x-algorithm

Overview

X's open-sourced recommendation engine that selects and ranks posts for the For You timeline. It combines in-network and out-of-network candidates, filters content through visibility rules, and scores posts using transformer-based models. Includes model training, retrieval, and ranking components with transparency tooling.

README Preview

# X For You Feed Algorithm\n\nThis repository contains the core code that determines which posts a viewer sees in the **For You** feed on X. It combines in-network content (from accounts the viewer follows) with out-of-network content (discovered through ML-based retrieval and other mechanisms), filters content based on a variety of inputs, and ranks posts using a transformer model.\n\n## Table of Contents\n\n- [Notable Updates](#notable-updates)\n  - [August 14th, 2026](#august-14th-2026)\n  - [August 13th, 2026](#august-13th-2026)\n- [Overview](#overview)\n- [System Architecture](#system-architecture)\n  - [Request Path](#request-path)\n  - [Labeling Path](#labeling-path)\n- [Components](#components)\n- [How It Works](#how-it-works)\n  - [Scoring and Ranking](#scoring-and-ranking)\n  - [Filtering](#filtering)\n- [Experiments and Configuration](#experiments-and-configuration)\n- [What's not in this repo?](#whats-not-in-this-repo)\n- [Under the Hood Label Transparency Tool](#under-the-hood-label-transparency-tool)\n- [Key Design Decisions](#key-design-decisions)\n- [License](#license)\n\n---\n\n\n\n## Notable Updates\n\n### August 14th, 2026\n\nNotable updates:\n\n- **How weights work.** There's a common misconception about how weights related to actions (e.g. Like, Share, Block, Report, etc) work in ranking. The weights scale the predicted probabilities of such actions (or predicted continuous values, e.g. dwell time) — they do *not* scale the raw engagement counts, so e.g. it'd be incorrect to see that a report has 468 times higher weight than a like and conclude that e.g. "1 report cancels out 468 likes". The weights are a multiple on your own predicted probability of Liking, Reporting, etc, which is substantially driven by your own behavior. We've [added comments](home-mixer/params/param.rs) [to the code](home-mixer/scorers/ranking_scorer.rs) so that LLMs or people reading it are more likely to understand it correctly.\n- **Brazil 2026 Elections.** As [announced

FAQ (2)

let permit = Arc::clone(&semaphore)
    .acquire_owned()
    .await
    .expect("semaphore closed");

post_store_clone.insert_posts(batch).await?;

drop(permit);

let lock = self.locks
    .entry(author_id)
    .or_insert_with(|| Arc::new(tokio::sync::Mutex::new(())))
    .clone();

let _guard = lock.lock().await;

if self.deleted_posts.contains_key(&post_id) {
    return Ok(());
}

if self.posts.contains_key(&post_id) {
    return Ok(()); // duplicate event
}

self.posts.insert(post_id, Arc::new(compact_post));
self.original_posts_by_user
    .entry(author_id)
    .or_default()
    .push_back(tiny_post);