That's strange, I've never thought it can be done this way. Normally I'd read the docs, maybe sketch up some diagrams, then maybe take a walk while thinking on how to solve the problem, and by the time I got back to the screen I'd already have a good idea on how to do it.

These days the only difference is that I feed my ideas to a few different LLMs to have "different opinions". Usually they're crap but sometimes they present something useful that I can implement.