AI-Proof: What Machines Can Never Be — Volume III: Transformation, Part VII: The Conscious Blueprint, Chapter 34: The Conscious Solopreneur
Part VII turns the book’s argument into plans for a person, a business and an organisation. After the individual, this chapter turns to someone who works alone: what one person can promise, at what price, in the hours she has.
A designer can spend two days researching a label and cannot send an invoice. A calligrapher fears her queue is for status. Working alone, you sell a promise, not a remainder machines cannot reach, and only your own hours can keep it.
In Brief For readers who work alone and wonder what they are selling once a machine can do most of the tasks. Chapter 34 asks what a person who works alone is selling once a machine can do most of the tasks, and who keeps the promise when she cannot. A designer in Mumbai avoids his invoices; a calligrapher in Jaipur doubts what families pay her for; a translator in Bogotá is paid less for work that needs more judgement. Cicero’s contempt for paid labour, and a Sanskrit verse that promises wealth to the fit, help show what a price says, and what it cannot. The chapter also asks what a promise needs besides skill: hours, money, terms that can be bargained, and someone to keep it when its maker cannot.
Working alone, you sell a promise, not a remainder the machine cannot reach: its price says what you will stand behind, its size must fit the hours you have to keep it, and no employer will give you those hours or keep the promise when you cannot.
Opening: The Unsent Invoice
The invoice was three weeks late. Arjun Mehta knew this because the client’s accounts department had written to him twice, the second time with the subject line marked urgent. The email sat among four others he had not answered. His accountant had a question about the quarterly tax return. A stationery client wanted a revised timeline. A journalist wanted to profile “independent designers thriving in the AI era”. And there was a note from his mother asking whether he had eaten.
He had answered none of them. It was the first week of October 2026. He had spent the morning at a paper merchant in Dadar, turning over handmade sheets that might suit the stationery client’s new line. The paper was beautiful. The visit was useful. The invoice remained unsent.
He could spend two days researching sharbat traditions before touching a screen. He could not spend twenty minutes on a spreadsheet.
Arjun is a brand and packaging designer in Mumbai. He works from two rooms above a printing shop in Parel and lives in Bandra. The business is his alone. Since May a younger designer, Neha, has worked for him as a freelancer on smaller jobs, first for a hotel chain in Goa that is one of his oldest clients. Since June she has been paid a monthly fee. She sends him an invoice every month, has a desk in the second room, and is not his employee. The late invoice was for six labels he had designed for a company that makes soap in small batches. The labels had gone to the printer in the middle of September.
The same habits that made him good at design made him bad at the rest. He gave each project his whole attention, and he refused anything that did not feel like real work. Somewhere he had not examined, he believed that the business side of his practice was beneath the creative side. The work was the real thing, and the invoice a necessary indignity. His business sense was not absent. It was avoided, which is a different problem and a harder one to fix, because avoidance feels like principle.
The avoidance had a cost, and not only to him. The client wanted to pay. Under India’s goods and services tax, a business may take credit for the tax on what it buys only when, among other conditions, it holds the supplier’s tax invoice. The supplier must also have reported it.1 The credit lets a business set the tax it pays on what it buys against the tax it owes on what it sells. Without his invoice, the tax on his fee could not enter that sum. Arjun’s delay sat in someone else’s books.
The belief he had not examined is very old. Marcus Tullius Cicero, the Roman statesman and philosopher, wrote On Duties (De Officiis) as a letter to his son Marcus, then twenty-one and studying philosophy in Athens.2 Near the end of its first book, Cicero ranks the ways a man may earn his living. In Walter Miller’s translation3: “Unbecoming to a gentleman, too, and vulgar are the means of livelihood of all hired workmen whom we pay for mere manual labour, not for artistic skill”. For such men, “the very wage they receive is a pledge of their slavery”. Of shopkeepers who buy from wholesalers to sell at once, he says that “they would get no profits without a great deal of downright lying”. And “no workshop can have anything liberal about it.”
The Latin turns on two words. A buyer pays either for a man’s labour (operae) or for his skill (artes). Pay for labour marks a man unfree. Cicero wrote for a gentleman, in a society that owned slaves, and his ranking is that society’s contempt for people who worked for others. This book refuses it. But the ranking describes, on my reading, the belief Arjun had never put into words. Someone who works alone is both the maker and the small trader of his own work. Cicero’s line cuts him in two. Arjun had taken the higher half for himself and left the lower half undone. The match is an analogy, not a line of descent. Cicero sorted livelihoods by status in a slave-owning city, not design from paperwork, and later in the same passage he softens towards trade on a large scale.
The machines have given Cicero’s split a new form. They now do much of what clients once paid a designer’s hands for: the layouts, the versions, the production files. So it is tempting to conclude that what remains to sell is the skill. Arjun’s case puts the question every person who works alone now faces. When a machine can do most of your tasks and you are the whole business, what are you selling, and who keeps the promise when you cannot?
The print edition drew Cicero’s conclusion without knowing it. “Judgement, taste, trust,” it said. “For the solo practitioner, these are not assets of the business. They are the business.” It added that all of it “rests on a single consciousness”. This edition cannot say that. Every person who works alone is conscious, so the word does not tell one practice from another. And judgement, taste and trust are not a remainder that machines cannot reach. What a person who works alone sells is a promise. It covers the tasks, which machines may do, and an eye that chooses. It also covers a person who answers for the result: who will explain the work and put it right when it fails. The price says what she will stand behind. The number of promises must fit the hours she has to keep them. When she cannot keep them, through illness, age or too much work, nobody will, unless she has said who. Nothing guarantees that anyone will buy any of it.
The chapter’s title needs one correction. In this book, “conscious” can mean only “knowingly”, the sense in which you can make a choice knowingly or blindly. Arjun knew that the invoice was late. What he lacked was the willingness to look at that part of his work. And “solopreneur”, first recorded in 1992, is a blend of “solo” and “entrepreneur”.4 This chapter says “someone who works alone”, and means by it the one person who carries a business’s promises to its clients. Machines, freelancers, printers and software may do some of the work. Arjun, as 34.1 shows, does not work entirely alone.
34.1 AI-Proofing a One-Person Business
To “AI-proof” a one-person business cannot mean making it safe from machines. Nothing does that. It can mean deciding what you promise, and making sure that someone can keep it.
The change machines have brought to working alone runs two ways at once. AI has made it possible for one person to operate at a scale that once required a team. It has also made it possible for one person to become indistinguishable from every other person using the same tools. The second is measurable. In a 2024 study, stories written with ideas from an AI were rated more creative one by one, and were more alike as a set.5 The tools lift the average and shrink the variety. That study asked for stories of exactly eight sentences, and the writers it helped most were the less creative ones.5 Whether it holds for design was not tested. I take it as a likely direction, not a finding: design has briefs, brand rules and printers that may pull work apart again.
Arjun has an order of work for the tools, learned in his first years with them. Before he gives a machine anything, he works out what a design must say, on paper, in his own rough drawings. Then he lets the machine make versions, and chooses among them. That keeps the design his own. It does not settle the harder question. What, specifically, is he providing that a client cannot get from a machine directly, from a larger firm using machines, or from another designer with the same tools? It is not quality alone, since machines make competent work. It is not speed, since machines are faster.
Here the machine is on the side of the part Arjun avoids. As of 2026, a model can draft an invoice from a quote in seconds. It can sort a month’s receipts for the accountant and write a polite reminder to a client who has not paid. The twenty minutes he could not spend on a spreadsheet are now five. His avoidance has lost its last excuse. Whatever was beneath him in the business side, it was not the effort.
So what does he sell? Chapter 26 split a design practice into three parts that travel differently. There are the tasks, which can be written down and done by a machine or another designer. There is the eye, which chooses and argues. And there is the person who stands behind the result. Each is part of what a client buys. None is a remainder that machines cannot reach. Machines already do many of the tasks, and a model, if asked, will argue with a brief. What a client buys from someone who works alone is the three bound into a promise by one person who can be found. The third part is not the old remainder under a new name. Answering is a position, held by someone a client can reach, question and hold to putting things right. A company that sells machine work holds it too, as 34.2 shows. What one person adds is that the one who answers is also the one who chose. Even that is narrower than it sounds. He chooses the result, not every part of what made it: the model, the font licence, the printer and the client’s brief are others’. He answers to the client for all of it, and may have to recover some of the cost from those others. For someone who works alone, the tasks, the eye and the answering still sit in one body. That is both what he sells and where he is weakest.
That changes what the business side is. A quote is the promise stated in money. It says what will be done, by when, for how much, and what will be put right if it fails. The invoice is its last line. It says that the work is done and that he stands behind it. Arjun had no trouble standing behind his work. He had trouble saying so in money. That leaves a promise half made, and calls it a higher form of care.
The invoice is not paperwork laid on top of art. It is the public form of the promise, the document that lets the client close its books. A promise also has parts that can come apart. There is the thing delivered, the standard it meets and the checking it gets. There is what is kept up after delivery, and what happens when it fails. A label can be beautiful and late. A corrected wedding card can arrive after the wedding. Being willing to answer is not the same as being able to repair. A person may stand behind the work and still lack the money, time or authority to make a client whole. A fixed price is a promise to both sides as well. Keeping it when the work overruns is honest. Letting every overrun go unpriced, quote after quote, is not care but a leak, and the cure is a better next quote.
The print edition proposed an “inversion”: hand the machine the routine hours, and keep your own for the thinking. But the inversion is a principle, not a partition. The judgement in an email to a client’s accounts department is not a different faculty from the judgement in a brand. An invoice sent on time, a date kept, a reply to the person who asked twice: these are the promise too, kept in small parts.
There is a harder point under this one. Every practice in this book needs time, and it matters whose hour it is. An employee’s hours are mostly his employer’s. A good employer gives back some of them: time to check, a safe refusal, a deadline set by someone who answers for it. When you work alone, nobody gives you those. When you are the whole business, the institution’s half is yours too. You set the hours for checking, the right to turn a client down, the record of what you promised, and the price. You decide who would keep your promises if you were ill for a month. The people who pay you hold the rest. They set the deadlines, write the contracts and, within what the law allows, decide how soon to pay. Some who work alone do not even set the price. Where an app sets what a driver or a delivery rider is paid for each job, the app holds much of the institution’s half. What this chapter says about a price belongs partly to it. Working alone does not mean bargaining on equal terms. A person may own the tools, choose the hours and pay the taxes while someone else controls access to customers, the price and when payment comes. Being free to refuse a client is not the same as being able to live through refusing one. Arjun had already learned, in other forms, how much his clients hold. A client’s nine-day deadline once ended in a reprint he paid for. A contract made him promise originality, and no toolmaker’s promise to defend its customers covered a designer on his own. So the risk ran one way, to him: his client’s contract pushed it down, and the toolmaker’s terms left it there. The invoice was the same lesson from the other side. Nobody but he could send it.
Payment is part of the promise too, and it runs both ways. A client with an accounts department can pay late and keep working. A person who works alone, paid late, may have to pay her own freelancer late, or her rent. Deposits, payment in stages and a written charge for late payment are how a promise is kept on both sides. In India, a buyer from a registered micro or small enterprise must pay within the agreed time, and never later than forty-five days after accepting the work. If it pays late, it owes compound interest at three times the bank rate set by the Reserve Bank of India.1 Arjun’s lateness ran the other way, but the law shows which side usually waits.
Nor had he written down who would keep his promises if he could not. If he were ill for a month, Neha could still draw the Goa chain’s menus. But the contract says that he looks at each one before it goes, so even those would wait. The brands, the labels and the stationery would stop. Saying who can be a small act. A line in each quote can name the colleague who would finish the work, or say that it would stop and what would be paid back. Neither is free. A colleague must know the work and be paid to stay ready, and can stand in only where the client agrees, as the Goa contract shows. Money paid back comes out of savings. A month is the easy case. If he could not come back at all, someone would need to reach his clients and finish or return the work. Someone would need to pay Neha what she was owed and close what could not be finished. A folder of files does not hold the judgement that made them. Sometimes the honest arrangement is that the work stops and the money comes back. That is a promise too, and a client should know it before it is needed. What Neha needs from him is what he would want from an employer: money on time, work she can see coming, and a word before it stops. He pays her on time. The rest he has not written down.
On paper Neha is not his employee. In practice she has a desk in his second room, a fee each month and work that waits on his eye. That is much of what an employee has, without what an employee is owed. A label does not settle it in law: Indian courts have asked, among other things, who controls how the work is done.1 Arjun, who has no employer, has become something close to one. The institution’s half that nobody holds for him, he now holds for someone younger. India’s social security code, in force since November 2025, may count her as a gig worker: someone who earns from work outside a traditional employer-employee relationship. It lets the central government frame schemes for such workers.1 It does not make Arjun owe her notice. Only he can write that down.
34.2 Pricing in an AI-Commoditised Market
Nandini Mathur’s hands ache in the mornings now. She does not mention this to clients. She runs them under warm water for ten minutes before the first stroke of the day, flexing them until the nib will move the way it must.
She writes wedding invitations by hand, in the old walled city of Jaipur, at a table in a room off the courtyard of her family’s house. She writes in Devanāgarī, the script of Hindi and Sanskrit; in nastaʿlīq, the flowing Persian hand in which Urdu is written; and in Roman letters. She began learning from her grandmother at eleven. Her waiting list runs more than a year.
That is hard to explain by price alone. Machine-lettered invitations cost a small part of her fee. In April 2026, one of the large model makers said that its new image model made “significant gains in non-Latin text rendering”, and named Hindi among the languages.6 That was the company’s claim about drawing the script, not a test of spelling a family’s names. In June a wedding planner showed Nandini a set of cards made that way. There was a couplet in Devanāgarī, the couple’s names and the date, in curling letters of the kind she had spent thirty years learning. She looked at them for a long time. She could not honestly say they were bad. The families who choose such cards want a beautiful invitation at a price they can pay, and that is a reasonable thing to want.
And yet the waiting list.
The standard explanation is that Nandini sells presence. Each card is written by a human hand. The pressure of the nib varies with her breath. No two letters are identical. The variation is the proof. Families pay to know that someone sat with this paper, this ink and these names.
The standard explanation is correct. It is also incomplete. An earlier chapter found that people pay a person for different reasons, which the word “presence” runs together. Some pay for the trace of a particular hand. Some pay for a relationship. Some pay for someone who can be questioned and made to put things right. And some pay for status. That chapter also found that such a premium is a preference, and preferences carry a date.
Nandini knows all of this from the inside, because she doubts it. She worries that her hands will fail before she is ready to stop. She worries that the families who come to her are paying for a status signal, the cachet of hand-lettered cards among their friends. If the fashion moves, the waiting list will go. She worries about the young women from the calligraphy schools in Rajasthan. They are skilled, they charge a third of her fee, and she cannot honestly say their work is bad. Some mornings, when the stiffness takes longer to go, she worries that she is a relic. She worries, too, that her refusal to train an apprentice quickly enough is a form of hoarding dressed up as standards.
One fact sharpens her doubt instead of settling it. In 2025 she raised her fee by a third, and the waiting list grew. She does not know what that means. If families paid only for her skill, a higher price should have sent some of them to the schools. If they pay partly for status, a higher price is part of what they buy. The same number fits both readings, and she cannot see inside the families who pay it.
An old Sanskrit verse states the promise she would like to believe. The Hitopadeśa, “Good Counsel”, is a book of fables by a writer named Nārāyaṇa, more than six hundred years old.7 In its frame story, a king of Pāṭaliputra despairs of his sons, who will not study, and a teacher undertakes to teach them through fables. Its opening verses praise learning. The sixth reads:
विद्या ददाति विनयं विनयाद् याति पात्रताम् । / पात्रत्वाद् धनम् आप्नोति धनाद् धर्मं ततः सुखम् ॥
vidyā dadāti vinayaṃ vinayād yāti pātratām | / pātratvād dhanam āpnoti dhanād dharmaṃ tataḥ sukham ||
Francis Johnson translated it in 1848. He taught Sanskrit, Bengali and Telugu at the college where the East India Company trained young men for its service in India.8 His translation was made “for the use of the Sanskrit student”.9 “Knowledge gives discretion. Through discretion a man attains fitness for employment. By fitness, he acquires wealth. From wealth, religious merit: thence proceeds felicity.” Max Müller’s word-by-word English of 1864 gives the key word differently, as “worthiness”.10
The word is pātratā. A pātra is a cup or a bowl, any vessel, and also a person fit to receive, worthy of a gift.11 Pātratā is being such a vessel. Johnson’s students were training for the Company’s service, and that may explain why he wrote “fitness for employment”. That is my reading, not his statement. Either way, the verse is the promise this book refuses, in one of its oldest Indian forms. Learn, behave well, and wealth follows. But the word itself, on my reading, puts the wealth in other hands. The verse’s verb, āpnoti, ‘obtains’, makes the fit man the one who gets it; a vessel, though, is filled by someone else. That is a reading of the image, not of the grammar, in which the fit man still obtains the wealth. And the book teaches princes, not markets. The verse’s chain from learning to wealth passes through conditions it does not name. Whether fitness brings wealth depends on who does the filling, and on what they think they are paying for. That is exactly what Nandini cannot know. She knows her fitness. She does not know what the families buy.
So what does her price say? It is not a measure of her worth, nor of everything her work is worth to a family. A price records what two sides agree under particular conditions: what the buyer wants, what else is on offer, what the seller can refuse and what each of them knows. A person’s worth is not a price, a wage or a fee, and these can move in different directions. Her price says what she will stand behind. How high it can go is set by the families as much as by her. What it covers is hers to say. She checks every name against the list the family writes for her, letter by letter. If she misspells one, she writes the card again at her own cost. She will not write a card she has not checked, and she takes no work in a hurry. The price covers those promises, and the hours they take.
The print edition explained her premium differently: “When you hire a machine, no one is responsible for the result.” That is not true. The planner’s printing firm answers for its cards too, as a company. It refunds and reprints when a name is wrong. What Nandini can sell is narrower: her own answering for her own work, which means that she will explain a card and put it right. Not every family needs that, and the families who do not are not mistaken. Her clients pay because her work is, in their eyes, more beautiful than the machine’s, and because she stands behind it. Both conditions are necessary. Neither is sufficient alone. Even both together do not keep the waiting list. Fashion can move, and her hands can fail.
Her slowness is also a choice, and it has a price. Each family on the list is a family she cannot serve this year. Some of them will go to the schools’ young women, or to the planner’s cards. Those young women stand behind their cards too, at a third of her fee. Nandini says the schools train dozens of them every year. Nandini trains nobody. And every season she keeps the whole of the work in her own two hands, the apprentice she has not trained is a season further away. If she ever looks for one, two questions will meet her. The first is money. Her own slowness is paid for by thirty years of reputation and a waiting list. A young calligrapher at a third of her fee cannot spend ninety unbilled minutes over tea, or throw away imperfect sheets, unless someone pays for her time while she learns. Standards have a payroll. The second is what she would be looking for. Someone who can reproduce her hand is one thing. Someone who can carry the practice on in a hand of her own is another. If the apprentice must become Nandini, nobody will qualify.
34.3 More Capacity, the Same Hours
Operating at ten times the scale without losing your soul promises both growth and safety. The stories in this section give neither.
Machines give someone who works alone more capacity. Research that took days takes hours. Production that needed specialists needs fewer of them. The paperwork that once needed a part-time assistant can be done by software. The trap is the obvious one: to fill the capacity with more clients. For an employee, the saved hours usually belong to the employer, who decides what fills them. For someone who works alone, they are hers. A wedding photographer in this book spent hers on more weddings. For a self-employed person that can be the right choice. It is a choice either way, and it has a price either way.
The new capacity brings a new dependence. Arjun does not need a team for versions and production files, but he needs a model maker, a software subscription and a cloud account to keep his promises. A change in their price, terms or access can reach his clients, who never knew those firms existed. He answers to his clients for work made partly with tools he cannot fully inspect, and he cannot honestly promise what he has no way to check. Being able to move his work elsewhere is part of being free to work alone.
Arjun had his own version of the trap. By the start of 2026 he had more work than two rooms could hold and was turning clients away. That winter a client asked for a new range of labels in nine days. He agreed, and an error he could not see on his screen reached the shop shelves. He paid for the reprint. What he changed afterwards was small. Every quote of his now includes the days a printed proof takes, and he no longer offers nine.
Catalina Vargas’s story is harder, because her choice does not leave her better off.
She lives in Bogotá. For fifteen years she has translated and localised marketing material, product documents and legal summaries from English and Portuguese into Colombian Spanish, for foreign companies selling in Colombia. As of 2026, a machine makes her first draft. The draft is fluent, and it frees about half her working time. For most product manuals the machine’s version is good enough, and a procurement department that says so is not being foolish.
In June 2026 a European pharmaceutical company sent her the Spanish text of its launch material for Colombia, already translated by a machine. It was fluent and well formed. On the second page it said that the medicine had received its autorización de comercialización. That is the European Union’s phrase in Spanish for permission to sell a medicine.12 In Colombia, the permission is a registro sanitario, the public document issued by the national medicines regulator, INVIMA.13 Any Colombian reader in the trade would see at once that the text had been written for somewhere else and not checked. Catalina changed it. She marked several other phrases that read as Spain’s Spanish or Mexico’s, not Bogotá’s. The text she sent back read as though it had been written in Colombia.
This call goes partly to the machine. Nobody had told it that the text was for INVIMA. Asked for Colombia’s regulatory terms, a model of 2026 might well have found the right one; nobody tested it. The work still needed someone to see that the country mattered, to know which regulator governed the text and to check the term against its source. Catalina’s work was to know what to ask, and then to check the answer.
Her fee tells the rest. The companies she works for compare her rates with machine translation, and their procurement departments ask why they should pay the difference. Her reply, that the machine’s output needs someone who knows Colombia, is true. It also makes her a checker of a machine’s work rather than its author. A checker is paid less per word than a translator. She is paid less for work that requires more judgement. There are reasons, and none of them is that the work is easier. A buyer counts the words the machine has already typed and pays for the rest. What she prevents is invisible when it works. And finding one wrong term in a fluent page can be harder than writing the page, because fluent text invites trust. When other work competes for their attention, people who watch automated work tend to check it less closely, a pattern researchers call automation complacency.14 The company keeps most of what her care saves it: a launch not delayed, a regulator not troubled. Her fee is set by the comparison with a machine’s draft, not by the cost she prevents.
She has changed one thing about how she signs, following a rule this book has stated before: a signature should say what it did not check. At the top of each job she now writes what she checked and what she did not. On the launch material she checked the regulatory terms, the names of medicines and the register, formal or familiar, of every passage a Colombian customer would read. She did not check every sentence of the two-hundred-page training manual for its sales staff that came with it. The note also tells a client exactly what the lower fee bought.
The print edition, a few paragraphs before it told her story, had promised the opposite. “Fewer clients means less revenue. In practice, the reverse occurs, because depth commands a higher fee than volume.” Her story refutes it. Depth commands a higher fee only where a buyer wants it and can tell it apart, and many buyers cannot, or will not pay to. In April 2024 Britain’s Society of Authors published a survey of its members and other authors, with 787 replies.14 More than a third of the translators said they had already lost work to generative AI. More than four in ten said that the income from their work had fallen in value because of it. It was one union’s survey, of those who chose to reply, in another country. It does not measure the whole trade, but it points the same way.
Catalina has made a choice, and it has a price. She lets the machine and the lower rates have the routine documents. She spends the freed time where a mistake about Colombia is expensive: a product launch, a filing to the regulator, a chief executive’s interview with the Colombian press. That work pays more. By moving her hours to it, she earns about what she earned before, for more judgement, from fewer clients. Each of those clients now matters more to her year. If one of them leaves, a large part of her year leaves with it. The routine documents used to do two quiet jobs. They paid the rent between launches, and they rested her mind between hard calls. Without them her income comes in lumps, and her rate on the hard work must also pay for the empty weeks and the recovery. She has changed what she does. She has not escaped the machine or the buyers. She worries that models trained on Colombian Spanish will take the expensive work too. Her worry is not irrational. The argument that human judgement will always be needed is an argument, not a law of nature.
The old verse’s ladder breaks at one rung in her story. Her fitness rose, and her pay fell. The people who do the filling now compare her with a machine. Cicero would have read her case another way. Near the end of the same passage he relents about trade: “Trade, if it is on a small scale, is to be considered vulgar; but if wholesale and on a large scale”, it “is not to be greatly disparaged”. For Cicero, size made commerce tolerable, though only with a condition: distributing to many “without misrepresentation”. Even for him, size was excused only with honesty, and Catalina’s note at the top of each job is the same honesty at the smallest scale. The machines offer even someone who works alone a kind of size. Catalina has refused some of it, and Arjun, who was turning clients away, has had to choose which promises his two rooms can keep. Neither chose because small is nobler. Each chose because a promise has to fit the hours that keep it.
That is the point of this section. A machine gives someone who works alone more capacity. It does not give her more hours in which to check, to explain and to put right. How many promises she can keep is set by those hours, by her health and by what each promise needs. An employer would set the number for her, badly or well. When you work alone, you set it yourself, or the work sets it for you. Call what the machine adds productive capacity, and the number of promises the hours can keep promise capacity. A business can raise its output while the time to check it falls, and its takings can rise as its ability to stand behind the work drops. The three people in this chapter meet three limits. Arjun’s is how he runs his own week. Nandini’s is her body. Catalina’s is a market she cannot set.
Part of every price is the owner’s future. A fee can look profitable and leave out the hours that are never billed: winning work, revisions, bookkeeping, chasing payment. It can leave out software, a new computer, sick days, holidays and a pension. An employer would pay towards some of these. When nobody does, a client who buys cheap is partly buying the seller’s unfunded old age. A rough test is to add up what a year really costs, pension and sick days included, and divide it by the hours that can be billed.
What would prove this wrong?
My account of Catalina’s fee is a claim about a market, and a market can be checked. I hold that checking a machine’s draft is paid less than writing the text, though it needs more judgement. I also hold that depth earns a higher fee only where buyers want it and can tell it apart.
Two findings would show me wrong. The first would be checkers in fields where an error is costly, such as medicines, law and finance. They would be paid at or above the rate for translating from scratch, once machine drafts have become normal. The second would be buyers paying more for a signed note of what was checked and what was not, like the one Catalina now writes. Either would mean that buyers had learned to see what checking prevents. My reasons would then describe an early stage of the trade, not how it works.
The place to look is the trade’s own surveys. The European Language Industry Survey, first run in 2013, asks translators, language companies and others in the trade what is changing. In its 2025 report, independent translators described a “move from human translation to discounted and less-valued post-editing”, which means checking and correcting a machine’s draft. The report warns that its data “are not objective and verifiable”. They record what people in the trade say. A survey that counted the rates actually paid, field by field, would test this section better than any story in it.
34.4 Building an Honest Personal Brand
Arjun hates self-promotion. He would describe it differently. He would say that he values the work over the performance of the work, or that good work speaks for itself. He might say that he has no time for social media because he is busy designing. These are true. They are also defences against a task that makes him uncomfortable.
The standard advice is relentless. Share your process. Post consistently. Show up authentically. The advice is not wrong. Followed by a thousand people at once, it produces a thousand posts that look alike. The vulnerability is curated, and the lessons learned are formulaic. Machines speed this up. As of 2026, a person can generate a month of “authentic” posts in an afternoon. It sounds like a person without sounding like any person in particular.
In November 2025 Arjun lost a project. A start-up that lends money to small shopkeepers through an app had asked him for its identity. Then it tested his design against louder ones a machine had made, and the machine’s won on every measure the client counted. He declined, in writing, to redesign towards them, and lost the project and the fee. In December he wrote about it: what he had believed, what the test showed, and why he still did not know whether he had been right. The post brought more enquiries than anything he had published.
The print edition drew a lesson from that. “The post was legible as human because it cost him something to write.” Readers, it said, “can feel the difference” between content made for attention and content lived before it was shared. “The distinction is not subtle, despite what the engagement metrics suggest.”
The evidence says otherwise. In six experiments with 4,600 participants, published in 2023, people “were unable to detect self-presentations generated by state-of-the-art AI language models” in dating, hospitality and professional settings.15 The study was first posted in June 2022, so the models it used are older than today’s. This call goes to the machine, plainly. As of 2026, a model can write a convincing account of a lost project, with doubt in the right places, in less than a minute. If those experiments are any guide, most readers could not tell it from Arjun’s. The experiments used short self-presentations; a long working relationship may show more, though nobody has shown how much.
There is a second problem, and it is Arjun’s own. Once he knew that a post about a loss brings enquiries, the next loss became harder to write about honestly. Candour that is known to win work stops telling anyone anything, because everyone can see why it is offered. In April 2026 he paid for the reprint, and he began a post about it. He stopped halfway, when he caught himself choosing the sentences that would read best. He has not posted about a loss since. It has cost him some enquiries.
Use AI for distribution: scheduling, formatting, cross-platform adaptation. Do not use AI for voice. The reason is not that readers can tell. Most cannot. The reason is that whatever goes out under his name is a claim he must answer for. That is true of an invoice too, and a machine may draft his invoices. The difference is what the claim is about. An invoice states facts he can check line by line. A post about a loss reports what he believed and doubted. A machine’s draft would offer him doubts in the right places that he may not have had. He could approve every sentence and not know which were his. A brand is a promise made to strangers before they meet you. An honest brand says what the work is, who did which part of it, and what you will put right. Arjun now sends each client one page at the end of a project, as Chapter 20 told, saying what a machine made and what he chose. Nobody asked for it, and it is the most honest thing he sends out. The page is not a ritual. It should say what changes a client’s picture of the work: which parts a machine made, whether their files went to a model, and what he checked himself. A model that only tidied the spacing needs no line.
The print edition also advised letting the contradictions show, because “An entity without contradictions is either a machine or a performance. Both are detectable.” Neither half holds. Machines produce contradictions easily, and the experiments above found people unable to tell machine-written self-presentations from human ones. Nothing in them shows that a performance is easier to catch. What can be shown honestly is smaller. The work travels. A record of who did what can travel with it. The rest is a bet on strangers.
That is all “distribution without performance” can mean for someone who works alone. No method for seeming honest survives being copied. What is left, when you stop trying to look honest, is the work and its record going out under a name that can be questioned. If strangers cannot tell honest self-presentation from a machine’s, they will lean on other evidence: referrals, work they can see, repeat business, terms in writing and errors corrected in public. Reputation is what people believe. Evidence is what can be shown.
34.5 What Unscalable Interactions Are For
Unscalable interactions do not become your biggest advantage. An advantage is something you hold over others. What this section finds is a cost you carry for the people you work for.
The print edition began with a sweeping claim: “Every business book written before 2025 tells the solo practitioner to scale.” Some did not. Paul Jarvis’s Company of One, published in 2019, was subtitled Why Staying Small Is the Next Big Thing for Business.16 But the observation behind the claim holds. When output is cheap, output stops telling one maker from another.
In July 2026 a family came to Nandini about a wedding in February 2027. Before any ink touched paper, she asked them about both families. She asked about the names and who had chosen them, which grandparents were still living, and whose name the bride’s grandmother would look for first. The conversation lasted about ninety minutes, and she charged nothing for it. She says that such conversations change what her hand does on the page, in ways she cannot describe and does not try to. That is her claim, and nothing in this book can test it.
What can be seen is smaller. In that conversation she learned that the groom’s grandfather wrote his name in an older spelling than the one on the family’s typed list. She asked which they wanted on the card. The family chose the old spelling. A planner’s form could have asked about spellings. This one had not, and a form may not hear what a person says while pouring tea. Whether a machine asking the same questions would have heard it too is open. Some families would tell a screen more than a stranger. Others would tell it less.
The print edition explained such conversations with an old idea, without naming its authors. In 1973 the economist Michael Spence argued that a costly education can work as a signal to employers, because it costs the less able more to acquire.17 Two years later the biologist Amotz Zahavi proposed that some animal displays, like a peacock’s tail, are believable because they are costly to carry.18 Biologists still argue about how far Zahavi’s idea holds.19 Applied to clients, the idea ran like this. A call returned on a Saturday is a costly signal, and so is a daughter’s name remembered. “A gesture that costs nothing communicates nothing.” Then came the verdict: “AI cannot send costly signals.” The passage called this “structural, not a limitation that better models will overcome.”
Three things are wrong with that. A machine bears no cost, but whoever sells it can. A company that offers a refund, a guarantee or a named person to call is sending a costly signal too. Nor is cost the only way a signal earns belief: a cheap claim that can be checked can be believed as well. Second, this book holds that cost describes presence but does not test it. A long call can be a performance, and a short one can be complete. Third, and worst, the same passage advised readers to “treat their time not as a cost to be minimised but as a signal to be deployed.” A kindness booked as a signal is a different act. A call made in order to be recommended has the recommendation in mind, and the person called is being worked. Nandini’s ninety minutes are not a signal she deploys. If they were, they would be marketing.
Arjun has his own test of this. His client software keeps notes on the people he works for. Before a call in September 2026, it reminded him that the soap company’s founder had a daughter starting university. He asked after her. The software had remembered what he would have forgotten. Was the asking still his? What was his was whether he meant it, and what he did with the answer. The founder said that the fees had made the company cut its first order. He did not raise his price.
There is a harder fact about what does not scale. It is rationed, and someone decides who gets it. Nandini’s ninety minutes go to the families who can wait a year and pay her fee. The schools’ young women give less time to more families, at a price more families can pay. A waiting list is a ration by the date of asking; a fee is a ration by money. Neither is unjust in itself. But no one who praises unscalable attention should forget that most people will never receive it. Less time need not mean less respect. A family that pays less may get a more standard service. It should not get hidden risks, or no way to have a mistake put right.
So what are unscalable interactions for? They are where the promise is kept in person, and sometimes where it is first found: Nandini’s question about a spelling changed the brief. In them the work is explained, adjusted and put right by the one who made it and answers for it. They cost the keeper’s hours, and some clients do not want them; they want the work done. Nothing guarantees that anyone pays for them. A business built from one person’s work can grow, but then it promises something other than that person. Staying small and growing are both honest choices. What matters is that the form of the business fits what it promises. A studio cannot sell Arjun himself, any more than a print shop can sell Nandini’s hand. What a person who works alone can promise is herself, for as long as her hours and her hands last.
This is where the two old voices part for good. Cicero ranked a livelihood by what was bought, labour or skill, and by its size. He thought pay for labour lowered the one paid, and that only large trade, honestly done, escaped his contempt. The Hitopadeśa treated wealth as a rung between fitness and dharma, right conduct. Earning stains nobody. It is what fitness brings. Both readings, and their use for a designer, a calligrapher and a translator, are mine. They disagree about whether being paid lowers the person paid. On one point they agree, and this chapter cannot. Both assume that the buyer pays for what the seller is: his skill, or his fitness. Neither was written to explain a procurement department comparing a translator with a machine, or an app that sets a rider’s fare. Catalina’s fitness rose while her pay fell. Nandini cannot know whether families pay for her hand or for their friends’ eyes. Arjun’s avoidance was Cicero’s contempt in a designer’s clothes. Neither voice wins. The price is neither a stain nor a reward. It is a promise, stated in money, that someone can hold you to. Your worth stays outside it.
Ending: The Quoted Figure
Nandini’s hands ache in the mornings, and they will ache more in the years ahead. Her grandmother’s hands became unreliable in her fifties. Nandini has perhaps a decade of work at her present standard, and she knows it. She has begun, reluctantly, to look for an apprentice. She does not want to make the work larger. She wants to hand the practice on before her hands decide the matter for her. The difficulty is her standards. She has met young calligraphers with technique. She has met none with the quality she cannot name: the willingness to sit with the discomfort of imperfection long enough for the work to become honest. She is looking. She has not found.
In the second week of October, Arjun went through the month’s papers. Neha’s invoice had come on the first working day, as it did every month: one page, on time. He had paid it that afternoon. He had never been late with anyone else’s money.
Then he opened the soap company’s file. His quote in July had promised six labels and two rounds of changes, for a fixed fee. The work had taken nearly twice the hours he had allowed, because he had redrawn the lettering twice after the client approved it. Nobody had asked him to. He could have added a line about the extra hours. He did not. The quote was what he had promised. The hours past it, about five days of his own time, were his.
He attached the invoice at the quoted figure, wrote two lines to the accounts department apologising for the delay, and sent it. He replied to the stationery client about the timeline and sent the accountant the figures. He wrote to his mother that he had, in fact, eaten. The journalist’s message he read twice, closed and left unanswered. Something about the phrase “thriving in the AI era” felt like a costume he was being asked to wear, and he was not ready to decide whether it fit.
Practice
Read your own invoice. Take your last three pieces of paid work. For each, write down what you promised: the tasks, the choices, and what you would explain and put right if it failed. Beside it write the price, and the hours the work really took, including the ones you never billed: messages, revisions, paperwork, chasing payment. Then write what putting it right would have cost you, and whether the price covered it. Next, count how many such promises your month can hold before one of them is kept by nobody. Write who would keep them if you were ill for a month. List what would stop the business if it vanished tomorrow: a client, a platform, a model maker, a collaborator. Write who could reach your clients, pay what you owe and settle unfinished work if you could not. If there is an invoice you have not sent, for no reason you can name, send it. The hour this takes is your own and unpaid, which is part of the point. It will not raise your fee or bring you a client. It shows what you are selling, and how many of those promises your hours can keep.
Words to keep
- artes and operae: Latin, “skills” and “labours”; Cicero’s two things a buyer may pay for, the second of which he thought made a man unfree.
- costly signal: an action believable because it costs the one who makes it; an idea from economics (Spence) and biology (Zahavi), still debated.
- Devanāgarī: the script in which Hindi, Sanskrit and Marathi are written.
- gig worker: in India’s social security code, someone who earns from work outside a traditional employer-employee relationship.
- Hitopadeśa: Sanskrit, “Good Counsel”; a book of fables by Nārāyaṇa, framed as lessons for a king’s sons.
- localisation: adapting a translated text to one country’s usage, law and register, not only its language.
- nastaʿlīq: the flowing Persian style of handwriting in which Urdu is written.
- pātra, pātratā: Sanskrit, a vessel or fit recipient, and being one; “worthiness” (Müller), “fitness for employment” (Johnson).
- promise capacity: the hours, money, skill and authority available to deliver, check, keep up and put right what has been sold; distinct from how much work can be produced.
- quote: here, a promise stated in money: what will be done, by when, for how much, and what will be put right.
- registro sanitario: Spanish, the public document by which Colombia’s medicines regulator permits a medicine to be made, imported or sold.
- solopreneur: a blend of “solo” and “entrepreneur”, first recorded in 1992; in this chapter, the one person who carries a business’s promises to its clients.
- vidyā, vinaya: Sanskrit, knowledge, and the discipline or right conduct it brings.
Sources
- Central Goods and Services Tax Act, 2017, s. 16(2)(a): a registered person may take input tax credit only if “he is in possession of a tax invoice or debit note issued by a supplier registered under this Act” (CBIC, taxinformation.cbic.gov.in); since 1 January 2022, s. 16(2)(aa) also requires that “the details of the invoice or debit note referred to in clause (a) has been furnished by the supplier” in his statement of outward supplies “and such details have been communicated to the recipient” (inserted by the Finance Act, 2021, s. 109; Notification 39/2021–Central Tax; in force from 1 January 2022; CBIC, taxinformation.cbic.gov.in, checked 31 October 2026). Section 49(4): “The amount available in the electronic credit ledger may be used for making any payment towards output tax” (CBIC, taxinformation.cbic.gov.in, checked 31 October 2026). On late payment: Micro, Small and Medium Enterprises Development Act, 2006, s. 15 (for goods or services, a buyer pays “on or before the date agreed upon between him and the supplier in writing”, or, with no agreement, within fifteen days of acceptance under s. 2(b); the agreed period may not “exceed forty-five days from the day of acceptance or the day of deemed acceptance”), s. 2(n) (a “supplier” is a micro or small enterprise “which has filed a memorandum” under s. 8, since July 2020 through Udyam registration) and s. 16 (compound interest with monthly rests “at three times of the bank rate notified by the Reserve Bank”); text as published by PRS Legislative Research (prsindia.org), checked 31 October 2026. On gig workers: Code on Social Security, 2020, s. 2(35) (a gig worker performs work or takes part in a work arrangement and earns from it outside a traditional employer-employee relationship) and s. 114(1) (“The Central Government may frame and notify” social security schemes for gig workers and platform workers), both in force from 21 November 2025 (Ministry of Labour and Employment, S.O. 5319(E), labour.gov.in). On the label: Dharangadhara Chemical Works Ltd v. State of Saurashtra, AIR 1957 SC 264, which took the right to control the manner of the work as the first test of employment; later courts have weighed other factors too.
- Introduction to Cicero, De Officiis, trans. Walter Miller, Loeb Classical Library (London: Heinemann; New York: Macmillan, 1913), LacusCurtius: “a letter addressed to his son Marcus ... at this time a youth of twenty-one, pursuing his university studies in the Peripatetic school of Cratippus in Athens”. The work was written in 44 BCE.
- Cicero, De Officiis I.150–151, trans. Walter Miller (Loeb Classical Library, 1913), pp. 153–155, LacusCurtius (penelope.uchicago.edu), and the Perseus Digital Library transcription of the same translation (checked 31 October 2026). Latin of I.150, from Miller’s facing text (p. 152) as transcribed by the Perseus Digital Library (phi0474.phi055.perseus-lat2, from the 1928 reprint; checked 31 October 2026): “Illiberales autem et sordidi quaestus mercennariorum omnium, quorum operae, non quorum artes emuntur; est enim in illis ipsa merces auctoramentum servitutis.” José Luis Vidal’s copy, “Visions del treball en Grècia i Roma” (Societat Catalana d’Estudis Clàssics, scec.iec.cat, PDF, item 4), spells it “Inliberales”. Winterbottom’s Oxford text (1994) not checked.
- Merriam-Webster, “solopreneur” (merriam-webster.com, checked 29 October 2026): “a solo entrepreneur”; “First Known Use: 1992”.
- Anil R. Doshi and Oliver P. Hauser, “Generative AI Enhances Individual Creativity but Reduces the Collective Diversity of Novel Content,” Science Advances 10, eadn5290 (12 July 2024): 293 writers and 600 evaluators; “The story must be exactly eight sentences long” (Methods); the model’s ideas helped “especially among less creative writers” (abstract); see Chapter 3.
- OpenAI, “Introducing ChatGPT Images 2.0,” 21 April 2026 (openai.com; a sample poster “combining Japanese, Arabic, Korean, Devanagari, Cyrillic, Bengali, Greek, Chinese, and Latin scripts”); OpenAI quoted in PetaPixel, “OpenAI Claims ChatGPT Images 2.0 Can Think,” 21 April 2026: “significant gains in non-Latin text rendering, particularly in Japanese, Korean, Chinese, Hindi, and Bengali”; also Engadget (Igor Bonifacic), “ChatGPT Images 2.0 is better at rendering non-Latin text,” 21 April 2026 (OpenAI’s claim of “significant gains” for Japanese, Korean, Chinese, Hindi and Bengali).
- Nārāyaṇa, Hitopadeśa, Prastāvikā (introduction), verse 6, Sanskrit text as at wisdomlib.org (Hitopadesha [Sanskrit], verse 0.6; sandhi normalised in pātratvād). The oldest known manuscript, found in Nepal, is dated 1373 (“Hitopadesha,” Wikipedia, accessed 23 October 2026). The frame story of King Sudarśana of Pāṭaliputra and his sons follows the introduction (Johnson, below).
- “JOHNSON, Francis,” Who Was Who in Indology (whowaswho-indology.info): “In 1824-55 Professor of Sanskrit, Bengali and Telugu at Haileybury”, the East India Company’s college.
- Francis Johnson, trans., Hitopadesa, or Salutary Counsels of Vishnu Sarman (London: Wm. H. Allen; Hertford: Stephen Austin, 1848), “translated literally from the original, into English, for the use of the Sanskrit student”, introduction, verse 6, p. 1; preface dated “East-India College, Herts., 25th March, 1848” (Internet Archive, hitopadeaorsalu00johngoog).
- Max Müller, ed., The First Book of the Hitopadesa: Containing the Sanskrit Text, with Interlinear Transliteration, Grammatical Analysis, and English Translation (London: Longman, Green, Longman, Roberts, & Green, 1864), p. 2, the interlinear English of the verse, where pātratām is “worthiness” (Internet Archive, firstbookofhitop00ml).
- Monier Monier-Williams, A Sanskrit-English Dictionary (Oxford: Clarendon Press, 1899), s.v. pātra: “a drinking-vessel, goblet, bowl, cup, dish, pot, plate, utensil etc., any vessel or receptacle”; “a capable or competent person ... any one worthy of or fit for”; s.v. pātratā: “the being a vessel or receptacle for”; V. S. Apte, The Practical Sanskrit-English Dictionary, s.v. pātra: “a person fit or worthy to receive gifts” (both as at wisdomlib.org, checked 23 October 2026).
- Directive 2001/83/EC of the European Parliament and of the Council of 6 November 2001 on the Community code relating to medicinal products for human use, Article 6(1), Spanish text (EUR-Lex): “No podrá comercializarse ningún medicamento en un Estado miembro sin que la autoridad competente de dicho Estado miembro haya concedido una autorización de comercialización” (in force as of October 2026; a new directive adopted by the Council in September 2026 will replace it once published).
- Republic of Colombia, Decreto 677 de 1995 (26 April 1995), art. 2 (“Registro sanitario: Es el documento público expedido por el Invima o la autoridad delegada”) and art. 19 (“Todo medicamento requiere para su producción, importación, exportación ... de registro sanitario”); text as published by the Secretaría Distrital de Salud de Bogotá.
- “A third of translators report losing work to generative AI systems, SoA survey reveals,” The Bookseller, April 2024; “A third of translators report losing work due to AI,” Books+Publishing, 15 April 2024 (787 responses to a survey of the Society of Authors’ members and other authors; “over a third of translators (36%) have lost work due to generative AI”); European Writers’ Council, “SoA Survey UK: A third of translators and quarter of illustrators losing work to AI” (europeanwriterscouncil.eu, 9 May 2024): the survey ran in January 2024, went to the Society’s “12,500 members and other authors” and drew 787 responses, and “over 4 in 10 translators (43%) say the income from their work has decreased in value because of generative AI”. On automation complacency: Raja Parasuraman and Dietrich H. Manzey, “Complacency and Bias in Human Use of Automation: An Attentional Integration,” Human Factors 52, no. 3 (2010): 381–410, doi:10.1177/0018720810376055; abstract: complacency “occurs under conditions of multiple-task load, when manual tasks compete with the automated task for the operator’s attention”. On the trade’s own surveys (for “What would prove this wrong?”, 34.3): European Language Industry Survey 2025: Trends, expectations and concerns of the European language industry (ELIS; knowledge-centre-translation-interpretation.ec.europa.eu, PDF, checked 31 October 2026), p. 5: “initiated in 2013 by EUATC” and co-organised with ELIA, FIT Europe, GALA, the EMT university network, the European Commission’s LIND group and Women in Localization; 1,322 respondents from 50 countries, 654 of them independent professionals; “Users of this report need to keep in mind that the data are not objective and verifiable.” P. 20, on independent professionals: “a drop in work volume and the generalised move from human translation to discounted and less-valued post-editing make 53% of the participants question their future as a freelance professional.”
- Maurice Jakesch, Jeffrey T. Hancock and Mor Naaman, “Human heuristics for AI-generated language are flawed,” Proceedings of the National Academy of Sciences 120, no. 11 (published online 7 March 2023; issue of 14 March 2023), e2208839120, doi:10.1073/pnas.2208839120; abstract: “In six experiments, participants (N = 4,600) were unable to detect self-presentations generated by state-of-the-art AI language models in professional, hospitality, and dating contexts.” First posted as arXiv:2206.07271 (June 2022).
- Paul Jarvis, Company of One: Why Staying Small Is the Next Big Thing for Business (Boston: Houghton Mifflin Harcourt, 2019), ISBN 978-1-328-97235-4.
- Michael Spence, “Job Market Signaling,” Quarterly Journal of Economics 87, no. 3 (1973): 355–374.
- Amotz Zahavi, “Mate selection: a selection for a handicap” (the title’s dash given as a colon), Journal of Theoretical Biology 53, no. 1 (1975): 205–214.
- Dustin J. Penn and Szabolcs Számadó, “The Handicap Principle: how an erroneous hypothesis became a scientific principle,” Biological Reviews 95, no. 1 (2020): 267–290 (published online October 2019); for the defence, Alan Grafen, “Biological signals as handicaps,” Journal of Theoretical Biology 144, no. 4 (1990): 517–546.
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 “Judgement, taste, trust. For the solo practitioner, these are not assets of the business. They are the business.”, that everything “rests on a single consciousness. And the quality of that consciousness determines everything”, “Consciousness first, intelligence second”, and that consciousness “is the whole fabric”; that “every hour you spend on work that a machine could do is an hour stolen from the work that only you can do”; that what clients pay for is accountability because “When you hire a machine, no one is responsible for the result”, a premium “for the willingness to be wrong, to be accountable, to bear consequences that a machine never bears”; that “Fewer clients means less revenue. In practice, the reverse occurs, because depth commands a higher fee than volume” and “They pay accordingly”; that “The premium on consciousness is real”; that readers “can feel the difference” between lived and optimised content and “The distinction is not subtle”, that the most effective brand is “built from the work itself”, and that “An entity without contradictions is either a machine or a performance. Both are detectable”; that “Every business book written before 2025 tells the solo practitioner to scale”, that “AI cannot send costly signals ... This is structural, not a limitation that better models will overcome”, and that the solo should treat time “as a signal to be deployed”. It also said that the Goa hotel chain had “cost him his biggest client” (the chain renewed him); told of a coffee brand and “four months” of machine-led work that no chapter contains; had a textile exporter from another chapter prompt the invoice with a call about a client’s labels; gave Arjun a sister whose verdict converted him; set a Bogotá professional flinching at droga, which Spanish uses for medicine and which Colombia’s pharmacies, the droguerías, take their name from; added a tax adviser in Lagos and three one-off professionals in Manila, Santiago and London; named the calligrapher Nandini Sharma and the translator Elena Vargas; and ended on “the eighteen months that would follow” and “the subject of the next chapter”.
- What I now hold: someone who works alone sells a promise: tasks (often done by machines), an eye that chooses, and a person who will explain and put right. Judgement and taste are how it is kept, not a remainder machines cannot reach. The business side is part of the promise: a quote states it in money and the invoice is its last line. When you are the whole business, the institution’s half is yours: the hours to check, the right to refuse, the record, the price and who keeps the promise if you cannot; the people who pay you hold the rest. A price says what you will stand behind, not what you are worth. A company that sells machine output answers for it too; one person can sell only her own answering, and not every buyer needs it. Capacity is not a promise: a machine adds capacity, not the hours in which to check and put right, and depth commands a fee only where buyers want it and can tell it apart. People mostly cannot tell machine-written self-presentations from human ones, and candour known to win work stops being evidence of candour; an honest brand says what the work is, who did which part, and what will be put right. Costly signalling (Spence; Zahavi) is credited and contested; cost describes presence and does not test it, and a kindness booked as a signal is a different act. Unscalable interactions are where the promise is kept in person, at the keeper’s cost; nothing guarantees that anyone pays for them. Answering is a position: a company that sells machine work holds it too, and one person adds only that the one who answers is the one who chose. A price’s height is set by buyers as much as by the seller; what it covers is hers to say. Where an app sets the price, much of the institution’s half is the app’s. A machine may draft what can be checked line by line, not a report of oneself. “Conscious” in the title can mean only “knowingly”. A promise has parts that come apart (delivery, standard, checking, upkeep, remedy), and willingness to answer is not capacity to repair. Productive capacity is not promise capacity. Working alone is not bargaining on equal terms, and independence in law can hide dependence on buyers, platforms and toolmakers. A price is an agreement under conditions, not a measure of worth or of contribution, and part of every price is the owner’s future. Someone who works alone with a freelancer at a desk has become close to an employer.
- What changed my mind: the Prologue’s rule against defining people by what machines cannot yet do; the print edition’s own story of Elena (now Catalina), whose fee fell as her judgement grew; Cicero’s split of pay for labour from pay for skill, which showed where the “remainder” argument came from; Johnson’s “fitness for employment”, which showed the promise of a wage in an old and loved form; the 2023 experiments on AI-written self-presentations; OpenAI’s April 2026 claim for Hindi lettering; Colombia’s medicines decree and the RAE’s entry for droga; Company of One (2019); the dispute over Zahavi’s handicap principle; the audit’s corrections; India’s laws on late payment to small enterprises and on gig workers; research on automation complacency; and three outside readings of an earlier version of this chapter, in October 2026.
- What survives: the opening, word for word in most of its sentences, with the invoice, the paper merchant in Dadar, the sharbat and the spreadsheet, and “avoidance feels like principle”; the two sentences on one person at a team’s scale and indistinguishable from every other; “what, specifically, am I providing ...?”; “The inversion is a principle, not a partition.”; Nandini’s warm water, the nib and her breath, “The standard explanation is correct. It is also incomplete”, the status signal, the young women whose work “she cannot honestly say is bad”, the relic, and “her refusal to train an apprentice quickly enough is a form of hoarding dressed up as standards”; “Both conditions are necessary. Neither is sufficient alone.”; the translator who earns less as a checker for work that requires more judgement, and “an argument, not a law of nature”; “It sounds like a person without sounding like any person in particular.”; “Use AI for distribution ... Do not use AI for voice.”; the ninety-minute conversation; the apprentice not found; the journalist’s “costume”.
- Section titles in the print edition: 34.3 was “Operating at 10x Scale Without Losing Your Soul”, 34.4 “Building a Personal Brand That Reflects Genuine Consciousness” and 34.5 “Why Unscalable Interactions Become Your Biggest Advantage”. 34.1 and 34.2 keep their print titles.