A technique for extracting sentiment and feature requests from a large, unstructured text source: feed the entire source — e.g., a full Discord channel's transcript — directly into a large-context LLM (Gemini was cited) and ask it analytical questions about sentiment and requested features, rather than manually reading or, more likely, skipping the analysis because of its volume. The source described results from this as 'a gold mine' precisely because the volume made manual reading impractical — the technique makes previously-too-large feedback sources analyzable at all. Related to Internal AI App for Sales-Call Feedback Mining (Gong Transcripts) and Launch-Readiness Cross-Referencing Prompt, which apply the same 'point a model at raw internal data' pattern to other sources.
A PM used Gemini's expanded context window to dump an entire Discord community's transcript in one pass and ask for sentiment and feature-request analysis, work that would never have gotten done manually and is described as unlocking a 'gold mine' of already-existing signal. Framing: 'I'm too busy to read all of this' is a leverage problem, not a time problem — long-context ingestion turns an infeasible manual-reading task into a single prompt.