The X Algorithm is Now Open Source

Your replies are worth
75x more
than your likes.

The algorithm doesn't care about your posting schedule.

We decoded the open-source X algorithm. Everything you've been told about growth is wrong. One reply chain = 150 likes worth of reach.

Join the waitlist β€’ No spam, ever

Weighted Scorer β€’ From Source Code
VERIFIED

// scored_tweets_model_weight β€” published April 5, 2023

reply_engaged_by_author75.0
reply13.5
good_profile_click12.0
good_click11.0
good_click_v210.0
retweet1.0
fav0.5
video_playback500.005
negative_feedback_v2-74.0
report-369.0

Bookmark is not in this formula. 22 signals are scored in ranking_scorer.rs β€” bookmark is not one of them. The word "bookmark" does not appear anywhere in the released code.

twitter/the-algorithm-ml β€’ xai-org/x-algorithmView full breakdown β†’
The Reality

Posting into the void?

Here is what happened to the person who built this, on his own account. You can go check every number.

Danks@danksterintelBuilt this

"I made this account in July 2025 and then basically sat on it. For five months almost nobody saw anything I posted. My weekly views were a flat line on the floor.

Then X put its ranking code on GitHub. I read it. Not somebody's summary of it, the actual files. Then I changed how I write posts to match what that code counts.

The flat line ended that same month. My best week went past 2.6 million views. The account sits at 31.9K followers. Same person, same topics, same face. The only thing that changed is that I stopped guessing."

This is the only testimonial on this page, and it is the founder's own account. Open the profile and check the join date and the follower count yourself. When customers have results, theirs go here too.

40-48%

Average engagement drop on X in 2025

0.029%

Median engagement rate (3 per 10K followers)

51%

Of full-time creators struggle to survive

The Solution

Stop guessing. Start knowing.

We built the first AI that actually understands the X algorithmβ€”because we read the source code.

Algorithm Score

See exactly how the algorithm will rank your content before you post. Real-time scoring based on actual algorithm weights.

reply_engaged_by_author75.0
good_profile_click12.0

Shadowban Detector

Know instantly if you're being throttled. Get specific diagnosis and a recovery planβ€”not just "wait it out."

report-369.0
negative_feedback_v2-74.0

Reply-Chain Optimizer

The highest-weighted signal in the algorithm: when your reply gets engaged by the author. We show you which replies to prioritize.

reply_engaged_by_author75.0
reply13.5
fav (like)0.5
The Process

How we cracked the code

X open-sourced their algorithm. We did what no one else bothered to do: actually read it.

01

We analyzed the Phoenix algorithm

X's recommendation system uses a Grok-based transformer that predicts engagement probabilities. We mapped every weight, every multiplier, every penalty.

02

We built a scoring engine

Our AI simulates how the algorithm will score your content before you post. No more guessingβ€”see your predicted reach in real-time.

03

We validate with real data

Every prediction is tested against actual X performance. Our model learns and improves with every post analyzed.

COMPLETE SCORING FORMULA

ranking_scorer.rs

// Final Score = Ξ£ (weight_i Γ— P(action_i)) β€” from xai-org/x-algorithm

Published Weights
twitter/the-algorithm-ml β€’ Heavy Ranker README β€’ April 5, 2023
75.0
reply_engaged
_by_author
13.5
reply
12.0
good_profile
_click
11.0
good_click
10.0
good_click_v2
1.0
retweet
0.5
fav
0.005
video
_playback50
-74.0
negative
_feedback_v2
-369.0
report
Additional Phoenix Signals
summed by ranking_scorer.rs β€’ weights not published
?
share
?
share_via_dm
?
share_via_copy_link
?
dwell
?
dwell_time
?
quote
?
quoted_click
?
photo_expand
?
follow_author

// params.rs is excluded from the release β€” the numeric weights used today are not public

vqv is the video score, and it can come out at exactly zero for some videos. The rule that decides when is not in the released code, so we will not put a number on it. Very short clips look like the risky case. β€” vm_ranker.rs:103

Not in the scoring formula

Bookmarks do not appear in ranking_scorer.rs. The scoring function sums exactly 22 signals β€” bookmark is not one of them. Searching the whole released codebase for the word "bookmark" returns nothing at all.

Bookmarking still helps X learn what you like, which shapes your own feed. It is simply not part of the formula that decides how far your post travels to everyone else.

[1]xai-org/x-algorithm/home-mixer/scorers/ranking_scorer.rs β€” 22 scored signals
[2]twitter/the-algorithm-ml/projects/home/recap/README.md β€” 10 published weights (April 5, 2023)
Early Access Pricing

Built for creators who are serious about growth

Waitlist members get 50% off for life. No commitment until we launch.

Creator

For individual creators

$29/mo
Waitlist: $15/mo
  • 100 Algorithm Scores/mo
  • Shadowban Detection
  • Basic Analytics
MOST POPULAR

Pro

For serious creators

$59/mo
Waitlist: $30/mo
  • Unlimited Algorithm Scores
  • Reply-Chain Optimizer
  • AI Content Assistant
  • Competitor Tracking (5)

Agency

For teams & agencies

$149/mo
Waitlist: $75/mo
  • Everything in Pro
  • 10 Team Members
  • White-label Reports
  • API Access

Ready to decode the algorithm?

Join the waitlist for early access and 50% off for life.

No credit card requiredCancel anytime50% off for life
FAQ

Frequently Asked Questions