A relative asked me at a wedding what I planned to study, listened to my answer, and said "beta, AI will do all that in five years." Then he went back to his plate. Very reassuring, thanks.

The thing is, he isn't making it up. Something real is happening. But I've since read a bit about what happened the last several times people were certain the jobs were ending, and the pattern is strange enough to be worth your ten minutes — because it changes what you should actually do about it.

---

The same sentence, over and over

This panic is not new. It's arguably the most repeated prediction in economic history, and it has a name: the lump of labour fallacy — the assumption that there is a fixed amount of work in the world, so any machine that does some of it permanently removes that much.

*Notice what these have in common: the task was automated, and the person ended up doing something one level higher up.

---

What actually happens: things get abstracted

Here's the pattern underneath all of it, and it's the only idea in this post you need to keep.

Technology doesn't delete work. It hides a layer of it, and pushes people up to the layer above.

Each new tool buries a layer. You end up standing on top of it.

Programming is the cleanest example. Early programmers wrote in assembly, which meant telling the machine every individual step. Then compilers arrived and did that translation automatically. Nobody said "programming is over" for long — instead, programmers stopped worrying about registers and started worrying about what the software should actually do.

The low layer got hidden. The number of programmers went up, not down.

Spreadsheets did it to accounting. The person who used to spend a week adding columns spent that week asking what the numbers meant instead. Same person, harder question, more valuable answer.

AI is doing this to a much wider set of tasks at once — drafting, summarising, first-pass code, routine analysis. Which is exactly why it feels bigger. But the direction is the familiar one: the doing gets cheaper, and the deciding gets more valuable.

---

The part that's genuinely true

I'm not going to tell you it's all fine, because that would be the same useless reassurance my relative offered in reverse.

Some jobs really do end. Not "transform" — end. Switchboard operators, once a huge workforce, were eliminated by automatic exchanges. Typesetters, lift operators, professional typists: those roles are gone, and the people in them mostly did not get promoted to something better. They absorbed the cost while everyone else enjoyed the cheaper prices.

These were real careers. They didn't move up a layer. They stopped.

So the honest version is this: the total amount of work has never run out, but individual people have absolutely been left behind by these shifts. Aggregate optimism is cold comfort if you're the one holding the obsolete skill.

Which means the useful question isn't "will there be jobs?" — historically, yes. It's "will I be standing on the layer that survives?"

---

So what do you actually do about it

Not "learn AI." That's the advice everyone gives and it's nearly useless — like being told to "learn computers" in 1998.

Aim one layer up instead. Concretely:

1. Become the person who decides what to ask for. AI is very good at producing an answer and completely indifferent to whether it was the right question. That gap is where you live.
2. Get good at judging output. Anyone can generate a draft now. Knowing that draft is wrong — in a subject you actually understand — is the scarce part.
3. Learn the fundamentals under the tool, not the tool. Tools change every eighteen months. Knowing why something works survives all of them.
4. Build things people can see. When output becomes cheap, evidence that you can produce it becomes the differentiator. A portfolio beats a claim.
5. Practise the things that don't automate — explaining an idea to a room, handling someone who disagrees, taking responsibility when a decision goes wrong.

The job is moving from producing the answer to knowing whether it's the right one.

None of that is a trick to dodge AI. It's just where the value went, the same way it went from arithmetic to analysis when the calculator showed up.

---

One more thing about your relatives

Every adult confidently predicting your future is extrapolating from theirs. The uncle who says AI will take everything watched computers arrive and is pattern-matching. He may be right about the shape and wrong about the outcome — that's usually how these predictions land.

Nobody in 1995 could have written down "cloud architect" or "UX researcher." The jobs you'll do have names that don't exist yet, and bas, that's genuinely how it has gone every single time.