Somewhere in most houses there is a frame with a document in it. It has a seal, a signature, a name spelled correctly at last, and a phrase in a dead or ceremonial language certifying that the bearer has completed a course of study. Nobody looks at it. It was not made to be looked at. It was made to be shown once, to an employer or a relative, and then to hang as proof that something happened during those years, something the frame calls education.

Ask the owner what happened, and the answer is rarely about ideas. It is about friends, a few teachers, exams survived, the long routine of getting up and going somewhere because everyone else did. Some of it was learning. Much of it was something else. The frame does not distinguish, and neither, for most of the last two centuries, did the institution that issued it.

I. Instruction Wearing a Better Name

The two words were once kept apart. Instruere, in Latin, means to build in, to furnish, to equip: you instruct a soldier, an apprentice, a clerk. Educere means to lead out, to draw something from a person that was not put there by the teacher. The first is transmission. The second is formation. Both are honorable, and they are not the same activity.

The French state knew the difference well enough to change its sign. For most of the nineteenth century the ministry in charge of schools was the Ministère de l’Instruction publique. In 1932 it became the Ministère de l’Éducation nationale. The buildings did not change, nor did the timetables, the dictées, or the examinations. The word did. Spain made the same move a few years later, and much of the world followed. It is hard to find a better emblem of the substitution: an institution built to instruct, renamed for the larger thing it hoped it was doing.

Mass schooling had good reasons to instruct. The Prussian system, and Horace Mann’s common schools in Massachusetts, set out to make citizens literate, numerate, and governable, and to give industrial economies workers who could read a notice and keep a schedule. Michel Foucault, in Discipline and Punish, would later list the school beside the barracks and the factory as an institution of timetables and examinations. None of this was a conspiracy. Societies needed to transmit skills, coordinate access to work, and give children somewhere safe to be. The trouble began when those functions borrowed the name of something they did not guarantee. John Dewey, in Democracy and Education (1916), complained that schools treated education as preparation for a later life instead of a life being lived. He was describing a gap the name had already papered over. Ivan Illich, in Deschooling Society (1971), was blunter: the pupil is schooled to confuse teaching with learning, promotion with education, a diploma with competence.

II. Leisure, Which Is to Say School

The word school carries its own small confession. It comes from the Greek scholē, which meant leisure: the free time of a person who did not have to work. Aristotle took for granted that inquiry belonged to people whose bread was made by others. The original school was not a building but a condition, and the condition was distributed by class.

That condition never became universal, even when attendance did. Virginia Woolf put a price on it in A Room of One’s Own: five hundred pounds a year and a door that locks. Time, security, books, and a household that did not need your wages were the real prerequisites for a life of sustained learning. The school could open its doors to everyone and still leave those prerequisites where they were. Pierre Bourdieu spent a career showing how it then graded children on what their homes had already given them and called the result merit.

It would be easy to conclude from this that the appetite for learning belongs to a minority. The historical record refuses that conclusion. Jonathan Rose’s The Intellectual Life of the British Working Classes documents miners pooling their pennies to stock libraries in the Welsh valleys, weavers reading Carlyle at the loom, mutual improvement societies meeting after twelve-hour shifts. The appetite was common. What was scarce was the time to feed it, and the people who fed it anyway did so mostly outside the school, often against it. The minority was never a minority of curious people. It was a minority of people with the conditions to stay curious.

III. The Credential Stands In

Once instruction had borrowed the name of education, a second substitution followed. The certificate began to stand in for the instruction. In 1973 the economist Michael Spence formalized the idea of job-market signaling: a degree can be valuable to an employer even if it teaches nothing, because it proves the holder could endure the process of getting it. Randall Collins, in The Credential Society (1979), argued that much of the expansion of schooling was an arms race of certificates, each generation needing more years to signal what fewer years once did. Bryan Caplan pushed the claim further, and probably too far, in The Case Against Education. The milder version is enough. A large part of what schooling sells is not knowledge but a sorting mechanism, and the student is rational to buy it.

The social psychologist Donald Campbell warned in 1976 that the more a quantitative indicator is used for decisions, the more it distorts the process it was meant to monitor. Campbell’s law was written with standardized testing in mind. Teaching to the test is the visible symptom. The quieter one is what grades do to appetite. In a well-known 1973 experiment, Mark Lepper and his colleagues promised preschoolers who already loved drawing a certificate for drawing; afterwards, those children drew less on their own than children who had never been promised anything. Psychologists call it the overjustification effect. An institution that pays for learning in grades risks teaching a person that learning is something one does for grades, and stops doing once they end.

IV. What the Machine Exposes

For a long time this arrangement was stable because the instruction underneath the credential was scarce. Somebody had to know how to reconcile the ledger, draft the contract, write the report, translate the letter. The certificate was a promissory note on those operations, and the operations kept their value.

Economists David Autor, Frank Levy, and Richard Murnane described in 2003 how computers absorb routine tasks: anything that can be specified as a procedure. Generative models have widened what counts as specifiable. A 2023 study by Tyna Eloundou and colleagues, GPTs are GPTs, found that most occupations contain some tasks exposed to language models. That is not the same as whole occupations disappearing, and the distinction matters. But it changes what the frame on the wall is worth. If the instruction can be decomposed and performed by a mechanism, the credential that certified the instruction loses its promise, while the word education stays printed on it, intact.

The tempting reply is to say that education is whatever machines cannot do. Judgment, creativity, empathy: the residue. That answer should be refused. Chess was once the emblem of the human intellect, until 1997. Go was the game a machine would not master for decades, until 2016. Defining education, or human worth, as what lies beyond the current frontier of automation makes both depend on a boundary that engineers move every year. It also concedes the institution’s premise, that education is a set of performances, and merely bets on which ones will stay scarce.

V. The Library Opens

Then came the language models, and after them the agents. For most of history, instruction was rationed by the supply of people able to give it. In 1984 the educational psychologist Benjamin Bloom reported that students tutored one-on-one performed about two standard deviations above students taught in a conventional classroom, and named the two sigma problem: how to get that result without a tutor for every child, which no society could afford. A language model is the first serious candidate answer. It explains the same procedure ten different ways at three in the morning, in the student’s own language, without impatience, and it gets better every year. An agent goes one step further and performs the procedure itself. The impact on instruction is not a forecast. It is already in every classroom, mostly through the students’ phones, and it will keep improving faster than the institutions around it. It may distribute instruction more evenly than any ministry ever managed. It does not distribute the leisure, the security, or the appetite.

Borges imagined something like this in 1941. “The Library of Babel” describes a universe made of hexagonal galleries, adjacent and endlessly connected, that holds every possible book, including “the faithful catalogues of the Library, thousands and thousands of false catalogues, the demonstration of the fallacy of those catalogues, the demonstration of the fallacy of the true catalogue.” When it was announced that the Library contained all books, the librarians’ first reaction was extravagant happiness: there was no personal or worldly problem whose eloquent solution did not exist in some hexagon. As Borges notes, that hope was followed by an equally excessive depression. Everything was available, and nothing could be found, or trusted once it was found. Some went in search of the Man of the Book, a librarian said to have read the catalogue of catalogues and to be, for that reason, like a god.

A language model is not the Library. The Library holds every possible string of letters, nearly all of them gibberish. A model is closer to the fabled librarian, the one who answers anything, fluently, from a shelf you will never see. The old hope returns with it, and so does the old risk: the faithful catalogue and the false one arrive in the same calm voice. When instruction becomes infinite and connected, it stops being scarce, and so it stops being what a school can sell. What stays scarce is what Borges’s librarians lacked: knowing which book to look for, noticing when the catalogue is lying, and having a reason to keep reading. None of that is instruction. It is the other thing the word was supposed to name.

VI. What the Word Was For

A different answer is available, and it is older than the machines. Education, if the word means anything beyond instruction, names the formation of a person able to judge what is worth doing, to notice what a procedure leaves out, and to keep asking questions after there is no certificate to earn. It is closer to an orientation than to a skill. Its evidence is not what someone can do on the day of the exam but what they still do, unprompted, twenty years later.

That definition does not belong to the university. A cabinetmaker who keeps reading about wood after the apprenticeship ends, a nurse who wants to know why a protocol exists and not only how to follow it, a retired mechanic working through a history of the Roman Republic at the kitchen table: these are educated people in the only sense that survives automation. A doctorate earned for the title and abandoned at the defense is not. Vocational learning, practical learning, and self-directed learning are not lesser forms. They are often where the appetite was best protected, because nobody was grading it.

Latin America built two institutions that were more honest about the name than most ministries. In January 1942, Getúlio Vargas’s government created Brazil’s SENAI, the Serviço Nacional de Aprendizagem Industrial, financed by a levy on industrial payrolls and run by the industrialists themselves; its model spread across the region. In June 1957, a military junta in Colombia created the SENA, the Servicio Nacional de Aprendizaje, on an idea of Rodolfo Martínez Tono, funded the same way and free to its students. Neither called itself a ministry of education. Their names say aprendizagem, aprendizaje: learning, apprenticeship. They promised a trade and delivered one, to millions of people the universities never counted, and a lathe operator or an electrician trained there owed nobody an apology for not holding a degree. That honesty is also what now puts them first in line. An institution built to transmit specifiable skills stands directly in the path of machines that perform them. What it cannot hand over to the machine is the electrician who keeps asking why.

The lie was never that schools instructed. Instruction is necessary and dignified. The lie, and it was mostly unintended, was the promise attached to it: that the institution delivered the larger thing, that everyone who passed through had received it, and that anyone who failed to become a lifelong learner had personally wasted the gift. The conditions for that kind of learning were distributed as unevenly as the leisure in scholē, and the blame was distributed to individuals. Automation has not made education obsolete. It has made the old promise legible, by taking away the part that was easiest to certify.

If nobody would ever again check what you know, what would you still want to learn?

Further reading