The reskilling problem: how fast can a country learn something new?

Suppose a country decides, today, that it must prepare its people for AI. Not eventually — now. How fast can that actually happen?

Under the way we do things today, the honest answer is: painfully, dangerously slowly. And the reason why is a chain with too many links, each one measured in years.

To teach AI in schools and colleges, you first need people who both understand it deeply and can teach it. There are very few of them, and finding or growing them takes years — before a single student has learned anything at all.

Then the chain runs, link by link, in sequence: experts train teachers → teachers become competent → teachers finally teach students → students slowly graduate with the skill. Each link takes years, and because they run in order, not in parallel, the total lag from national decision to a genuinely skilled cohort is easily a decade.

Notice where the jam is. It isn't a shortage of students wanting to learn AI — there are plenty. It's the shortage of people qualified to teach it. A country cannot reskill its population faster than it can manufacture the trainers, and manufacturing trainers is the slowest step there is.

It is worth grounding this in something that already exists. The engine at the centre of this proposal is not hypothetical for one whole domain: it assesses language today, from A1 to C1, giving the thousandth learner the same criteria-based judgement as the first — a score on each criterion, every error corrected in the learner's own text, the exact next step named. Which is to say, for the integration backlog the technology to clear it is already here. What is missing is not the capability but the structural permission to deploy it — the decision to let a proven grader carry the load a scarce human currently rations.

Worst of all, in a fast-moving field the chain guarantees you arrive late. By the time the teacher-training pipeline has finally produced enough AI teachers, AI has moved on. Students end up learning the version of the field that existed when the pipeline started — not the one waiting for them when they graduate. You are perpetually teaching yesterday.

This isn't only about AI. Every time the economy needs a new skill — a new technology, a new trade, a new regulation — the entire slow chain starts over from the beginning. Which means a country is structurally slow to adapt to anything new, forever a decade behind whatever the world just changed to.

Now do it the other way. One genuine expert — the best there is — records the lessons once. From that moment, every school and college can teach AI from day one, at the level of that expert, without first growing an army of AI teachers. The teacher-training bottleneck simply doesn't exist, because there is no teacher to train — there is a recording to press play on, and an observer in the room to keep the learners moving.

And when the field moves, the expert re-records the parts that changed, and every classroom is current the next day. No decade-long catch-up. The system sits at the frontier by default.

This works for colleges and adult retraining as much as for schools — anything after age eleven where "correct" can be taught. The one honest exception is the research frontier itself: PhD-level, original inquiry, where the knowledge is still being made and there is no settled best explanation to record. Everything short of that frontier, the model reaches.

And because the engine never closes, an adult whose whole field is disrupted can be retrained by the same system, immediately, repeatably, for life — not left waiting months for a course that may never be staffed.

Lay the two side by side, on the three axes that matter.

Time. The old way: roughly a decade, sequential, from decision to a skilled cohort. Our way: start immediately — time-to-competence becomes just the length of the course itself, not the length of the course plus the years of building teachers first.

Human resources. The old way: you must divert your scarcest experts into training thousands of teachers before one student benefits. Our way: one expert's recorded time, once. The thousands of teacher-training hours are never spent — and those freed people (see the last piece) go to the work the country is actually short of.

The cost of not moving. Every year a country runs the slow chain instead of the fast one, it pays: the salaries and hours of the training pipeline; the output lost while its workforce isn't ready and its competitors' are; the businesses that can't hire the skills; and the compounding disadvantage of teaching a stale version of a moving field. The slow chain isn't free. Its price is a workforce that arrives late to every new technology — paid, quietly, in lost competitiveness, year after year.

A nation's ability to adapt to change should not be capped by how fast it can clone teachers. Remove that cap — record the expert once, distribute instantly, update overnight — and a society can finally re-skill itself at the speed the world is actually changing.

Next: what all this reclaimed time and mind are actually for — and why we may be teaching the wrong things entirely.

Written from a working platform and a concept put up to be tested — not from a result already claimed. — Crosshire.