Interesting! I did something similar with 12+ years of my writing in an attempt to understand the trajectories of my attitudes. Downloaded my Reddit data and ran some extra processing to contextualize my comments within as much of the original discussion as possible. Created some specialty prompts, did some tests and checks, then let it rip.<p>It makes me happy to see someone else thinking about using a LLM to analyze stories and writing to understand people/characters.