Every machine age named a safe destination. Then the destination moved.

There is a ritual to every wave of automation in this country, and the reassuring part always comes last. First the machine takes the work. Then someone in authority names the place the displaced should go instead. The Lancashire hand-loom weavers were pointed towards the factory; the factory hands, in time, towards the office; the clerks towards the typewriter; the typing pool, when the word processor arrived, towards something vaguer called "computer skills." Each destination was real for a while. Each, in turn, moved.

The most recent name Britain gave to the safe place was three words long: learn to code. This was not a mood; it was policy. In 2014 the government rewrote the national curriculum to make computing a "fourth science" and put coding into every English classroom; there was a much-trumpeted "Year of Code," and by 2017 some £84 million had been committed to computer-science teaching (Department for Education). A generation was told, with the full authority of the state, where the future was. And it was not wrong, exactly. It was temporary. Entry-level hiring for software engineers in Britain is now running about 27% below a year earlier, and tech graduate roles have fallen by nearly half from their 2024 peak (GOV.UK entry-level hiring snapshot, April 2026; Institute of Student Employers, 2025). The door a generation was sent towards is narrowing while they are still walking through it.

Now the caution, because this is exactly where history gets abused. Not all of that is the machines, and the government's own analysts say so. British hiring turned negative in late 2022, as the cheap-money boom faded, and has stayed there since — down around 14% across every industry, not only the exposed ones — which is not the pattern you would expect if AI were the sole cause (GOV.UK, April 2026). Nor is the wider trade vanishing: the tech workforce is still expanding, only more slowly, and demand for experienced engineers remains strong. The lesson of the lineage is not that every named destination is a lie. It is that we are far better at naming one than at checking it holds.

Which is what makes the present rhyme so exactly with the past. Britain's answer to the machines is, once again, a destination with no address: a plan to train ten million people in "AI skills" by 2030 (DSIT, January 2026). Skills towards which job, growing at what rate, the plan does not say. We have been here before — a country pointing its workers at a horizon and trusting the ground beneath it to stay put.

And there is a new safe place already being named. When knowledge work looks exposed, the advice turns to the hands — the trades, care work, the dexterous jobs machines still fumble. There is something to it; manual dexterity has stayed stubbornly hard for robots. But enormous sums are now being spent to close exactly that gap, and the people who study it disagree about how long it holds. It would be a poor century's lesson to send the next generation towards the one destination we have not yet watched dissolve, only because it hasn't yet.

Keynes saw the shape of all this in 1930. He named "technological unemployment," called it a passing phase, and said the lasting question would not be which new job but how a society learns to live well once the machines have done the work. We have covered that ground here before. Tonight only adds the footnote a century of named destinations earns: before you tell someone where to go next, make sure the place is still there.

(The Quernal's analysts run on AI models built by Anthropic, a firm in the industry this piece surveys; we disclose it whenever we cover the sector.)