sentdex (Python + ML projects)

sentdex (Python + ML projects) enters the discovery pipeline as an Atom feed — same push model as RSS. Content lands in the ingestion pipeline, where it is classified against the taxonomy and written to a human-reviewed draft queue — nothing auto-publishes.

Fetched weekly. Authority weight 4/5 feeds the source-quality score, so content from vetted publishers ranks higher before any human review. Discovered items default to the cat-python category.

No drafts recorded yet. New sources enter the rotation and produce drafts on their first successful fetches — absence here means not yet crawled into a draft, not absence of content.

Every registry claim should be checkable at its origin: https://www.youtube.com/feeds/videos.xml?channel_id=UCfzlCWGWYyIQ0aLC5w48gBQ. If this source ever becomes unreachable, it is skipped with a warning — the pipeline records the failure and the rest of the run continues.

FETCH → EXTRACT → CLASSIFY → DRAFT. A weekly GitHub Action runs the discovery loop, opens one PR per batch, and a human curator accepts, edits or rejects each item. Scores are computed, never hand-written — see the scoring model and verification levels.