It seems like there's not much point to LLM+RAG over a good local search engine.

At least, that's what I've rapidly been concluding.

I'm still holding out a bit of hope for there being more though.

Could you expand on the reasons and possibly relevant papers?