X updates timeline recommendations with new AI ranking model

X, the platform formerly known as Twitter, has once again tinkered with the machinery deciding what lands in your feed. On October 20, 2026, Elon Musk announced two updates to the timeline recommendation system. One is a new AI/SI ranking model built to better predict how long users will spend on content. The other is a retrieval model that favors posts people actually read to the end. Musk paired the news with a bold engagement claim. He said active user seconds on X are growing exponentially and hitting new all-time highs nearly every day. What changed in the feed The retrieval model pulls in posts with high completion, meaning posts that users tend to read all the way through rather than abandon halfway. The new ranking model’s job is to predict dwell time, which is the industry term for how long someone lingers on a piece of content. Phoenix, Grok and the open-source pitch The updates sit inside X’s broader ranking framework, known as the Phoenix model. Phoenix runs on a transformer architecture adapted from Grok, the AI model tied to Musk’s xAI venture. Phoenix doesn’t produce a single verdict on a post. Instead, it calculates probabilities for a range of possible user actions...