X Updates Timeline With New AI Ranking Models
The platform has deployed new machine learning models designed to prioritize posts based on predicted user engagement.
Updated on Oct. 20, 2026 in Artificial Intelligence
Live Poll
Do you trust that social media algorithm changes prioritize the content you actually want to see?
X has updated its timeline recommendation system to include two new AI models focused on predicting dwell time. These changes aim to improve user engagement metrics across the platform.
Why it matters
By implementing models that favor posts read to the end, the company seeks to refine its content delivery to boost user interaction. This shift prioritizes high-completion content within the user experience.
The Phoenix ranking framework utilizes a transformer architecture adapted from Grok to calculate probability scores for user actions. This system aggregates engagement data to produce an overall score for individual posts.
The players
X
This social media platform utilizes machine learning algorithms to curate and display content for its global user base.
Elon Musk
He is the owner of X and frequently communicates updates regarding the platform's infrastructure and user metrics.
The details
The retrieval model identifies and elevates posts based on high user completion rates, while the ranking model predicts the total time spent on specific entries. X also made its ranking weights, filters, and training code available to the public in August 2026.
Timeline
January 2026: X deployed the first major update to its ranking model.
May 2026: The platform shipped its second major update to the ranking model.
August 13, 2026: X open-sourced its ranking weights, filters, and training code.
October 20, 2026: Elon Musk announced the latest timeline recommendation updates.
The Tech Race
The adaptation of Grok architecture for X timeline ranking marks a shift in how social platforms leverage proprietary large language model foundations to optimize feed relevance. This evolution moves past static algorithmic filtering toward dynamic, model-driven content prediction.
Users can expect to see more content in their feeds that the system predicts will keep them reading until the end. This update may change the variety of posts appearing on the timeline as the platform emphasizes predicted dwell time over other metrics.
The takeaway
The move toward completion-based ranking suggests a strategy focused on increasing the depth of user engagement rather than just the volume of posts viewed. Readers should monitor their own feed changes to see if the content they prioritize reading is surfaced more frequently.
Further reading
Learn more about evolving Artificial Intelligence trends and their integration into major social platforms.
Live Poll
Do you trust that social media algorithm changes prioritize the content you actually want to see?
