Courses are watched once but referenced for years. This system archives every course locally (recordings, transcripts, decks) before access lapses, then extracts the knowledge into structured records so it can be searched, cross-referenced, and revisited at the exact moment in the video where something was taught.
How content is organised
Source
A session, office hours, deck, YouTube video, article, or book — where knowledge came from.
Concept
An idea or technique (chain-of-thought, RAG). The primary way in — concepts link every source that teaches them.
Tool
A product or service (NotebookLM). Tracks demonstrated vs explained vs mentioned, plus freshness: Current / Shifted / Superseded / Discontinued.
Category
Topical groupings of concepts and tools (Prompting, Automation, Vibe coding).
The pipeline
1 · A nightly job checks the course platform and archives anything new. 2 · Transcripts are condensed and knowledge is extracted into YAML records — the only permanent data format, git-versioned. 3 · Every page here is regenerated from those records; nothing is hand-edited. 4 · A search index (SQLite full-text + vector) is being added for transcript-level and semantic search.