TikTok and the Sorting Hat
Source: https://www.eugenewei.com/blog/2020/8/3/tiktok-and-the-sorting-hat Author: Eugene Wei Date: 2026-02-10 (saved, originally 2020)
Summary
Eugene Wei’s definitive analysis of TikTok’s algorithm. The “sorting hat” metaphor: TikTok puts every user through a rapid sorting process to find their taste graph, then serves optimal content. The speed and precision of this sorting is TikTok’s core innovation.
Key Claims
- Cold start problem: most recommendation systems take months to understand a new user. TikTok sorts users within minutes.
- The interest graph vs. social graph: TikTok’s recommendation is based on what you like, not who you know. This removes the bootstrapping problem (you need friends to get content from) that plagued earlier social networks.
- Infinite scroll as a test harness: every swipe is a signal. TikTok processes billions of micro-signals to build precise taste models.
- The “sorting hat” experience: users describe finding their “niche” on TikTok surprisingly fast — sports, crafts, humor, education.
- Implication for AI products: the same “rapid taste sorting” approach could apply to any personalized AI experience.
Concepts
- Agent Memory — rapid user modeling from behavioral signals