Source: LessWrong
The shift from engagement-optimized algorithmic feeds to LLM-driven personalized curation threatens platforms like Meta and TikTok, which monetize attention extraction rather than relevance matching. A new class of startups can now offer superior discovery by using language models to understand user intent and content nuance in ways that traditional collaborative filtering cannot. This collapses the gap between what algorithms currently show you and what you actually want to read. Whoever owns the interface between users and their information diet first—and trains an LLM on actual preference data rather than engagement metrics—can fragment the oligopoly's hold on how we encounter media.