1999 (EurekAlert): accountability to an audience whose views are known “elicits less effort and more biased judgments”, and effects “depend on the type of accountability, the type of task and the characteristics of the decision-maker”.
2. Richard E. Nisbett and Timothy DeCamp Wilson, “Telling More Than We Can Know: Verbal Reports on Mental Processes,” *Psychological Review* 84, no. 3 (1977): 231–259; see Chapter 1.
3. Simine Vazire, “Who Knows What About a Person? The Self-Other Knowledge Asymmetry (SOKA) Model,” *Journal of Personality and Social Psychology* 98, no. 2 (2010): 281–300; as summarised by Washington University in St. Louis, “Others may know us better than we know ourselves, study finds” (March 2010): friends better on “intellect-related traits, such as intelligence and creativity”, the self on “internal, or neurotic traits, such as anxiety”.
4. Wu Youyou, Michal Kosinski and David Stillwell, “Computer-based personality judgments are more accurate than those made by humans,” *Proceedings of the National Academy of Sciences* 112, no. 4 (2015): 1036–1040, published 12 January 2015; sample of 86,220 Facebook volunteers; 10, 70, 150 and 300 Likes to outperform a work colleague, a friend or cohabitant, a family member and a spouse (University of Cambridge and Stanford University releases, 12 January 2015); the abstract also reports higher external validity “when predicting life outcomes such as substance use, political attitudes, and physical health”, and that “for some outcomes, they even outperform the self-rated personality scores” (Europe PMC record).
5. Heinrich Peters and Sandra C. Matz, “Large language models can infer psychological dispositions of social media users,” *PNAS Nexus* 3, no. 6 (2024): pgae231, published 13 June 2024, doi:10.1093/pnasnexus/pgae231; preprint arXiv:2309.08631 (v2, 5 June 2024): GPT-3.5 (gpt-3.5-turbo-0301) and GPT-4 (gpt-4-0314), both of March 2023, zero-shot, from Facebook status updates; “an average correlation of r = .29 (range = \[.22, .33\])” with self-reports, “similar to that of supervised machine learning models”.
6. Plato, *Laches* 187e–188b, trans. Benjamin Jowett, Project Gutenberg eBook 1584 (the edition is not stated). The Greek at 187e has *didonai peri hautou logon*, “to give an account of himself” (Burnet’s text, Perseus Digital Library).
7. Plato, *Apology* 21c–e, 31b–c and 37e–38a, trans. Harold North Fowler, in *Plato in Twelve Volumes*, vol. 1 (Loeb Classical Library; Cambridge, MA: Harvard University Press; London: William Heinemann, 1914), Perseus Digital Library text (perseus-eng2). At 21d: “I became hateful to him and to many of those present”.
8. Madhav Goyal et al., “Meditation Programs for Psychological Stress and Well-being: A Systematic Review and Meta-analysis,” *JAMA Internal Medicine* 174, no. 3 (2014): 357–368: moderate evidence for anxiety, depression and pain; “low evidence of no effect or insufficient evidence of any effect” on positive mood, attention, substance use, eating habits, sleep and weight; “We found no evidence that meditation programs were better than any active treatment (ie, drugs, exercise, and other behavioral therapies).” On *yako-zen*: *Digital Daijisen* and *Seisen Nihon Kokugo Daijiten* (Kotobank, accessed 31 October 2026) define it as a Zen practitioner’s conceit that he is awakened when he is not; *Kojiseigo wo Shiru Jiten* (same page) traces it to the koan of Hyakujō’s fox, *Wumenguan* (*The Gateless Gate*, compiled by Wumen Huikai in 1228), case 2, paraphrased here.
9. Miguel Farias, Everton Maraldi, Karl C. Wallenkampf and Giancarlo Lucchetti, “Adverse events in meditation practices and meditation-based therapies: a systematic review,” *Acta Psychiatrica Scandinavica* 142, no. 5 (2020): 374–393, doi:10.1111/acps.13225: “The total prevalence of adverse events was 8.3%” (3.7 per cent in experimental studies, 33.2 per cent in observational ones); 83 studies, 55 of which reported at least one adverse event; anxiety (in 18 of those 55 studies) and depression (in 15) the commonest (Coventry University release, 17 September 2020); the authors note that such events “may occur in individuals with no previous history of mental health problems”.
10. Lal Ded, *vākh* 2, in *Lallā-Vākyāni, or The Wise Sayings of Lal Dĕd, a Mystic Poetess of Ancient Kashmīr*, ed. and trans. Sir George Grierson and Lionel D. Barnett (London: Royal Asiatic Society, 1920), p. 24; the editors’ note on p. 25: “‘There’, i.e. in the Supreme.”; the Supreme “is not even Siva and his Sakti, or energic power, for these have form and name”; vows of silence “do not lead directly to Him”; “The utmost they can do is to lead the mind to that knowledge of the Supreme which brings it into union with Him.” The verse’s next line is “Not there dwell Siva and his Sakti.”
11. Nicky Terblanche, Joanna Molyn, Erik de Haan and Viktor O. Nilsson, “Comparing artificial intelligence and human coaching goal attainment efficacy,” *PLOS ONE* 17, no. 6 (2022): e0270255, published 21 June 2022: two randomised trials over ten months; “Coach Vici”, “based on expert system (ANI) principles using chatbot technology”, employing “a scripted conversation”; “the AI coach was as effective as human coaches at the end of the trials”; undergraduate participants and self-reported goal attainment. On flattery: Myra Cheng et al., “Sycophantic AI decreases prosocial intentions and promotes dependence,” *Science* 391, no. 6792 (March 2026): eleven models; users’ actions affirmed about 49 per cent more often than by humans; see Chapter 12.
12. Paul D. Trapnell and Jennifer D. Campbell, “Private self-consciousness and the five-factor model of personality: Distinguishing rumination from reflection,” *Journal of Personality and Social Psychology* 76, no. 2 (1999): 284–304, abstract.
13. Simone Weil, letter to Joë Bousquet, 13 April 1942, in Simone Weil and Joë Bousquet, *Correspondance* (Lausanne: L’Âge d’Homme, 1982); see Chapter 3. The reasoning model’s pause: OpenAI, “Learning to Reason with LLMs,” 12 September 2024 (performance improves “with more time spent thinking (test-time compute)”); see Chapter 1. The source reports better scores with more computation; it claims nothing about attention or awareness.
14. Lal Ded, *vākh* 21, in Grierson and Barnett, *Lallā-Vākyāni* (1920), pp. 42–43 (translation and the editors’ note, p. 43). The note also calls the rendering of the third line doubtful and suggests “let him worship me with flowers of reality, i.e. with real flowers”.
15. Lal Ded, *vākh* 3, in Grierson and Barnett, *Lallā-Vākyāni* (1920), pp. 25–26: “With passionate longing did I, Lalla, go forth. Seeking and searching did I pass the day and night. Then, lo, saw I in mine own house a learned man”; the vocabulary glosses the word as “a learned man; esp. a guru or spiritual preceptor”, and the commentary reads the house as her own soul.
16. Lal Ded, *vākh* 24, in Grierson and Barnett, *Lallā-Vākyāni* (1920), p. 46, read whole: “Integrity and high repute are but water carried in a basket. If some mighty man can grasp the wind within his fist, Or if he can tether an elephant with a hair of his head, Only if one be skilled in such feats as these, will he be successful (in retaining integrity and high repute).”
17. NHS, “Low mood, sadness and depression” (nhs.uk, page last reviewed 29 July 2026; accessed 31 October 2026): see a GP if “you’re struggling to cope with a low mood” or “things you’re trying yourself are not helping”.

*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 “In the age of artificial intelligence, self-knowledge is a competitive advantage”, that “Every quality identified as irreplaceable depends on it”, and that holding Mori’s question “is the foundation of everything the AI age demands from a human being”; that “capability claims are precisely what the AI age dissolves”; that the daily practice “of consciousness” is “the substrate of the only kind of work the age demands”; that consciousness is “the recursive capacity not just to monitor your thinking but to find meaning in the monitoring”; that the traditions differ in method but share one “architecture”; that a junior engineer’s question showed “a kind of understanding no model possessed”; that the machine “will always be faster at processing” and that “clarity is the scarcer one”; that a life is “a particular, irreducible, conscious existence that only you can live”, and that being awake is something “no machine can replicate”; that a career on judgement “compounds”, a business on trust “endures” and work that demands taste “resists automation”, and that a lived life is “the deepest form of AI-proofing”; and that Socrates’ “word for the unlived life ... was waste”. The print edition named its Zen consultant Takeshi Mori (a name shared by several real people) and trained him at Daitoku-ji, a real monastery, under a teacher who could be identified; his years did not add up. It credited “the view from above” to the Stoics (the phrase is Pierre Hadot’s), used “morning pages” (Julia Cameron’s term) without credit, and echoed a line about the gap “between stimulus and response” that is often misattributed to Viktor Frankl. It also gave Priya insomnia, the 500D bus, a branded notebook, a decade at the company and an epiphany on the bus; made her pause a triumph that audited the model; and told the cases of a founder in Tokyo, a founder in Singapore and Winston Churchill. It ended on a bridge to the next Part.
- *What I now hold:* describing yourself is a dated doing. Introspection is unreliable (Nisbett and Wilson; the bias blind spot), friends see some traits better (Vazire), and machines read personality from traces (2015; 2024). The examined life, in Socrates’ sense, is an account of how one lives, given to people who may ask why, and changed where it cannot be defended (*Laches*; *Apology* 38a). A machine can help prepare it but cannot give it, because the life is yours (the first never, by meaning). Self-knowledge sought for advantage is self-optimisation, which a machine coach can largely run and which is no fault. The evidence for meditation is moderate for anxiety, depression and pain and no better than other active treatments; some people are harmed; a practice promises no calm. A practice paid for by those who want a result is pulled towards producing it, and towards looking as if it does; the examining becomes a performance once the payer’s result becomes its reason. A pause is a stretch of time that holds whatever is put into it, so slowness is not awareness by itself. “A life that requires a human to live it” is true only by meaning. An account is more than an explanation: it must face justification, consequences and repair, and its giver is not its only witness. Answerability is not exposure: being able to observe a person gives no right to her account. Socrates and Lal Ded agree that the outward signs of a practice are not the thing itself; one asks for the reasons owed to others, the other refuses to perform for a crowd, and they part on what the examining is for. Socrates’ “not worth living” measures no one’s worth.
- *What changed my mind:* the Prologue’s rule against defining people by what machines cannot yet do; the psychology of self-knowledge (Pronin, Vazire, Trapnell and Campbell) and the personality studies of 2015 and 2024; the 2014 review of meditation programmes and the 2020 review of their adverse effects; the 2022 trials of a chatbot coach; Plato’s own text, which has Socrates examining himself and others aloud, refusing fees and refusing silence; the audit’s corrections (the misquotation, the arithmetic, the real monastery, the uncredited phrases, the bus); the fact that the print edition’s best warning, the teacher’s, condemned its own first section; Japanese Zen’s own name for a fault like it; research on models that flatter their users; and three outside readings of an earlier version of this chapter, in October 2026.
- *What survives:* Mori’s doubt, now Sonoda’s, and his teacher’s warning verbatim: “Beautiful practice, his teacher said, is practice that has learned to perform itself.”; “He is very good at this. He is also not entirely sure it is honest.”; “You have turned awareness into an accomplishment, and accomplishment is the opposite of what we are doing here.”; the not-knowing as either the most honest thing about him or a convenient way of avoiding an uncomfortable answer; “His clients do not know he does not have the answer.”; his plain instruction and “a high-end simulation of the real thing, impressive and empty”; Priya’s notebook and its question, “What did I notice today?”; the four seconds and “is this about this bug, or is this about something larger?”; the distinction between self-knowledge and self-optimisation, “does not optimise” and “two orientations toward the self”; “The examined life does not get solved.”
- *Section titles in the print edition:* 32.1 was “Self-Knowledge as Strategy” and 32.2 “The Daily Practice of Consciousness”. The other three keep their print titles.

*Next in the series:* [*Part VII: The Conscious Blueprint (overview)*](https://nithinpalal.com/ai/ai-proof/part-7-the-conscious-blueprint)*.*