AI in the World
Recommenders: The Algorithm That Knows You
6 min+55 XPStep 1/11
The engine behind your feed
Netflix doesn't ask what you want to watch — it watches what you watch. Every play, pause, rewind, and abandoned episode is a data point. Recommenders learn from behavior, not opinions.
The core trick is collaborative filtering: find people whose taste overlaps with yours, then surface what they loved and you haven't seen. Nobody tags a film 'moody-but-hopeful' — the model discovers taste dimensions no human ever named, just from who-watched-what.
insight
Estimates suggest recommendations drive roughly 80% of Netflix viewing and 70% of YouTube watch time. For these services the algorithm isn't a feature — it's the front door.