Free Resource · Algorithm Intelligence

The X Algorithm Playbook

xAI publishes core code for X's For You feed. This guide separates what the repository documents from practical content experiments for founders and creators. Source facts first. Creator tactics clearly labeled.

Source: xai-org/x-algorithm · Reviewed September 27, 2026 · Some components are omitted; configuration changes

Download Free Playbook Read in browser

Share on X·No email · No paywall · Save as PDF

The Scoring Formula

A simplified view of ranking predictions.

Phoenix predicts viewer-specific actions and values. Ranking combines weighted predictions alongside other processing and filters. This diagram is explanatory, not a complete production scoring specification.

Illustration: Σ ( weighti × P(actioni) )

Weights scale predictions, not counts — there is no fixed report-to-like exchange rate

👤

Follow Author

positive

💬

positive

🔁

Quote Tweet

positive

📤

positive

❤️

Favorite / Like

positive

⏱️

Dwell Time

predicted value

🔍

Profile Click

positive

🖼️

Photo Expand

positive

⚠️

Report

negative

🚫

Block Author

negative

🔇

Mute Author

negative

👎

Not Interested

negative

What's Inside

Source notes, mechanics and experiments you can test.

How prediction scores work

Ranking combines weighted predictions for a viewer. Raw engagement counts are not a score or an exchange rate.

The 5 levers that build reach

Reply, Quote, Follow, Dwell, Share — creator experiments to test, not a universal ranking of actions.

The 4 signals that bury you

Predicted negative feedback can reduce ranking scores. One report does not cancel a fixed number of likes.

7-day execution plan

A practical writing and measurement routine. Posting frequency and timing are hypotheses to test with your audience.

Five creator experiments

A practical starting point.

Write for useful conversation

Ask a specific question when it helps the reader. Reply counts cannot be converted into a guaranteed ranking advantage.

Give readers a reason to return

Make your subject and recurring value clear. Following is a business outcome to measure, not a guaranteed feed multiplier.

Make attention worthwhile

Use readable text, useful visuals and clear structure. There is no established universal dwell-time target.

Test a sustainable cadence

Try one or two useful posts per day as a starting experiment, not an algorithmic maximum.

Measure over several posts

Compare audience response across formats and times. The public code does not establish a universal first-15-minute cutoff.

Free. No email. No paywall.

Open it in your browser and hit “Save as PDF.” Share it with anyone building a brand on X.

Get the Playbook Share on X

// More free resources at jeremyknox.ai

Browse all resources

Explore the Tesseract Labs Ecosystem

Tesseract Intelligence →Competitive intelligence and strategic foresight.Rewired Minds →Psychology and the hidden mechanics of high performance.Blueprint →AI-powered idea execution.Clarity →A council of AI agents debates your career and financial decisions.