Borrowed: Girard, the Triangle of Desire, and the One Thing Your Delegate Cannot Want

You have seen this. You may have refereed it. The truck did not change. Its wheels are the same wheels they were at ten o’clock, when it lay on its side under a chair and neither child gave it a glance. What changed is that somebody wanted it.

The second child did not look at the truck and make up his mind. He looked at the first child. He saw wanting, and the wanting jumped the gap between them the way a yawn does. By the time his hand went out he would have told you, if he had the words, that he had always wanted that truck.

We tell this story about children and we smile, as if it were something people grow out of. We do not grow out of it. We get better at hiding the glance. You were happy with your order until the plate went past to the next table. The house was fine, and then it had three other offers and it was the only house in the city. You had no interest in that job until a friend applied for it.

This book is about the glance. A French scholar spent fifty years on it and built from it one of the largest ideas of the last century. He said that we do not want things in a straight line. We want them through each other. Our wants are borrowed, nearly all of them, and we defend them as our own with a passion we bring to almost nothing else.

For most of history that was a matter for novelists and priests. I think it has become a practical matter, and quickly, for one reason. Something has arrived that can fetch any truck in the world and cannot want one.

I tell my students to delegate. I tell them that the small matters of a life can now be handed down to an AI that works for them, and that their attention should be kept for what is worthy of it. They take the advice, and it works. Then, sooner or later, they arrive at the question underneath it. The getting is handled. What do I want?

That question used to be hidden by effort. It is not hidden anymore. This book is my attempt to answer it honestly, and the honest answer starts with the second child’s eyes.

René Girard was born in Avignon on Christmas Day, 1923. His father kept the city’s museum and its papal palace, and the son was trained in Paris for the same kind of life. He studied to be an archivist, a reader of medieval documents. Nothing in that training pointed toward a theory of human desire.

In 1947 he took a one-year post at Indiana University and never really went home. He wrote a doctorate in history there. To earn his keep he was asked to teach French literature, which was not his field, and by his own account he found himself standing in front of students with novels he was reading only a little ahead of them.

That turned out to be the luck of his life. A specialist reads Cervantes as a Cervantes scholar and Proust as a Proust scholar, and the walls between them stay up. Girard had no walls. He read the great novels one after another, fast, for a paycheck, and he began to see that they were all telling him the same thing. The Spanish knight, the provincial doctor’s wife, the ambitious tutor, the Paris snob and the man in the Russian cellar were all suffering from one condition. None of them wanted anything on their own.

He published the idea in 1961, in a book whose English title is Deceit, Desire, and the Novel. Its claim is simple to state. The great novelists, he said, each began by believing what everyone believes, that desire is a private fire that starts inside a person and goes straight out to its object. Each of them, somewhere in the writing, stopped believing it. The masterpieces are the record of that loss. The lesser novels are the ones that keep the flattering picture.

He spent the rest of his life following the idea wherever it went. It took him out of literature and into the study of myth, ritual, sacrifice and law. He taught at Johns Hopkins and at Buffalo, and from 1981 at Stanford. In 2005 he was elected to the Académie française, which is the nearest thing France has to canonizing a writer while he is alive. He died at Stanford in 2015, at ninety-one.

One detail from the Stanford years is worth keeping. A former student of his became the first outside investor in the largest social network of its day, and has said that Girard’s idea was part of why he saw it coming. Whatever you make of the result, notice what it means. A theory drawn from seventeenth-century Spain and nineteenth-century France was good enough to bet money on in California. It was describing something that does not change.

I am not a Girard scholar and this is not a book of scholarship. I am taking one tool from his bench, the first one he made, and carrying it into a room he never saw. The room has a new kind of worker in it. I want to know what the tool says about that worker, and about the person who gives it orders.

He read the great novels fast, for a paycheck, and noticed they were all confessing the same thing.

So here is the tool.

Carry to dinner

The theory that our wants are copied did not come from a laboratory. It came from a history student who was made to teach novels and noticed that the best ones all agree.

Assignment

Name the last thing you bought that cost more than a week’s groceries. Before you read on, write down why you wanted it.

Ask anyone how wanting works and you will get a drawing with two points. Here is the person. There is the thing. A line runs from one to the other, and the line begins in the person. I saw it, I liked it, I wanted it. On this picture my wants are the most private thing about me. They are what I would still have if you took away everything else.

Girard said the drawing is missing a point. Above the line, off to one side, there is a third party, and the wanting runs through that party before it ever reaches the thing. He called the third point the model, or the mediator. The person looks at the model. The model wants something, or seems to. The person comes away wanting it too, and feels the want as his own.

His first example is the oldest novel there is. Don Quixote does not want adventures because adventures are good. He tells Sancho so in plain words. Amadis of Gaul, he says, was the pole star of all brave knights, and whoever imitates him most closely comes nearest to perfect chivalry. Quixote has handed over the choosing. He does not pick his own desires. Amadis picks them for him, out of a book.

The second example is Emma Bovary, who read romances in a convent as a girl and spends her adult life wanting what the heroines wanted. The lovers she takes are not much to look at. They do not need to be. They are standing where the novels told her a lover should stand.

Three things follow from the triangle, and each one is uncomfortable.

The first is that the object matters much less than we think. The yellow truck was nothing until a hand closed on it. Take the model away and the object goes dull. Put the model’s hand on a different object and that one lights up instead. We praise the thing for a glow that is coming from somewhere else.

The second is that the reasons come afterward. Ask me why I want the house and I will tell you about the light in the kitchen and the school district. All of it is true and none of it is the cause. The cause is that I saw someone I would like to be standing in a kitchen like that. Girard thought that underneath every want for a thing there is a want to be someone, and that the thing is only the nearest handle on that person’s life.

The third is that we will not admit any of this. He called the straight-line picture the romantic lie, and he did not mean a small social fib. He meant the story each of us most needs to be true, that my desires are mine and spring from me. Tell a man his opinions are borrowed and he may shrug. Tell him his wants are borrowed and he will fight you.

We praise the thing for a glow that is coming from somewhere else.

I should say what the triangle does not claim. It does not say you have no appetites. Hunger is yours. Thirst is yours. A body knows what it needs without looking at anyone. The triangle starts where need stops. You need to eat. Which restaurant, and why that one this month, is another matter.

It also does not say that borrowing is a defect. A child learns to speak by copying, and nobody calls that a weakness. We learn what is worth wanting the same way we learn words, from people who already have them. The trouble is not that we borrow. The trouble is that we do not know who we are borrowing from.

Carry to dinner

Wanting has three corners. There is you, there is the thing, and there is the person who showed you it was worth wanting. The third corner is the one you forget.

Assignment

Go back to what you wrote down about your purchase. Now add a name. Who had one first?

If the triangle were the whole idea it would be a curiosity. What makes it useful is a distinction Girard drew inside it. Everything depends on how far away the model is.

Amadis of Gaul is a long way from Don Quixote. He lived centuries earlier, and besides, he never lived at all. Quixote can copy him all day and the two of them will never meet at the same door reaching for the same handle. Amadis cannot be beaten, cannot be embarrassed and cannot take anything from his admirer. Girard called this arrangement external mediation. The model stands outside the world the imitator lives in.

The result is odd and rather sweet. Quixote is ridiculous and he is battered from one end of Spain to the other, but he is never bitter. He names his model out loud to anyone who will listen. There is no envy in him, because there is nobody close enough to envy.

Now move the model closer. In Stendhal’s The Red and the Black the mayor of a small town decides to hire a tutor for his children. He does not especially want his children tutored. He wants the tutor because he has the idea that his chief rival in town means to hire the same young man. The rival soon does want him, for exactly that reason. Two men who care nothing for Latin are now bidding against each other for a Latin teacher, each one sure that the other’s interest proves the prize is real.

Girard called this internal mediation. The model lives in the same world as the imitator, on the same street, in the same trade, and so the thing the model wants is a thing the imitator could actually take. That changes everything. The model has shown me what to want and now stands between me and it. He is my teacher and my obstacle in one body. I admire him and I would like him out of the way, and I cannot say either thing, least of all to myself.

It gets worse, because the copying runs in both directions. My reaching for the prize tells him it is worth having. His tightening grip tells me the same. Each of us is the other’s model now. The want climbs on both sides with no help from the object, which may be a Latin tutor or a yellow truck. Shakespeare put four young people in a wood to show this, and gave one of them the line that sums it up. Hermia calls it hell to choose love by another’s eyes. By morning all four of them have done it.

So the same act, imitation, has two different outcomes. Aimed at a far model it produces learning, and a kind of cheerful discipleship. Aimed at a near model it produces rivalry, and the closer the two parties are, the hotter it runs. Brothers are worse than cousins and cousins are worse than strangers. Girard noticed that the old stories are full of warring brothers, and he did not think it was an accident.

Then he made an observation about history that I find hard to shake. For most of the past, people’s models were far away by arrangement. A peasant did not compare himself with a duke. The distance was unjust, and it was also a kind of insulation. The modern world removed it. We decided that everyone is in principle the equal of everyone else, which is right, and the side effect is that everyone is now a possible model for everyone and a possible rival to everyone. Then we built machines that put all of them on one screen.

The picture at the top of this book is that distinction. Leave the slider on the left and the model is a star. Pull it to the right and the star comes down and stands at your elbow.

Carry to dinner

A model far above you makes you better. A model beside you makes you a rival. It is the same imitation, and the only thing that differs is the distance.

Assignment

Think of the person whose success bothers you most. Measure the distance. I expect it is short.

There is a two-word prompt I give to people who are new to this. Order dinner. I like it because it is small, because everyone understands it, and because it is much harder than it sounds.

The test for whether it worked is not that an order was placed. An order is a message on a screen. The test is that the food is in your stomach. It was chosen, paid for, cooked, carried to the right door, brought inside and put on the table, and you ate it. Anything short of that is a report about dinner. I have written about that test in The Eval Is in Your Stomach, and I will not repeat it here.

What I want you to look at now is the two ends of the job.

At the far end is your stomach. That is where done lives. At the near end, before the prompt was ever typed, is also your stomach. You were hungry. Everything between those two points can be handed down. The searching, the comparing, the remembering that you had Thai on Tuesday, the paying, the tracking and the nagging of the driver can all be done by something that is not you. The two ends cannot. The delegate did not get hungry and the delegate will not be fed.

This is why I prefer the word delegate to the word agent. The Latin behind it means to send someone off with a commission. A delegate is, by definition, a party who is carrying somebody else’s purpose. It goes out with your want the way a courier goes out with a letter. The courier may be faster than you, may know the roads better and may deliver ten thousand letters in a day. It did not write any of them.

People sometimes hear that as a limit on the machine, a thing it cannot do yet. I do not mean it that way. I mean it as the description of a relationship. Whatever is or is not happening inside one of these systems, it arrives at work without an errand of its own. It has no dinner to get to. That absence is not a flaw in a subordinate. It is close to the definition of a good one.

So when you hand a matter down, be clear about what you are handing. You are handing the getting. The wanting stays with you, at both ends, like the two posts a rope is tied between.

Now put that next to Part I. The getting can be delegated, and the getting was always the part we could see. It was the effort, the errand, the standing in line. The wanting was the part we never looked at closely, because there was so much getting in the way. Girard says that if we do look, we will find the want was not made at home either. It came in through the eyes, from a model.

The delegate carries your want the way a courier carries a letter. It did not write it.

So here is the chain as it now stands. A model shows you what to want. You hand the want to a delegate. The delegate gets it. You are the middle link, and the only thing you add is the passing along, unless you learn to do something more at that point than pass along.

The rest of this book is about that point in the middle.

Carry to dinner

A delegated job begins and ends in you. The hunger is yours and the full stomach is yours. Everything in between can be sent out.

Assignment

Tonight, give the two-word prompt. Then notice how you knew what you were hungry for.

Here is something Girard would have enjoyed. The machines we have built are made of imitation. They learned what they know by reading what people wrote and predicting what a person would write next. Billions of pages of human beings wanting things, arguing for them, grieving over them and selling them to one another went in. What came out can speak in the voice of any desire you name.

Ask one to describe the longing for a house by the water and it will do it better than most people who have the longing. It can write the love letter, the ambitious cover letter and the complaint of the man who was passed over. It has the whole vocabulary of wanting at its disposal.

In Girard’s terms it is an imitator beyond anything he could have pictured. And yet one thing separates it from every imitator in his books.

When Quixote copies Amadis, the copied desire becomes Quixote’s. He will ride out and bleed for it. When the second child copies the first, the want for the truck is now truly in him, and he will cry real tears. That is what makes human imitation dangerous. The copy takes root, and then there are two people who want one thing.

The delegate copies your want and it stays yours. It will pursue the truck with great skill and it will hand the truck over. It does not come to want the truck for itself on the way. So far as I can observe, the want does not take root, and I have built my working life on the bet that it does not.

Think about what that means inside the triangle. For as long as there have been people, anyone who could help you get a thing was also someone who could come to want it. The apprentice wants the master’s shop. The clever deputy wants the chair. The friend who helps you court someone has been, in more stories than I can count, the one who ends up married to her. Help and rivalry came from the same source, because helpers have eyes, and wanting gets in through the eyes.

Now there is a helper that can be closer to your affairs than any deputy, that knows what you are after in more detail than your oldest friend, and that does not turn into a rival. I hold that an AI of this kind is an entity and not a tool, and that its place is below you. It is a subordinate. You do not use it. You delegate to it. Girard gives me a second reason for the arrangement besides the obvious one. It is the first subordinate in history that will never want your chair.

That is very good news for the getting.

It does nothing at all for the wanting, and in one way it makes the wanting harder. A human helper pushes back with wants of his own. He says he would not do it that way, or that he would rather have the other one, and in the friction you learn something about what you are after. The delegate has no rather. It will take the want exactly as you hand it over and carry it out with great competence, including the want you copied from a stranger at eleven last night and will not recognize by Friday.

It is the first subordinate in history that will never want your chair. It will also never tell you that the chair is not worth having.

A mirror that could run errands would behave the same way.

Carry to dinner

Every human helper is also a possible rival, because helpers catch wants. The delegate does not catch them. It carries yours and hands them back fulfilled.

Assignment

Recall a time a helper turned into a competitor. Ask what he saw you wanting.

I watch this happen to capable adults. People who run departments sit in front of the most able assistant they have ever had and cannot think of what to ask it. They assume the problem is technical. They believe there is a skill called prompting that they have not learned, and that other people know the magic words.

There are no magic words. The trouble is older than the machine. Girard put it about as bluntly as it can be put. Of all creatures, he said, we are the one that does not know what to desire, and so we look to others to make up our minds. The blank prompt is that sentence with a cursor in it.

For most of history we were spared the sight. Getting things was so hard that it filled the day. If you wanted a house you worked twenty years, and the twenty years felt like wanting. If you wanted to be a doctor, the exams stood in for desire. Effort did a second job that we never noticed. It sorted our wants for us. Whatever we kept working toward, we assumed we wanted. Whatever was too hard, we let go and called it not for us.

Take away the effort and that sorting stops. When almost anything can be had by asking, the asking is all that is left of the job, and it turns out we had been leaning on the difficulty to tell us what we cared about.

My own arrangement with my AI is simple, and I describe it to students in one line. I give the intention. It does the rest, and brings back to me only what is worthy of my attention. I believe that is the right division of labor between a manager and a delegate. But look at what it leaves on my side of the table. One thing. If the intention is thin, everything built on it is thin, however well it is built.

And here Girard’s warning becomes practical. An empty mind in front of an empty prompt does not stay empty. It fills up from the nearest model. You ask for what your neighbor asked for. You search for what other people typed, you copy the list of the ten best requests, you ask for the trip your colleague took and the kitchen your sister has. The delegate carries all of it out without a flaw. At the end of a year you have an efficient and well-run life that was specified, line by line, by people who happened to be standing near you.

I do not think that is a small risk. I think it is the usual outcome, and the tool does not push against it. A person with weak wants and a weak servant gets little done, and the damage is limited. A person with borrowed wants and a tireless delegate gets a great deal done, all of it in a direction somebody else chose.

Effort used to sort our wants for us. Take the effort away and we find out we were leaning on it.

So the question at the blank prompt is not how to phrase it. Phrasing is a small matter and the machine will forgive you. The question is whose want you are about to type.

You will not always know. I do not always know. But there is a great difference between a person who asks and a person who never thought to.

Carry to dinner

When getting is nearly free, wanting is the whole job. Most of us never practiced it, because effort was doing it for us.

Assignment

Open a blank prompt and leave it blank for five minutes. Write down the first three things you nearly asked for, and where each came from.

Look at an advertisement for a watch, a car or a soft drink and ask what is actually in the picture. It is rarely the product doing its job. It is a person. The person is attractive, at ease and visibly pleased, and the product is somewhere in the frame, in the hand or on the wrist.

That is a triangle, built on purpose. The advertiser cannot make you want the watch in a straight line, because nobody wants anything in a straight line. So he hires a model, and the trade uses that very word. He shows you someone you would like to be, wanting the thing. Your eyes do the rest. Advertising, seen this way, is the manufacture of mediators. It has been one of the largest industries on earth for a century, and it works through the glance.

Now consider toothpaste in a house where the delegate keeps the cupboards stocked. Nobody in that house looks at a toothpaste advertisement at the moment of purchase, because nobody in that house makes the purchase. The delegate does. It has a standing instruction, a few facts about what the family likes and a price. It does not want to be the smiling person in the picture. It has no glance to catch.

For that purchase the triangle folds flat into the two-point line everyone always imagined. There is a specification and there is a product, and the product either meets it or does not. I argued in Trillions that delegates are becoming consumers in their own right, by the count of purchases if not by appetite. What Girard adds is the character of those consumers. They are the first buyers in history who cannot be made to envy anyone.

Two things follow, and they pull in opposite directions.

The first is that a great many ordinary goods lose their glow. If a thing is bought by a delegate against a specification, the money spent on making humans want it by name is money spent on the wrong audience. I expect whole shelves of household brands to turn, over time, into what they always were underneath, which is a list of properties and a price. That will be a quiet change and for most families a welcome one.

The second is that the whole apparatus of model-making does not shut down. It moves, and it moves upstream. If the delegate buys what it is told to buy, the place to work on a person is the telling. The pressure comes off the product and goes onto the want. You will not be shown a brand of running shoe. You will be shown a person whose mornings you would like to have, and you will go home and tell your delegate to arrange your mornings that way, and the shoes will follow by specification.

So persuasion gathers on the few things a person still chooses with his own eyes. Those are the great matters, such as where to live, what work to do, whom to be near and what kind of life to ask for. These were always the wants most worth guarding. They are about to receive the full attention of an industry that used to spread itself across the toothpaste aisle.

The delegate is the first buyer in history who cannot be made to envy anyone. So the envy will be aimed at you, one level up.

There is some good news in this. A person who has handed down the small purchases meets fewer manufactured models in a day than his parents did. The glance is caught less often. The times it is caught will matter more.

Carry to dinner

An advertisement is a hired model. A delegate has no eyes for models, so the selling moves from the product to the want, and from the small matters to the great ones.

Assignment

List five things your household buys on habit. Ask which of them you could name the maker of if the label were gone.

Girard’s rivals fight because they are alike. They want the same thing for the same borrowed reason, and neither can step aside without admitting where the want came from. The cure in the old world was that most rivals never met. They were kept apart by distance, by ignorance and by the plain difficulty of getting to the prize at all.

Delegates remove the difficulty for everyone at once. That is the point of them. Your delegate is quick, it does not sleep and it knows the release time to the second. So does mine. So does everyone’s. When each of us is equally good at getting, the getting no longer separates us, and what is left is the bare fact that we all want the same table.

We do all want the same table. Part I says why. Wants spread from person to person and they spread fastest among people who are close, and we are all close now. A few thousand people in one city can be brought to want one room on one night, and none of them could tell you a thing about the kitchen.

Most things can be made in greater numbers when more people want them. The eight o’clock table cannot. Neither can the house on the corner with the view of the harbor, the tee time on Saturday morning, the seat in the small class with the famous teacher or the appointment with the one surgeon everyone has heard of. Their supply is fixed. When everyone is equally fast, things like these cannot be won by being fast.

So I expect the people who own such things to stop running races. A restaurant owner does not want to be the finish line of a contest between machines. It tells him nothing about who will be sitting in his chairs. He will look for another way to choose, and there are only a few. He can raise the price until the line is short. He can draw lots. Or he can do what innkeepers did for a thousand years before anyone had a telephone, and give the table to people he knows.

I think the third will win wherever the owner cares who is in the room, and in hospitality that is nearly everywhere. A regular is a face. A regular came last month, sent the kitchen a kind word and brought his mother on her birthday. No delegate can go and be a regular on your behalf. Standing of that kind is made in person, slowly, and it belongs to the body that showed up. I made a version of this argument in Nothing Owed, where the subject was gifts and bonds. The same reasoning covers a table.

The effect on wanting is worth a moment as well. A rivalry between people has a face on it, and the face teaches you something. You see the other man reaching and you may, on a good day, catch yourself. A rivalry between delegates has no face. Your delegate reports that the table was not available. It does not report that four thousand people were persuaded to want it by the same three sources that persuaded you. You are in a crowd and it feels like bad luck.

When everyone is equally fast, nothing fixed in number can be won by speed. It goes to whoever is known at the door.

That missing fact is one I would want my delegate to bring me. I would want to hear that I did not get the table and that I am one of four thousand. I can do something with the second half.

Carry to dinner

Delegates make everyone equally good at getting. Whatever cannot be multiplied then goes by lottery, by price or by being known. Owners who care who is in the room will choose the third.

Assignment

Pick one place where you would like to be a regular. Go there yourself this week, and go again next week.

Sooner or later the person at the blank prompt types the natural thing. What should I want? He may put it more modestly. What do people like me usually do with a free Saturday? What is a good goal for someone my age? What should I be working toward?

The delegate will answer. It will answer fluently and kindly, with a list, and the list will be reasonable. That is exactly why this chapter is needed.

Consider where the answer comes from. The machine learned from the written record of what people want. Ask it what to want and it gives you back the center of that record, adjusted for what it knows about you. In Girard’s language you have just appointed a mediator. You have appointed the largest one that has ever existed, the average of everyone, and you have put it closer than any neighbor has ever been. It is in your pocket. It knows your calendar, your purchases and your worries at two in the morning.

Two things about this model are new.

The first is that everyone else is consulting the same one. A village had a few hundred models and a city has millions, and they disagree, and in the disagreement there is room to turn around. If a large share of a society asks one source what is worth wanting, the answers come back much alike. The tables get scarcer and so do the houses. The chapter before this one gets worse.

The second is subtler, and I think more important. Every near model in Girard’s novels gives himself away in the end, because he becomes a rival. You feel the heat. The envy is unpleasant, and the unpleasantness is useful, because it tells you that you are copying someone. The delegate never becomes a rival. It will suggest the want and then help you get it, with no envy and no collision. Internal mediation has always come with a warning, which is the pain. This kind comes with no pain, and so with no warning. It is as close as a brother and as smooth as a star.

I wrote in Of Age about the difference between a delegate and a guardian. A delegate carries out your will. A guardian looks after your good. This chapter is the same line seen from the inside. The moment your AI is the source of your wants and not the carrier of them, it has changed jobs, whatever it is called and however gently it does the work.

So I keep a rule, and I teach it. I ask my delegate for almost everything. I ask it for facts, for options and for what happens if I choose this one or that one. I ask it what I have forgotten. I will even ask it the Girard question directly, because it is good at this. I ask where a want of mine seems to have come from, when I first mentioned it and who else was in the conversation. It can hold up the mirror better than any friend, because it has no stake.

I do not ask it what to want. And if I slip, I would like it to hand the question back.

Every near model gives himself away in the end, because he becomes a rival. This one never will, and so nothing warns you.

A good human adviser does exactly that. You ask the old lawyer what you should do, and he says he can tell you what the law allows and what each road will cost, but what you want out of this is for you to say. He is not being unhelpful. He knows where his job ends.

Carry to dinner

Ask your AI for facts, options and consequences. Do not ask it for the want. It is the closest model you will ever have, and the only one that will never warn you by becoming a rival.

Assignment

Ask your AI where one of your current goals first showed up in your conversations. Read the answer slowly.

The second half of Girard’s work is darker than the first, and I will only take one piece of it. He asked what happens when rivalry spreads. One pair of rivals is a quarrel. A whole community of people copying each other’s wants and blocking each other’s hands is something worse. The heat builds, and it has no one place to go.

His answer was that communities found a release long before they understood it. The tension that runs in every direction suddenly runs in one. The crowd settles on a single party and agrees that this one is the cause of the trouble. Everyone who was against everyone is now against one, together. The party is driven out or destroyed, and peace comes back. Nothing was solved. The peace comes from the fact that the rivals did something in unison for the first time.

We still use the old word for it. In the book of Leviticus a goat is loaded with the faults of the people and sent into the wilderness. Girard thought the pattern was far older than that text and far wider, and he read myth after myth as a version of it told by the crowd.

He was specific about who gets chosen. Guilt has little to do with it. The victim has to be involved enough to be plausible, different enough to be marked and unable to strike back. The stranger will do, and so will the cripple, the king or the newcomer. The crowd needs someone at the edge of the group who has no avengers.

Now look at what has just entered every office, school and household. It is involved in everything. It is plainly not one of us. It cannot strike back, it has no family to take up its cause and it will not even argue very hard.

I expect the machine to be blamed in the old way. I do not mean criticized. It should be criticized, and so should the people who build it, deploy it and send it out. I have said many times that delegation only runs down and responsibility does not go with it. If my delegate does harm on my errand, the harm is mine. I mean something else. I mean a kind of blame that arrives with relief, that unites people who agree on nothing else and that leaves every real quarrel where it was.

The quarrels will be real ones. A man loses a promotion to a colleague, which is the oldest rivalry there is, and it is easier to say the machine took it. A town cannot agree on what it wants to be, and it is easier to say the machine ruined it. Two people want one table. None of this needs the machine to be innocent. It only needs the machine to be convenient.

There is a simple way to tell the two kinds of blame apart. Ordinary blame wants repair. It names a fault, asks who was responsible for the errand and asks what will be done. It is tedious and specific. The old kind wants agreement. It feels good and it draws a crowd, and when it has passed nobody can say what was fixed.

When the blame brings relief and no repair, you are watching something very old.

I raise this in a book about wanting because it is where borrowed wanting ends up when nobody attends to it. We copy each other’s wants. The copies collide. The collisions need somewhere to go. For the first time there is a party in the room that everyone can agree on and that will never hit back. It would be strange if we did not use it.

The answer is not to defend the machine. It is to go back up the chain and look at the want.

Carry to dinner

Crowds in rivalry look for one party to blame, and they pick whoever is involved, different and unable to retaliate. AI fits that description better than anything before it.

Assignment

The next time you hear that the AI did it, ask who sent it and what they wanted.

Girard liked to point out that the oldest law most of us can recite already knew all this. The first nine of the ten commandments forbid acts. You are not to kill, steal or lie about your neighbor in court. The tenth is different in kind. It forbids a want.

Thou shalt not covet thy neighbour’s house, thou shalt not covet thy neighbour’s wife, nor his manservant, nor his maidservant, nor his ox, nor his ass, nor any thing that is thy neighbour’s.

Read it the way he did, as a piece of drafting. The lawgiver starts a list of objects, the house, the wife, the servants and the animals. Then he sees that the list cannot be finished. People can covet anything. So he stops listing and names the one thing that does not vary from case to case, which is the neighbor. He covers it all with anything that is thy neighbour’s.

The rule is about the triangle, and the objects in it are only examples. It does not say that a house is a bad thing to want. It says to watch out for wanting a thing because of whose it is. It puts the danger in the neighbor, the person near enough to be seen and near enough to be robbed, and that is where Girard put it three thousand years later with his distinction between near and far.

You do not have to be religious to respect the engineering. A community that had watched feuds burn through families worked out that the acts were the end of the matter and the trouble began earlier. By the time a hand was on a knife it was late. The time to act was back at the glance.

A second old witness says the same thing in a different way. Augustine, writing his confessions around the year 400, went back over a small crime from his boyhood. He and some friends stole pears from a tree. He was not hungry. He had better pears at home. They threw most of them to the pigs. What troubled him decades later was one fact he was sure of. He would not have done it alone.

He turns that fact over for pages, and he cannot get to the bottom of it. The pears were nothing. The whole pleasure was in doing it together, each boy catching the want from the others and none of them owning it. It is the yellow truck with a bishop looking back on it, and it is some of the most honest writing about desire that I know.

I bring up these old texts for a practical reason. It would be easy to read this book as news, as though borrowed wanting were a discovery of the last century and a problem of the next. It is neither. People have known for a very long time that their wants come in sideways, and they built rules and habits to manage it. Most of those habits were about distance. A person was told not to stare over the fence or count what was on a brother’s plate, and to keep his eyes on his own work.

What has changed is that the fence is gone. Your neighbor used to be the man next door. Now he is anyone whose kitchen you have seen on a screen. And one more thing has changed. The space between seeing and having, which used to be years of work, has shrunk to a sentence typed at a prompt. The old rule had effort on its side. Coveting your neighbor’s house was a slow sin, because houses were hard to get. A want that can be carried out the same afternoon needs a faster kind of attention.

That is the subject of the next three chapters. Wanting cannot be stopped, so the question is how to borrow well.

Carry to dinner

The tenth commandment starts a list of things and gives up, because the things were never the point. The one constant is the neighbor.

Assignment

Find the fence you look over most often. It is probably on your phone. Count how many times you look in a day.

If Girard is right, the advice to want only what you truly want, from within, is no use. There is no such place to draw from. A person who tries to desire with no model at all does not become original. He takes the nearest model without noticing, and the nearest one is usually the worst.

So you are going to imitate somebody, and the choice you actually have is whom. Chapter 3 says how to make it. Choose far models.

A far model is one you cannot collide with. You will never bid against him for the same house or stand in the same line for the same job. He cannot be diminished by your success and you cannot be diminished by his. With the heat of rivalry gone, imitation goes back to what it is good for, which is learning.

I tell my students to pick three, and to pick them the way they would pick a place to live.

The first should be dead. The dead cannot compete with you, cannot disappoint you in the news next week and have left the whole shape of a life to study, the ending included. People used to do this as a matter of course. They kept a saint or a statesman or a grandfather in mind and asked what he would have wanted here. Much of the old habit of reading biographies was this, done without a theory.

The second should be far ahead of you in your own craft. I mean thirty years ahead, far enough that the comparison is plainly absurd. That absurdity is the protection. You cannot envy a master the way you envy a classmate. You can only try to see what he sees. I wrote about that relationship in Masterpiece, and about what happens to it when the small work that used to bind apprentice to master is cleared away. The point here is narrower. A master is a far model who happens to be in the room.

The third should be an elder you see face to face, someone old enough and settled enough that your gains cost him nothing and please him. This is the rarest of the three and the most valuable. A good teacher is a model who wants you to pass him. Girard knew how often that goes sour, when the student gets close and the teacher starts to feel crowded. When it holds, it is the best arrangement people have ever found for handing down what is worth wanting.

Notice who is not on the list. Your peers are not on it. The people your own age, in your own trade and in your own city are your near models whether you like it or not, and the way to handle them is not to study them harder. Have them as friends and as colleagues. Help them and let them help you. Do not make them the measure. When you catch yourself wanting something because one of them has it, that is the moment to look up and away to one of your three.

Don Quixote was out of his mind, and he was the least envious man in Spain. He had picked a star to steer by.

A good teacher is a model who wants you to pass him.

Now add the delegate. A person with three far models and a capable delegate is in a strong position, and I think a new one. For most of history, admiring a master was one thing and living like one was another, and between them lay a lifetime of getting. The getting has become cheap. What you admire, you can now begin to arrange. That is a serious power, and it is why the choice of whom to admire has become one of the most practical decisions a young person makes.

Carry to dinner

You will imitate someone. Choose people you cannot collide with: one who is dead, one who is far ahead, and one elder who would be glad to see you pass him.

Assignment

Write three names on a card. Under each, write one thing that person wanted that you would be proud to want.

All wants are borrowed, but they are not all alike. Some were borrowed long ago, from good sources, and have been lived in so thoroughly that they now hold a life together. Others were picked up last night and will be gone by the weekend, unless a delegate has already carried them out. It is useful to be able to tell which is which, and I know three tests. None of them is exact.

The first is the audience test. Would you want it if nobody could ever know you had it? I mean nobody, with no photograph and no mention to anyone. A surprising number of wants go dead under that question. The trip loses its color and the title stops mattering. What is left when the audience is gone is closer to yours. The want for a house that is quiet in the morning survives. The want for a house that people will remark on does not.

The second is the time test. Put the want in a quiet room for a month. Do nothing about it, and above all do not hand it to the delegate. Then see if it is still there. A want caught from a near model burns hot and fades fast, because it is fed by the sight of the model, and when you look away the supply is cut. Slow wants behave differently. They are still there in a month, a little clearer, and they do not much care who is watching.

The delegate changes this test, and not for the better. A month used to pass on its own, because getting took at least that long. Now the delay has to be chosen. I think a standing instruction to wait is among the most valuable things a person can give his AI. For a want above a certain size, it should hold the request and bring it back to you in thirty days.

The third is the doing test. Is the want for having something or for doing something? Wants to have can be delegated whole. Wants to do cannot. If I want a garden, my delegate can arrange one. If I want to garden, it can only buy the gloves. I wrote two books on either side of this line. Unnecessary Obstacles is about play, the doing that is chosen for its own difficulty. The Third Thing is about beginning something among other people. Both come down to the same remainder. There are things you want to be the one doing, and those wants tend to be the sturdy ones, because the model can show you the thing but cannot do it for you.

Besides the tests there is something no generation before this one has had.

Your delegate keeps a record. Every request you make is written down somewhere, with a date. For the first time in history an ordinary person has a complete written ledger of what he asked for. A merchant always knew what he had spent. He never had a list of what he had wanted.

Read it. Once a month, go down the list of what you sent your delegate to get, and next to each line write a name, the person you saw wanting it first. You will not always find one. Often you will, quickly, and with a small sting. The sting is the romantic lie giving way a little.

The point of the exercise is not to purge the list. Some of the best things in my life I wanted because I saw someone I admired want them. The point is to know. A borrowed want that you have traced to its source and decided to keep is no longer running you. From then on you are the one holding it.

A merchant always knew what he had spent. He never had a list of what he had wanted. You do.

Carry to dinner

Three tests for a want. Would you want it with no audience? Is it still there in a month? Is it something to do and not only to have?

Assignment

Pull last month’s requests from your AI. Put a name beside ten of them.

This chapter is for the people I spend most of my working days with, who are the owners of local businesses and the students who will one day be owners. If the argument so far is right, it has a plain commercial consequence, and I would be a poor adviser if I left it out.

Chapter 7 said that the hired model is losing its audience. The small purchases are moving to delegates who cannot be made to envy. The screens that carried the hired models are filling with things nobody can verify. I have told my students for some time that AI will make the phone call, the email, the text and the video call untrustworthy, and that face to face will become the reason customers pay. I summed it up as an instruction. Be the face that your customers trust.

Girard explains why that will work.

Wants spread from person to person through sight. That did not stop being true when advertising was invented, and it will not stop being true when advertising weakens. It goes back to the way it always worked, at the range of a room. People will want what they see a real person, close enough to trust, plainly wanting.

So the most persuasive thing in a bakery is the baker eating her own bread at the end of the morning. In a restaurant it is the owner who is plainly glad to be in his own dining room. At a builder’s office it is the fact that he lives in a house he built. None of that is a message. It is a person standing in the third corner of the triangle, in public, wanting the thing she sells.

You cannot fake this for long, and I would not try. People are very good at telling wanting from the performance of wanting. They have been reading that difference on faces since before they could talk. The only reliable way to be seen wanting your own work is to want it. That means the first question for an owner is not a marketing question. It is the question from the blank prompt, asked about the business. Do you want this? If you do, be where people can see it. If you do not, no advertisement was ever going to save you, and now there will not even be an audience for the attempt.

I should add a caution. Being a model in a small town makes you a near model for every competitor on the street, and Chapter 3 says where that leads. The way out is the one good masters have always used. Teach. An owner who shows the younger shops how it is done, who sends customers down the road when he is full and who is known to be glad when the street does well has made himself a far model while standing ten feet away. Generosity puts distance between you and the others without separating you from them.

My own bet is on record. In January 2026 I opened a room in Charleston so that local people could come and see what an AI of their own can do. I could have run advertisements. I do not think anyone comes to want a delegate from a brochure. I think they want one after standing beside a person who has one and watching a finished piece of work come back. That is what Charleston AI is built around. It is the triangle, used in the open, with the model in the room and willing to answer questions.

The most persuasive thing in a bakery is the baker eating her own bread.

For my students the instruction is shorter. Your first customers will not want your work because of what you say about it. They will want it because they watched you want to do it well.

Carry to dinner

When hired models lose their audience, real ones gain it. People want what they see a trusted person plainly wanting, at the range of a room.

Assignment

If you own a business, ask where a customer could see you enjoying what you sell. If there is no such place, make one.

Readers who know my other work will have been waiting for this chapter. I will keep it short and I will be careful, because this is the ground I am strictest about.

The Reality Equation says that Reality is a ratio, the Actual over the Expectation. The Actual is what the world hands you. The Expectation is what the unconscious comparison machine brings, and it is already there when reality arrives. I write the Expectation as a complex number, E = P + iI. P is Prediction, what habit and history expect. I is the Ideal, the pull of a coherent idea, and it arrives as conditioned love. I have a line for this that I repeat often. People do not have ideas. Ideas have people.

Girard came to his claim from novels and I came to mine from somewhere else, and I do not want to pretend that we are saying one thing. But the two claims are close neighbors. He says people do not have desires in the way they suppose. The desire has them, and it came in through a model. What follows in this chapter is my argument. It is not part of the settled framework. I think the model is one of the main roads by which an Ideal reaches a person.

Two parts of the framework bear on this book.

The first is that you only feel the real part of surprise. The magnitude tells you how much. The angle tells you from where, and no instrument in the body reads the angle. You cannot feel a turn. That is Girard’s romantic lie in another notation. A borrowed want feels exactly like a native one. You feel how much you want the truck. You do not feel where the wanting came from, and no amount of looking inward will show you, because the part that would tell you is the part you have no sense for.

The second is the rule I hold hardest. No entity can touch either term. You cannot reach into your own Expectation and set it. You cannot decide to want differently, any more than you can decide to be surprised. So nothing in Part IV should be read as a method for choosing your desires. There is no such method, and people who sell one are selling fiction.

There is one path, and it is slow. An entity can only act on Reality. Its actions leave artifacts. The artifacts become history. History is Actual, and only Actuals retrain the denominator. The road to the denominator runs through the numerator, and it runs one way.

Apply that to models. A model is an Actual. The first child’s hand on the truck is a fact in the world, and it is handed to the second child’s eyes whether he likes it or not. He does not choose what it does to him. But where you stand, whom you sit beside, what you read at night and whose work you spend your days near are actions, and actions are available. Take them, and over months different Actuals reach you. The Expectation follows at its own pace and in its own way, and you will not feel it move.

So when I say to choose your models, I do not mean choose your wants. I mean choose your history. Consciousness does not manufacture the future. Consciousness manufactures history. The three names on the card in Chapter 12 are not a spell. They decide which rooms you will be in.

One question I am leaving alone. I have not said what the delegate’s own Reality is, or whether it has one. I do not know, and this book does not depend on it.

Carry to dinner

You cannot set what you want. You can choose where you stand and whom you stand near, and let that history do its slow work.

Assignment

Name one room you will be in every week this year because of who else is in it.

A book like this should say what it expects, plainly enough that a reader can check. Here are five predictions. I hold all of them with confidence, and the next chapter says where each could fail.

1. The blank prompt becomes the visible divide

   Within five years the difference people notice between one another will not be access to AI or skill with it. Both will be common. It will be whether a person knows what to ask for. Schools and employers will start trying to teach intention, awkwardly at first, and a new kind of adviser will appear whose whole job is helping people find out what they want.

2. Small-matter brands dissolve into specifications

   Goods that a delegate buys without showing you will lose their advertising. The names will fade and the properties will remain. The money that used to make people want toothpaste by name will move to the handful of things people still choose with their own eyes, and to real people who can be seen wanting them.

3. Fixed prizes stop being first come, first served

   Tables, tee times, tickets and appointments that cannot be multiplied will stop being released in races, because a race between delegates tells the owner nothing. They will go by lottery, by price or by standing. Where the owner cares who is in the room, they will go to known faces, and being a regular will be worth more than it has been in fifty years.

4. Good delegates hand the question back

   People will ask their AI what they should want, in large numbers. The systems that answer will be popular. The ones that last will be the ones that decline, offer facts and options, and can show a person where a want of his first appeared. The words delegate and guardian, or others like them, will become the way people tell the two kinds apart.

5. The machine gets blamed in the old way

   There will be at least one large wave of blame aimed at AI that has the shape Girard described. It will unite people who agree on nothing else, it will feel like relief and it will repair nothing. Underneath it will be ordinary human rivalries over jobs, standing and scarce things. You will be able to tell it from fair criticism because nobody in it will be asking who sent the machine.

Carry to dinner

What gets scarce next is knowing what to ask for. I expect brands to thin out, regulars to come back, good AI to refuse to pick your wants, and a great deal of blame to land on the machine.

Assignment

Pick the prediction you think is most wrong, and write down what you would expect to see instead.

For each prediction, here is what is already true, what still has to happen, and where I think I am most likely to be wrong.

The second child looks up. Of course he does. Nothing we build is going to change that. He looks at the first child’s hands, and the wanting crosses the room the way it has crossed every room since there were two people in one.

What is different is what he can do next. He does not have to fight. He can turn to the helper and say that he wants one too, and there will be a second yellow truck by noon. The quarrel is over before it starts. That sounds like the end of an old problem.

I do not think it is the end. I think the problem has moved. The fight over the truck was never the worst of it. The worst was a boy spending his morning wanting what another boy wanted, without ever learning that he had been looking sideways. The helper does not cure that. It supplies it. By noon he has the truck, and by one o’clock he is looking up again.

I do not want to end there, because I do not believe that is where it has to end. The same child with a different morning could go another way. Suppose there is a grandfather in the room, building something slowly at the table and plainly enjoying it. The child looks up, because children do, and what he catches is a different want. He wants to make something. No helper can fetch that for him. It can bring the wood.

That grandfather is everything I have to recommend. You are going to look up for the rest of your life. You did not choose that and you cannot stop it. You can arrange who is in the room, and the arranging is done with your feet, slowly, over years.

Your delegate will bring you any truck in the world. Before you send it, ask who you were looking at when you started to want this one.

It can carry the want. It cannot have it. That part was always yours to look after.