In this new world of seemingly endless possibilities, I’m finding that more of my work involves imposing constraints, and that the job has got harder, not easier.

Jack White once said that the whole point of the White Stripes was the liberation of limiting yourself. The band followed self-imposed rules: no bass, no guitar solos, no slide guitar, no covers, three colours, two people. Out of those rules came one of the most distinctive rock bands of the last twenty-five years. Years later, on his solo album Lazaretto, he did it again with the physical record itself: a vinyl LP with hidden tracks under the centre label, a side that plays from the inside out, and a hand-etched hologram in the dead wax that only appears when the record spins. He’s spoken openly about this being a pattern. “I’m always imposing restrictions on myself,” he said in one interview. The constraints aren’t obstacles he’s working around. They’re the thing he’s working with. They’re how he gets to the work he wants to make.

Great design has usually been shaped by constraints. The size of the page, the budget, the brief, the team, the platform, the knowledge I happen to have at the time, the hours I have to make the work good. They give the work something to push against. They stop design becoming endless. They force decisions, and decisions are what design actually is.

AI seems to change that. It opens doors. It gives me access to more ideas, more references, more technical possibilities, more ways of testing an approach. It lets me make things I might not have been able to make before, or at least not as quickly. The common framing, and you hear it everywhere now, is that the only remaining constraint is your imagination.

I don’t think that’s true, and I don’t think it’s even close to true.

What’s actually happened, in my experience, is that the constraints feel like they’ve moved rather than disappeared. They used to come from outside me; now more often than not, they have to come from me. The page size used to be fixed by the printer. The budget used to be fixed by the client. The technical possibilities used to be fixed by what I personally could build. Those external limits did a lot of quiet work. They eliminated options before I ever had to think about them. They made the space of possible designs small enough that I could hold it in my head and make a choice.

AI removes that quiet work. It hands me a space of possible designs that is effectively infinite, and it hands it to me cheerfully, with no opinion about which corner of it I should be in. So if I want the work to cohere, if I want it to feel like something rather than like a competent average of everything, I have to provide the constraints myself. I have to decide what this brand sounds like, what it never does, how it spaces things, how it treats edges, what its relationship to the grid is, what it owes its users, how it behaves when things go wrong. I have to hold all of that in my head and bring it to bear on every output the tool gives me, because the tool will not bring it to bear on its own.

That is more demanding than the work I used to do, not less. I work harder now than I did before AI. The prevailing story is the opposite. The story is that AI makes creative work faster and easier, that it frees designers from the grunt work so we can focus on the ideas. That hasn’t been my experience. AI has made the grunt work fast and made the thinking heavier. There is more to shape, more to question, more to refine. The cost of producing a thing has collapsed; the cost of producing the right thing has, if anything, gone up, because the right thing now has to be picked out of a much larger field of plausible-looking wrong things.

There’s a version of this argument that sounds like a retreat: the designer as editor, the designer as curator, the designer as the person who says no to the machine. That isn’t what I mean, and it isn’t what I do. I still design. The system the AI is working inside came from me. The rules it’s adhering to came from me. The style is still mine. I make hundreds of choices a day that have nothing to do with constraining a model and everything to do with the ordinary work of being a designer: looking at something, feeling that it isn’t right, and changing it until it is.

What’s changed is that the constraint-setting is now a much larger part of the load, and it’s a kind of design work in itself. Deciding the rules of a system is not less creative than executing inside one — it might be more creative, because the rules determine everything that follows. When I write a guideline that says we never use shadows on this surface or all our buttons inherit this radius or this is how we handle a low-vision user reaching this screen, I’m doing design. I’m just doing it at a level the AI can’t reach on its own, because the AI has no view of the whole. It generates locally. I have to think globally. The coherence of the work, across screens and components and contexts, is something only I can be responsible for.

So that’s how I think about it now. AI hasn’t made the job easier and it hasn’t made me less of a designer. It’s made the job bigger and more layered, and it’s made the role of the designer more intentional, because intention is the one thing the tool can’t supply. Jack White was right. Limiting yourself is liberating — but only if the limits are yours. The constraints used to come for free, from the world. Now I have to bring them. That’s harder. It’s also more interesting.