Local LLM with the entirety of Wikipedia accessible via RAG (or better, newer technique) is legitimately a little super-powered assistant. It works when the grid goes down, and the searchability is orders of magnitude better - especially when you're not sure what you're looking for.