I keep reading opinions about why we should continue writing code ourselves. One argument is that we need the practice to develop our minds.
I think writing code yourself is useful practice. You make decisions, get stuck, and work out why something fails. But I think the detailed thinking about the code is what develops us. There is room for that when AI writes it too.
I have never done so much thinking as I have in the past year working with AI. Like, ever.
Suppose you are deciding whether a system should be one application or several services. Where do the responsibilities belong? Which parts really need to change independently? Would splitting them make the system easier to understand, or just move the complexity into how the parts communicate?
Working through those questions, comparing solutions, and questioning the approach takes thought, even if AI writes every line.
Of course, accepting the result can skip the learning. A small coding experiment found worse immediate understanding with AI assistance overall. Participants who used it to explore concepts and understand the generated code tended to do better, though those usage patterns were observational. That is a useful distinction, not proof that reviewing and writing teach the same skills.
I am optimistic about the capacity AI unlocks in us. It is just a different type of thinking.