fixedprog

Hmm, I’ve been trying to work this out myself.

I do use LLMs to try and help me learn, use inquiry-based learning and have it guide me to solutions without telling me the answer, but I’m yet to work out whether I’m actually learning more.

I think it was Socrates who talked about needing a certain level of discomfort in order to learn effectively. I do feel that even with that approach, LLMs don’t provide enough discomfort.

The difficulty for me is that they are simply excellent at retrieving information about certain topics, so it also seems silly to avoid them entirely.

I’d be curious to hear other people’s experiences as well.

I like to use AI for research and learning about new topics, and for occasional code snippets.

I'm not completely sold on the "agent" thing yet, although I do admit they can be pretty fun to use.