The Om Universe.
I exported my entire YouTube watch history from Google Takeout — 13,655 videos — and sorted every title into topics with a rule-based classifier. The result is a fairly honest map of what my brain gravitates to: CFA prep and gaming neck-and-neck at the top, then music, movies, gadgets, comedy, and Formula 1. Drag the planets. Hover for counts.
Where the gravity is.
The full ledger.
“Other” is the honest bucket: titles with no useful keywords and one-off rabbit holes. “Noise” is ads YouTube served me and videos that have since been deleted or made private (Takeout only leaves a URL for those).
| Topic | Theme | Share | Videos | % |
|---|
How the map is built.
Google Takeout
Takeout → YouTube and YouTube Music → history → watch-history.html. My own account, my own machine.
Titles only
A small Node script pulls the title of every “Watched …” entry. Video IDs, channels, and timestamps are dropped on the spot.
Ordered rules
A Python classifier runs ~50 regex rules in priority order (finance before tech before entertainment); the first match wins. Topics roll up into themes.
Canvas physics
Repulsion + springs + gravity in ~150 lines of vanilla JS. No libraries. Node size ∝ √(count).
Provenance.
Every number on this page traces back to one file I exported myself. Nothing is scraped, estimated, or pulled from an API — and nothing personal leaves the pipeline except counts.
The first version of this map (Feb 2026) reported 1,433 “AI / ML” videos. That was a bug: the rule matched the letters ai inside words like Captain and Britain. The re-classifier uses whole-word rules; the real AI/ML count is around 170. Gaming, on the other hand, turned out to be the single biggest bucket — which the old map missed entirely.