I’ve been working on a learning / incremental reading tool for a while, and I’ve found LLM and LLM adjacent tech useful, but as ways of resolving ambiguity within a product that doesn’t otherwise show any use of LLM. It’s like LLM-as-parser.