The System That Learns From Itself: Recommender Systems from GroupLens to Two Towers

Deep End Era. Engineering with Sankalp · Oct 28, 2026 · 52m

Lantern's new music recommender has raised plays per session by eleven percent, and Tunde has a chart showing the number of artists reaching listeners has collapsed from forty thousand to twenty-six thousand. Part nine of The Deep End's Machine Learning series is about the machine learning that decides what you see next. We go from Xerox PARC's Tapestry and Minnesota's GroupLens to Amazon's "customers who bought this also bought", then to the Netflix Prize, where a blogger called Simon Funk gave away the key idea and the million dollars was decided by a twenty-minute gap. Matrix factorisation…