Even ignoring the compounding of learned skills over time, I've noticed that when relying on AI that my brain switches off and I stop thinking critically. This is a bad habit, and works against the idea of checking the AI results for quality. When things don't work, why turn on the brain to figure it out, when you can just ask the AI to try harder. This is often where a lot of time is spent. I once went back and forth with the AI for quite a while trying to solve a little bug in a script I wanted, but didn't know the language that well. It was a helper tool for me personally, not a deliverable for my job. After getting fed up, I finally looked more at the code and saw that it was an off-by-one error. Even telling the AI this, it couldn't fix it. I ended up fixing it myself. I had written a similar script in another language in the past. Mine had less error handling, and was a bit less fancy, but it was good enough for what I needed, I understood what the code was doing, and I didn't get stuck on any silly errors along the way. Debugging other people's code is not fun, and that's mostly what AI coding turns into.