X Deploys New AI Ranking Model and Retrieval System to Improve Timeline Feeds
X has rolled out a new AI ranking model and retrieval system designed to predict dwell time and surface posts users read to the end.
X Deploys New AI Ranking Model and Retrieval System to Improve Timeline Feeds
X has rolled out a new AI ranking model and retrieval system designed to predict dwell time and surface posts users read to the end.
X AI ranking model introduced for timeline feeds
Social media platform X has introduced significant updates to its timeline recommendation system, focusing heavily on user engagement metrics. According to Crypto Briefing, platform owner Elon Musk announced the deployment of a new X AI ranking model alongside a separate retrieval system on October 20, 2026.
The new ranking model utilizes artificial intelligence to better predict dwell time, which measures the duration users spend lingering on specific content. Simultaneously, the newly implemented retrieval model is designed to surface posts that users read completely rather than abandon halfway through.
Integration with the Phoenix framework and Grok architecture
These recent adjustments operate within X's broader recommendation framework known as the Phoenix model. Phoenix is built on a transformer architecture that was adapted from Grok, the artificial intelligence model developed by Musk's xAI venture.
Rather than issuing a single verdict for individual posts, the Phoenix system calculates probabilities for various user actions. These calculated probabilities are then combined into a comprehensive engagement score that dictates the placement of posts within user timelines.
Ongoing algorithmic evolution and engagement claims
The October modifications follow a series of earlier adjustments made to X's ranking models throughout 2026, with prior updates shipping in January, May, and August. Previously in August, the company made headlines by open-sourcing its model-training code, filters, and ranking weights under the Apache 2.0 license.
Alongside the algorithmic update, Musk reported that active user seconds on the platform are experiencing exponential growth and reaching new all-time highs on a near-daily basis. However, these engagement claims were presented without independent figures to verify the scale of the reported growth.
About The Author
