An interesting use case of LLMs is as a conversational interface to some body of text, as a complement to other forms of navigation: from a table of content or index, via full-text search, or reading specific sections or even the whole text cover-to-cover.
Of course the LLM can still make up nonsense, but this is a similar limitation to a full-text search leading to a section irrelevant to your current question if the search terms match too broadly. With any navigation method, eventually you need to read the material once you find the section you need.
The Phenix documentation, papers, newsletters and tutorial videos have been fed to an LLM, so now we can navigate it as a conversation: https://phenix-online.org/version_docs/2.0-5725/reference/chatbot.html
A quick test trying to answer a question I knew the answer to suggests that it is working pretty well for sufficiently specific questions. This will likely be useful.