I've found that if you give it pairs of "Why..." questions it'll happily argue both sides of the coin ("Why is X better than Y", "Why is Y better than X"). But in giving me justifications, it gives me fodder for further traditional searches. I don't trust it to give me a correct answer, but I do trust it to define the problem space a bit.

You're typing out 300 queries to ChatGPT every day?!

Asked it a situation with examples where I’d use type erasure in swift. It looked good but was a mesh mash off various related things