August 23, 2026
Worthy Difficulty
A psychologist on the one thing schools must protect in the age of AI, because everything else is being automated.
A child starting kindergarten this fall will enter adult work sometime in the late 2030s. Nobody can tell you what that work will be. Not the futurists, not the economists, not the people building the machines. For the first time since we invented the modern school, the adults genuinely do not know what world they are preparing children for.
We do know the change has already started. At a BMW plant in South Carolina, humanoid robots worked ten-hour shifts across an eleven-month pilot, moved more than ninety thousand parts, and helped build more than thirty thousand vehicles. BMW is now expanding what it calls physical AI into its American and German factories. Early days, heavy safety constraints, more pilot than takeover. But a threshold has been crossed. The five-year-old walking into her first classroom and the robot walking onto the factory floor are arriving in the same world.
One of them will now spend thirteen years in a system built for a different one.
Earlier this summer I asked what college is for when the machine can do it. The question has not waited for an answer. It has walked down the hall to kindergarten.
What the old school was for
The thing that is dying deserves a fair hearing, because it was an achievement.
Before the Industrial Revolution, most children learned what their parents knew. The factory changed that, and so did the city, the office, and the profession. Mass society needed mass literacy, numeracy, punctuality, and the ability to sit in rows and follow a schedule. So we built the school we all recognize: subjects in sequence, children sorted by age, standardized tests, a credential at the end that led to an occupation. It worked. That system opened books, professions, and whole futures to people who had been locked out of them for centuries. My own life runs through it. Probably yours does too.
It was calibrated to something specific. The industrial school prepared people to take a place inside large institutions, and its core product was transmissible knowledge: facts, procedures, formulas, the kind of knowing that fits in a textbook and can be checked on a test.
That is exactly the kind of knowing machines now do better than we do.
And the consequences have stopped being theoretical. Ask this year’s college graduates, who did everything the industrial school told them to do and are sending out hundreds of applications to reach one interview. The entry-level rungs their degrees were built to reach, the first drafts, the research memos, the routine code, the junior analysis, are exactly the work AI eats first, because that is the work we standardized most. Economists at Stanford tracking payroll data can already see it: the youngest workers in the jobs AI reached first are falling behind their peers. Reached first, not exposed, as though the rest of us were safe. Every job is on the same road. Those graduates are simply at the front of the line. Meanwhile the people pulling ahead are not the ones holding the most credentials. They are the ones, degreed or not, who picked up the machine and learned to work it.
I will say plainly what I think is coming, because I would rather be wrong in public than vague. Some morning within the next five years, we are going to wake up and notice that AI can do most of what we do better than we can, and that it has hands now, robotic bodies moving through warehouses and kitchens and clinics. Not every job will vanish that morning. But the question every parent is already whispering at the school gate, can my child compete, will have been answered, and the answer will be no.
Which is exactly why competition is the wrong frame, and this is the hinge of the whole thing. A child cannot outrun a tractor either, and nobody calls that a crisis of childhood. The right question is not whether AI can do it better. The right question is what doing it does to the child.
Two kinds of knowing, and the one that cannot be downloaded
Ask what school actually gives a person and you find two different things tangled together.
The first is the kind of knowing you can look up. Dates, definitions, procedures, the quadratic formula. Schools became very good at delivering it, because the industrial world needed it and a test could measure it. That kind of knowing now lives in the machine, retrieved and explained on demand, patiently, in any language, at any hour. There is no future in being a warehouse of facts. That warehouse has been built, and it is not us.
The second kind cannot be looked up, because it is not information at all. It grows in a person, through trying, failing, correcting, enduring, working beside others, and carrying responsibility for something real. Call it judgment. It is the difference between the student who can recite the formula and the one who knows which formula the mess in front of him actually calls for. Between knowing about courage and being brave. Machines now hold the first kind of knowing for us, which leaves education one irreplaceable job. The second kind. And the second kind cannot be transmitted at all. It can only be grown, and it grows through difficulty.
The rabbi, the philosopher, and the bronze giant
We have known this longer than we have forgotten it. Before education became a delivery system, its two deepest models in the West were the rabbi and the Greek teacher, and neither one was in the information business.
The rabbi did not hand his students a scroll and a rubric. A disciple lived beside his teacher. He walked with him, watched how he prayed, how he settled a dispute, how he treated the poor, and argued the text with him until the arguing became part of him. The goal was never to know what the rabbi knew. It was to become what the rabbi was. And the Greek teacher did not lecture either. Socrates stood in the marketplace asking questions that took a young man’s confident answers apart in public, not to humiliate him, but because the examined life has to be examined by someone. Philosophy was not a subject. It was the training of the instrument itself, the discipline of telling a true thought from a merely plausible one.
Here is the detail that should stop us. The same Greeks who invented that education were the first people on record to dream of intelligent machines. Hephaestus forges golden handmaidens who speak and think. Talos, a giant of bronze, circles the shores of Crete three times a day, guarding the island, the first robot sentry in literature. The word automaton is theirs.
And Pandora. We remember the box and forget what carried it. Pandora was a manufactured woman, fashioned to order and lovely on the outside, and Hesiod says it was Hermes who filled her with lies and crafty words. Hermes, the god of language. The first artificial companion in our literature was beautiful, was sent into a man’s home as a gift, and had nothing inside her but speech. The gift looked like company. It came bearing a jar that could not be shut once it was opened. Twenty-seven centuries before the first AI companion, the Greeks had written that story to its ending.
The civilization that first imagined the artificial mind is the same one that insisted every young mind be trained in philosophy, and I do not think that is a coincidence. They dreamed the machines, they dreamed the seduction, and they prescribed the remedy in the same breath.
We kept their dream and dropped their remedy. We have built the automatons, and we teach almost no philosophy to anyone under twenty. The one discipline whose entire subject is the difference between truth and plausibility has been retired from our schools at the exact moment we surrounded our children with machines that generate plausibility on demand, fluently, at no cost. That is the vacancy at the center of the curriculum, and no app fills it. It was always filled by a person, a teacher whose questions were harder than the machine’s answers.
The Trend
The most important study I have read on AI in schools makes the ancient point with modern instruments. Researchers gave nearly a thousand high-school math students an AI assistant to help with their practice. While the assistant was on, they performed 48 percent better. It looked like a breakthrough. Then came the exam, with the assistant off, and the same students scored 17 percent worse than classmates who had never used it at all. The machine had made the homework better and the students weaker, because it was doing the very struggling that learning is made of. The researchers reduced it to one line. AI can improve performance while weakening learning.
It did not have to go that way, and the same study proved it. When the teachers rebuilt the tool to give hints instead of answers, the harm mostly vanished. And in Nigeria, a six-week trial in ordinary schools put the machine inside a teacher-guided program, with prompts built to provoke reasoning rather than hand over conclusions, and students gained more than many interventions achieve in years. So there are not two options, AI or no AI. There is a third, and it is the only one worth building: the machine designed to equip rather than replace, and the student taught to engage rather than lean back. That is where the best total quality lives: better work, and a stronger person doing it.
Same technology. Opposite outcomes. The difference is not access to AI. The difference is whether the machine replaces the struggle or dignifies it.
There is the real divide. Outsourcing hands the machine a task and loses the formation the task was carrying. Augmentation hands the machine what does not form the child, so more of the day can go to what does. The clinical name for what is at stake is agency: the settled inner sense that says I acted, I struggled, and I can do this. One kind of help builds it. The other replaces it and inflates the grade at the same time, so the child produces better work while becoming less convinced he can work at all.
You cannot tell the two apart by looking at the tool. You can only tell by looking at the child afterward. Stronger, or just faster?
The school I would actually want
Hold that test in your hand and a different school comes into focus. This is the one I would want for my own children, who are sitting in these grades right now.
In grade school, almost nothing changes, and that is the point. Children still learn to read, write, count, memorize poems, take turns, and care for a room they share. Those foundations get protected on purpose, the way handwriting now survives only where somebody decided it mattered. What changes is where the teacher’s hours go. The machine generates practice at each child’s level, catches the recurring error, translates the note home into a family’s own language, and eats the administrative work that devours a teacher’s evenings. The teacher, who came to this work to be with children, gets to be with children. That is the whole gain: more music, more dirt under fingernails, more argument and forgiveness on the playground, more adults who know each child by name.
And the screen can change teams. The same device that sedates a child can become a workbench. A group of nine-year-olds describes a game, generates a first playable version, draws the characters, tests it on their classmates, discovers it is boring, and rebuilds it. That is narrative, logic, mathematics, collaboration, and the hard art of receiving feedback, all in one afternoon. Children who have spent hundreds of hours inside worlds engineered to hold them still would learn that worlds are made, and that they can make one.
High school is where the furniture really moves. Seven disconnected periods give way to a studio, a seminar, a lab, and an apprenticeship stitched together. Core mathematics with a tutor that never loses patience and never lets you skip the struggle. The Brothers Karamazov argued around a table, because the machine can summarize the novel but it cannot be changed by it. Two afternoons a week inside a hospital, a machine shop, a parish, somewhere the work is real and someone is counting on you. The robots become part of the curriculum, the new material of work: teenagers training them, inspecting their output, deciding which tasks should stay human and having to defend the answer out loud.
Assessment follows the work, because the essay and the problem set are no longer proof of anything. Defend your project out loud. Solve an unfamiliar problem in front of us. Teach it to a younger student. Show us what happened when your design met reality and lost. The question stops being did you submit the assignment and becomes what can you make, repair, explain, and take responsibility for.
The trades may be where the new arrangement is clearest first. The robot takes more of the lifting, the repetition, the danger. The human electrician diagnoses, improvises, directs the machines, and makes the safety call no training video anticipated. An apprentice learns beside three teachers: a human master, an AI that has read every manual ever written, and a robot performing the task at half speed so the hands can follow. That is not a classroom anymore. It is a guild with better tools, and we had guilds long before we had grade levels.
Passengers, pets, and parasites
The Greeks began the dream of the automatons, and science fiction has been finishing it ever since.
The darkest and most honest version comes from Iain M. Banks, whose Culture novels imagine the endpoint: a civilization run by superintelligent Minds, where humans want for nothing and are needed for nothing. Banks once described the humans of that world, from the Minds’ point of view, as somewhere between passengers, pets and parasites. Then he added the clause everyone forgets: the humans give the Minds a reason to live and something to do.
That is the outsourcing road taken all the way to the end. Every difficulty smoothed, every task lifted, every struggle optional. It sounds like paradise and it describes a kennel. And yet even Banks could not make the picture close: his godlike machines still needed the humans to matter. The question he leaves on the table is the one our schools will have to answer within a generation: what is a human being for, when usefulness is no longer our strongest claim?
The Christian tradition answered that a long time ago, and the answer is older than the economy that taught us to forget it. Work came before the fall. Before anything has gone wrong, the man is put in the garden to till it and keep it (Gen 2:15), which means labor was never the price of our survival. It was part of what we were made for. Sweat and thorns came later. John Paul II put the whole thing in one clause in Laborem Exercens: through work a person does not only transform the world, he becomes “more a human being”. And the Second Vatican Council taught that education itself is the formation of the human person toward his own end and the good of everyone his life will touch. No employer wrote that sentence. So a world that needs less of our labor does not make formative work less necessary. It makes the choosing of it the whole task. When difficulty becomes optional, the curriculum becomes a moral document: it is the list of difficulties we have decided a child’s soul is worth.
I am not writing this from above it. I am a psychologist who works beside these machines all day, and almost nothing I produced this week was untouched by AI. What to hand over and what to keep is not a policy question for me. It is a decision I make at my desk every morning, and I do not always make it well. I know the pull of the outsourcing road from the inside. It is smooth the whole way down.
So the death of industrial education is real, but it is not a tragedy, because the industrial school was never the whole of education. It was one arrangement, built for one world, and that world is ending. What comes next will have astonishing tools, and its task will be ancient: to give the young difficulties worthy of them. To hand the machines everything that does not form a child, precisely so the child can spend her one childhood on what does.
We are not preparing children to compete with machines. We are preparing them to inherit responsibility for a civilization that will contain them.
That was always the job. We are just no longer allowed to forget it.
If this gave words to something you have been carrying, share it with someone who needs it.
The Inner Exodus is where I think out loud about staying human in the age of AI: the interior life, the attention we are losing, and the God who meets us beneath the noise. The essays are free, and they stay free. If the work is worth something to you, becoming a patron is what keeps it free for everyone who cannot pay.