This post was originally written in Italian by Paolo Gambardella. It has been translated by an AI agent and may contain inaccuracies.
When there’s a new design problem, my first move can be to open Claude and ask it to structure things for me. Sometimes that happens out of laziness, sometimes out of a genuine fear of getting it wrong.
This week I read two articles:
The first, on the Atlantic, argues that when “intelligence” becomes abundant, what actually makes the difference is volition. The winners aren’t those who use AI to do less work — they’re the ones who use it to push themselves harder, to demand more of themselves in the process.
The second, by Lars Faye, argues that AI coding tools remove exactly the friction that builds real expertise. The paradox is that you need expertise to use these tools well, but using the tools can prevent that expertise from forming in the first place.
The HOW matters
For years I’ve written the first draft of a design pillar or a one-pager by hand, before I even open an editor. That’s the effort I need — the thing that makes me truly understand what I’m actually proposing.
If an algorithm decides how a problem is framed, we end up as mere executors, and we lose the part that’s supposed to make us grow as designers. Claude and other platforms bring the difficulty of framing down to nearly zero, but that’s precisely the bit that should stay ours. Identifying a problem is slow work by its very nature — that’s where judgement gets trained.
So the rule I set for myself is a simple one: I sketch out the framing on my own first, even if it’s rough, even if it’s wrong. Then I bring it to Claude to sharpen.
AI isn’t just a tool — I’ve written about this before. The real risk is losing sight of when you’re using it to dodge the kind of struggle that actually matters.