Open YouTube on any random Tuesday and you’ll find a thousand variations of the same video. Build a SaaS in one hour, no skills needed. Make a logo in five minutes, no design knowledge required. Write a novel in a weekend, no experience necessary. Launch a course in one day, no expertise needed. The X keeps getting more specialized. The promised time keeps shrinking. The skills bar keeps dropping. And under all of it, quietly, something serious is happening — in my IT information bubble at least.
You’ve felt it too, probably. That faint background unease when you see one of those videos and think: wait, isn’t that something people used to study for years?
Yes. It is. That’s the point of the video.
Let’s start with what’s obviously true
Yes, the content in those videos is built on stolen work. The model was trained on every designer, every novelist, every developer, every course creator, every photographer who ever put their work online. None of them were asked. None of them were paid. The video is, in a real sense, the redistribution of skills that other people spent careers building.
Yes, it takes jobs. The freelance designer who used to charge €800 for a logo now competes with a teenager who generated thirty of them between maths classes. The freelance copywriter who used to charge by the day now competes with a SaaS that produces an entire content calendar before lunch. The boutique web agency now competes with a one-prompt site builder. None of this is a hypothetical future. It is happening right now, this quarter, in your inbox if you run any of those businesses.
But that’s not actually the part of this that keeps me up at night.
The deeper question (and it’s worse)
The part that keeps me up at night is this: if a skill takes ten years to learn properly, and AI can do 80% of what that skill produces with no learning at all — why would anyone choose to spend the ten years?
You can answer that for yourself, today, as an adult who already has skills. You’d say well, for love of the craft, or because I want to be excellent, or because the 20% AI can’t do is where the real value lives. All fair. All true. All things I’d say.
Now answer that for a fifteen-year-old.
A fifteen-year-old looking at the path you took — a decade of practice, repeated failure, an apprentice income, the slow grind to competence — can see, very clearly, the alternative. The alternative is: skip all of that, type a prompt, ship a result that’s 80% as good as yours, and use the time you saved to do something else with your life.
Why would they choose the ten years? Why would anyone, given the choice, choose the ten years?
The cruel asymmetry
Let’s suppose you push through. You’re the kid who decides, against all the incentives, to spend the decade. You become a master. You can do things in your field that AI demonstrably cannot.
Then you publish your work. Or you teach it. Or you stream yourself doing it. Or a client commissions it and someone takes a screenshot. Or it ends up indexed by a scraper. Or you talk about it in a podcast. Or your prompt sequence leaks. Or you train a junior who later goes home and uses AI to make sense of their notes.
Any one of these is enough. Your novel thing, the thing only you can do, is now training data. The next model version sees it. The version after that has internalised it. By the version after that, the thing only you could do is — structurally — the thing the model can reproduce on demand, for anyone, for the cost of a few tokens.
What I tell my daughter?
I’ve been thinking about this a lot, because I have a kid, and at some point she’s going to ask me what she should learn.
The honest version of my answer is roughly: learn things because you want to know them, not because you expect to be paid for them. Learn things that involve your body, because that’s the thing AI can’t do. Learn things that involve being legally or emotionally on the hook, because that’s the thing nobody wants AI to do. Learn how to talk to people who don’t want to be talked to. Learn how to notice things. Learn how to wait. Learn how to be wrong without breaking.
None of that is on a YouTube list of in-demand skills. Most of it is unfashionable. Some of it sounds suspiciously like advice my grandfather would have given me. I don’t love that. But I think it’s the most honest answer I’ve got.
The skills that used to make a person economically valuable — technical specialisation, expertise in a domain, the ability to produce a polished artifact — are precisely the skills that AI is best at, and getting better at fastest. The skills AI is worst at are the unmeasurable ones. Judgement. Taste. Presence. Care.
Those happen to also be the skills that take the longest to build, that nobody can teach you in an hour, and that no YouTube video promises.
Where this lands
I don’t have a clean conclusion for this one and I don’t want to pretend I do.
The short version is: a lot of the skills my generation built our identities around are being devalued at a speed that’s hard to process emotionally, let alone economically. The long version is: I think this is going to produce a generation that learns less, makes less, builds less of an identity around what they do for a living, and has to find meaning somewhere else. I don’t know if that’s good or bad. It’s probably both.
What I do know is that “just learn AI” is not the answer either. The same logic applies one level up: today’s prompt engineer is tomorrow’s commoditised role. The model gets better at prompting itself, and the human-in-the-loop premium shrinks. There is no safe terminal skill in this picture. There is only the moving frontier.
The series so far has been about AI being dumb, AI being miscalibrated, AI being misused. This post and the one before it are about something a little different — not what AI gets wrong, but what it gets right, and what that costs. Because the loud failures are easy to laugh at. The quiet successes are the ones that change the shape of a life.
If you want the version of this question with much higher stakes — where the consequences of AI being trusted run from “awkward inbox” all the way to “catastrophic and deniable” — the closing piece in this series is Oopsie with deadly consequences — or plain evil?.
English 
Dutch