X Open-Sources Core Recommendation Algorithm, Developers Parse Weights with Grok

X (formerly Twitter) has officially open-sourced the core recommendation algorithm code for its "For You" feed on GitHub, aiming to enhance platform transparency, actively address criticism, and improve recommendation quality. The open-source repository has already garnered over 27,000 Stars. Simultaneously, the platform launched a new transparency initiative, allowing users to directly view the labels affecting post visibility and the reasons for traffic restrictions.

已确认

• 开源内容：Disclosed the core ranking system, model configurations, feed filters, and content labels impacting post visibility. The codebase also features a bidirectional follow graph.
• 持续更新：X pledged to publish future algorithm updates to the repository, enabling users to track specific changes via code diffs.
• Grok解析成果：Developers fed the entire 370,000 lines of code into the Grok model for deep analysis. Grok successfully extracted core mechanisms such as scoring weights, diversity penalties, and negative demotion signals.
• 排名机制：According to Grok's summary, the algorithm doesn't directly assign a "quality score" to posts. Instead, it predicts likely interactions from each reader, multiplies them by corresponding weights, and sums them up for a total score. The highest-scoring posts get recommended.

为什么重要

This move is regarded as a significant step toward algorithmic transparency for mainstream social media platforms. By leveraging large AI models to directly parse open-source code, developers have successfully cracked open the black box of social media algorithms. This not only validates the practical value of open-sourcing but also provides an excellent use case for understanding complex engineering systems.

2026-08-14 ~ 2026-08-14 · 7 related posts

• [source] X open-sources For You timeline algorithm and introduces visibility labels — GlenBradley · 2026-08-14
• X Open-Sources 'For You' Feed Algorithm for Full Transparency — tetsuoai · 2026-08-14
• [source] Musk Open-Sources X Algorithm: Grok Summarizes Interaction Weights — elonmusk · 2026-08-14
• X Open-Sources Recommendation Algorithm, Developer Uses Grok to Analyze Ranking — prasenx · 2026-08-14
• Elon Musk Announces X Open-Sources the For You Timeline Algorithm — elonmusk · 2026-08-14
• [source] xAI Open-Sources X Platform Recommendation Algorithm, Hits 27k Stars — prasenx · 2026-08-14
• User Feeds X's Open-Source Algorithm to Grok, Extracts Scoring Weights — prasenx · 2026-08-14