Google seems to be better at bringing up a variety of stackoverflow and blogposts relevant to my search queries. Qwant seems to struggle exactly with that: it's great at giving exactly what I was searching for but that's sometimes not what I was looking for, if you get what I mean.

In a sense, LLM's are actually perfect for that. But like you say, the super confident hallucinations are just too frustrating. Literally every time I've asked it a serious programming question, it's hallucinated an API that doesn't exist. Everyone seems to be focusing on letting LLM's solve math and thinking problems. That's exactly what _I'm_ good at. I would much rather have an LLM that is good at combining sources and giving me facts (knowledge, rather than thinking) while most of all being able to say "I don't know".