Using LLMs as a learning tool isn’t what causes skill atrophy. It’s using them to solve entire problems without understanding what they’ve done.

And not even just understanding, but verifying that they’ve implemented the optimal solution.

It's partly that, but also reading and surface level understanding something vs generating yourself are different skills with different depths. If you're learning a language, you can get good at listening without getting good at speaking for example.