The "transcribe the tutorial" workflow
Conference talks and YouTube tutorials are how most engineers actually learn new frameworks. Watching them is fine; remembering them six months later is hard; getting Cursor to write code that reflects them is impossible without the content being in the workspace. The fix: paste the talk's YouTube URL into the converter, save the resulting .md as docs/learning/2026-03-tanstack-router-deep-dive.md, and Cursor's indexer picks it up automatically.
From that point, every Cursor chat in the workspace can reference the tutorial. @docs/learning/tanstack-router-deep-dive.md what was the recommended pattern for nested routes? returns the actual exchange, with timestamps. Asks like "implement nested routing per the patterns in the linked tutorial" produce code that follows the talk's recommendations.
Per-framework learning libraries
Build out docs/learning/ as a per-framework reference. Convert the official launch talk, the deep-dive workshop, the maintainer's podcast appearance — three Markdown files per framework, indexed by Cursor, available to every chat. This compounds: after a few months you have a curated learning library that genuinely changes the AI's code suggestions.
Pair with PDF documentation (PDF for Cursor), official docs scraped from the web (URL for Cursor), and audio podcasts (Audio for Cursor) for a complete framework reference.