The Future of Education

A dated document is worth more than a confident opinion.

Anyone can explain a change once it has happened. In the early part of the 2010s TEN3 published two short papers saying what was coming, and they are reproduced here in full, at the bottom of this page, so you can judge them yourself rather than take our word for it.

They are not impressive documents. They are four pages and three pages, written plainly for school administrators, with hand-drawn diagrams. That is rather the point: the argument did not need to be clever, it needed to be right about the direction.

Some of it was right, one part of it has been overtaken, and one call has gone the other way so far. All three are set out below, scored in a table. The part we got wrong turns out to be the most useful thing on this page.

The industrial model is not working.

A school year, a class of thirty, a fixed syllabus delivered at one pace, and an artefact collected at the end to prove it arrived. That design was not handed down from antiquity. It was built to educate large numbers of people affordably, and it did.

It was already straining before anyone had heard of a large language model. Students arrive at university unable to read a complex argument and carry it; employers say the same thing about graduates every year; and everybody's proposed remedy is another layer of measurement laid over the same machine.

Everything in these two papers follows from one observation: when the scarce thing changes, the institution built around the old scarce thing stops making sense.

Universities exist because knowledge was rare and expensive to reach. You went to where the books and the scholars were, because there was no other way. The second paper says this outright — "the rare resource then was knowledge rather than equipment" — and it was written before knowledge became free to anyone with a connection. That is the whole argument, and artificial intelligence has only finished the job the Internet started.

The credential comes loose from the school.

The first paper's claim was that education was moving to a test-based system — that what would matter is the exam a person can pass, not the institution that walked them to it.

It began by drawing the arrangement everyone knew. A student climbs: primary school, then an exam, then secondary, then another exam, then university, and only then the professional certification. Every step is gated by the step below it, and the school controls the gate. The paper noted one exception already visible at the time — a technical certification like the A+ could be taken by anyone who could pass it, straight from primary school, with no institution's permission at all.

Then it named the thing that would break the arrangement, in a sentence that reads very differently now than it did then: "The Internet is opening up the opportunity for people to study on their own and pass exams."

If that is true, the paper argued, the institutions stop being the road and become optional stops beside it. A student would not need a secondary or tertiary diploma to sit the exam that actually gets them work. It gave a concrete example: the California application to become an Investigative Certified Public Accountant asked for the "equivalent to graduation from college" — not graduation itself. And it pointed out that in living memory, medicine and law were both entered by examination and apprenticeship rather than by degree.

From that one shift the paper drew the consequence that mattered to a school, and it is the sentence worth carrying away:

"We know that for a school to succeed in the future, they are going to have to show 'value-added' to the prospective students, or they will not have students."

Its third diagram put the two pictures together, and this is where the paper is at its sharpest. The tables turn. Students stop struggling to pass a test in order to be let into the next institution; institutions start working to attract students who no longer need them. Both routes exist side by side, and the school's job changes from gatekeeping to being demonstrably worth attending. The paper also listed, almost in passing, the two things it expected schools to need: portfolios of the student's actual work, and exams arranged in levels. One of those has since been overtaken, and it is the subject of the scorecard below.

Then the second paper said what a school should do.

School Response refused the obvious answer. The obvious answer in the 2010s was to chase science and technology, and to buy back the timetable hours from art, history and reading.

The paper called that more damaging than the Internet, on the grounds that the mind runs on language and thinking is tied to stories — so a curriculum that trades away language to make room for technique produces students who cannot sustain a long thought about anything, technical subjects included.

Its recommendation for primary school was the opposite: build language and reasoning first, integrate science into exploring the world, and do not begin formal mathematics until around grade five, citing Bénézet's experiments in which arithmetic taught later was learned faster and understood better. Stated plainly, its aim for a primary school was that "the student should learn to read, reason and have a relationship with Christ, their Creator."

And for secondary school it put one thing underneath everything else. A school may specialise — pick which exams it prepares students for, and treat that as its niche in the market. But nothing it offers stands up without the layer beneath.

And the second paper named the failure precisely, in a sentence that has aged better than anything else in either document: "Students learn to read and can tell you what they read, but they do not know what it means. They have no idea how to apply information usefully."

What we called, and what became of it.

Six claims, scored. Two held, one is partly there, one turned into a programme we now publish, one has been overtaken by events — and one has gone both ways at once.

The overtaken row is the most valuable line in the table. TEN3 named the problem in the 2010s, proposed the portfolio as the answer, and the same organisation now says plainly that the portfolio is no longer sufficient. The evidence has had to move off the artefact and onto the student — which is what the fifth entry of Build a Wisdom Centre is about. We would rather be corrected than consistent.

The tests split the sector in two.

A footnote added to Educational Direction around 2020 reads: "the SAT has not morphed and universities are beginning to drop the SAT as a requirement." Since then the picture has divided.

Most of the sector did move away. Hundreds of institutions went test-optional, and in 2021 the University of California — the largest system of its kind in the United States — went further and stopped considering the scores at all. Then a small group of highly selective private universities went the other way and reinstated their requirements: the Massachusetts Institute of Technology in 2022, and Dartmouth, Yale, Brown, Harvard and the California Institute of Technology in 2024.

So the score is split, not reversed. The reinstatements are the visible half because of who made them rather than how many.

A reinstatement shows that a university found the score useful for sorting the applicants in front of it. Whether the thing the score measures is still the thing that matters is a separate question — and it is the one these papers were asking.

The distinction is between logic and reason, and it is worth naming because it recurs everywhere in this argument. Logic is the machinery for getting validly from one statement to the next; it never breaks. Reason is that machinery plus the commitment to check whether what you fed into it is true. An argument can be perfectly logical and still absurd — all spiders have eight legs; my dog is a spider; therefore my dog has eight legs — because the wiring was flawless and nobody inspected what it was plugged into.

Reinstating an entrance test is a valid inference from the premise that scoring high measures the scarce and valuable thing in a person. That premise held for a long time. It is precisely the premise artificial intelligence has taken away: the machine outscores every student any of us will ever teach, at recall, at speed, at pattern-matching, and it does so for a fraction of a penny. What the test measures is exactly the capacity that has become cheap.

An institution can run a flawless argument for years on a premise it has stopped going upstairs to check.

Those universities may well be right about the applicants in front of them this year. The longer question is a different one, and our answer to it is an argument rather than a fact: a test of the cheap thing is unlikely to stay at the centre of admissions once everyone can buy the cheap thing. We said the SAT would have to change what it is or be dropped. We were early, and part of the sector has gone the other way.

Grading by listening is the older thing.

Examination by conversation is what assessment was before we industrialised it.

Socrates did not collect submissions. Aristotle's students were not marked on documents. For most of the history of education, finding out what someone understood meant asking them and listening to the answer, and the written artefact submitted for marking is the recent innovation — adopted because it scales, at exactly the moment schooling needed to scale.

The machine has not driven us somewhere strange. It has removed the reason we ever left. We accepted the artefact as a proxy for the person because asking each person was expensive. The proxy has failed, and the original method is waiting where we left it.

It is still expensive. Two things pay for it. The first is that the hour spent delivering knowledge is an hour that no longer has to be spent that way, which is where the time comes from in Build Your Own AI Policy. The second is structural: put students in a room doing real work under supervision all week, and the conversation costs nothing extra, because the person who would conduct it is already sitting there. That is Build a Wisdom Centre, and the two papers on this page are why it exists.

The two papers, as written.

Both were produced in the early part of the 2010 decade and circulated to school administrators. They are reproduced here in full. Three obvious typographical slips have been corrected and nothing else has been changed; the diagrams above are new renderings of the originals, drawn so they can be read on a phone.

The two papers ended without a method. These do not.

Both papers close by telling a school to develop a philosophy and a strategic plan. That was the right advice and it was not much help, because neither paper said how.

What has been written since is the how, and each piece answers a question these papers left open:

Build Your Own AI Policy — eight decisions a school has to settle about the machine, written to faculty rather than to students, each with wording ready to be used. It is where the question "what counts as the student's own work now?" gets an answer a school can actually apply on Monday.

Build a Wisdom Centre — nine decisions for a department that wants students in real work under supervision, assessed by explaining what they did. It is the "value-added" of the first paper, made concrete, and its fifth entry is where the portfolio is replaced.

TEN3 Primary Maths — the second paper's recommendation about language, reasoning and delayed arithmetic, built into a curriculum with practice activities. Free, and it works with the network off.