Skip to content

darkresearch/x-algorithm-breakdown

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

3 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

X Algorithm Breakdown

Grow your X (Twitter) account using insights from the actual algorithm source code.

X open-sourced their recommendation algorithm. We analyzed it so you don't have to.


The Short Version

The algorithm predicts one thing: "If I show this post to this user, will they engage?"

It scores every post based on 18 different actions users might take. Your job is simple: create content people want to engage with.

What Gets You Reach

Action Impact
Likes, Replies, Retweets, Quotes High
Shares (especially via DM) Very High
Profile clicks → Follows Very High
Time spent reading (dwell time) Medium

What Kills Your Reach

Action Impact
"Not Interested" clicks Bad
Mutes Worse
Blocks Very Bad
Reports Worst

The Two Audiences

  1. Your followers (in-network): Your posts appear at full strength
  2. Everyone else (out-of-network): Your posts get a penalty multiplier

This is why building a loyal follower base matters - you're always fighting uphill with non-followers.


Quick Wins

  1. Optimize for shares - DM shares are weighted heavily. Create content people want to send to friends.

  2. Don't spam - The algorithm penalizes multiple posts from the same author. Quality > quantity.

  3. Post when your audience is online - Newer posts rank higher. Get early engagement.

  4. Avoid negative signals - A few blocks hurt more than many likes help.

  5. Make people curious about you - Profile clicks lead to follows, which is one of the strongest signals.


Go Deeper

Guide Who It's For
Growth Guide Anyone who wants actionable strategies to grow their account
Technical Breakdown Engineers who want to understand how the system works
Internals Deep dive into Kafka pipelines, ML inference, and data structures

What's in This Repo

x-algorithm/           # The actual algorithm source code
├── phoenix/           # ML ranking model (Grok-based transformer)
├── thunder/           # In-network post storage and retrieval
├── home-mixer/        # Feed assembly and scoring pipeline
└── candidate-pipeline/# Reusable recommendation framework

This is a mirror of X's open source release with our analysis docs added.


The Bottom Line

The algorithm is sophisticated, but the strategy is simple:

Create content that makes people want to engage, share, and follow you.

Everything else is implementation details.


Analysis by Dark Research

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published