If AI Suddenly Went Kaput, What Would You Still Know How to Do?
AI agents have moved past the chat box. The useful ones can control your entire computer and do whatever you ask.
And I don’t want to make every call. God no. I just don’t want to lose the ability to know when something’s wrong.
That worry comes from a strange place. As a kid I wanted to be an archaeologist. Tintin was my favourite comic, and I was obsessed with old ruins and what people knew thousands of years before us. Two sites still sit at the top of my list, Göbeklitepe, around 7,000 years older than the pyramids, and Abu Simbel in Egypt.
And no, I’m not talking about ancient aliens.

What gets me is how much people built and understood, then lost and had to rediscover later. Knowledge doesn’t usually vanish in one go. It fades when the thing keeps getting done for you, and nobody practises it anymore.
That’s what I worry about with AI. If I let it make the call too often because it’s faster, making the call myself starts feeling like a chore. You don’t wake up stupid overnight. You just get dumber without noticing.
Progress turns both ways
Like those iridescent wind spinners, the same motion can look like it’s going up or down depending on how you look at it. A tool can be progress and a trade-off at the same time.
With AI, which way it spins depends on who makes the call. If it helps you think and you still decide, you get sharper. If it decides, explains, remembers and fixes everything for you, you slowly forget how to do it yourself.
The difference isn’t AI or no AI
Most of us will end up using AI somewhere, even if we don’t realise it. The real difference is between people who stay in control and people who hand it over because it’s easier.
From the outside they can look just as productive for a while. Until something breaks, and then you find out who still knows how to think it through.
How I design for it
This is why I build AI tools that keep people making the call. In Lola Studio, every scene gets checked against the story before anything is generated, so I’m still the director, not just the person paying for takes. In Cozy Builder, the AI’s layout comes in as a draft you look over first, so you’re never stuck accepting a messy scene or binning the lot.
The AI does the heavy lifting. The judgement stays with you.
So where are we heading?
This is the closest I’ve come to doing the thing I wanted to do as a kid, which is funny because I’m not standing in a desert with a brush in my hand; I’m sitting here looking at AI and wondering what future humans might say about us.
I don’t think we’re heading toward one clean future. AI will make some people sharper, and some people will lean on it more than others.
So “should we use AI?” feels like the wrong question. The better question is, what are we still practising while we use it? If I stop practising something for long enough, I should expect to lose it.
That’s the part I keep going back to with Göbeklitepe and Abu Simbel. What survives? What disappears? What did people once understand so well that nobody left enough behind for us to see the full picture?
If you’re using an agent and it disappeared tomorrow, what would you still know how to do, and what would future-you have to dig back up?