X Opens More of Its Ranking Algorithm and Adds Shadowban Transparency Tools

X is significantly expanding its open source codebase, including the app’s “For You” algorithm and core ranking engine. The source code for the default For You timeline is being made available on GitHub under the Apache v2 license.

The release adds more detail than prior open source efforts, including model configuration, filters, core ranking system details, and parameters used to weight different signals. TechCrunch reports the expanded codebase is roughly 10 to 15 times larger than before.

X is also adding a transparency tool designed to show whether its ranking systems have affected a user’s account or posts. The feature is rolling out to an “Under the Hood” page in the app’s settings.

Users who have posted 10 or more times over the past month will be able to download aggregate stats as a JSON file. That file will show whether labels were applied to their account or posts over the past calendar month, with the tool initially piloted among a test group of accounts at least a year old.

X previewed the open source codebase to external researchers familiar with recommendation systems, who were able to train and run X’s Phoenix scoring system using the released code. Developers will also be able to submit pull requests for X engineers to consider.

The release does not include every system. TechCrunch reports that some tools, including systems that use Grok to predict whether a post could violate rules, are excluded to reduce the risk of bad actors using the information to evade rules or flood the network with spam.

The move directly addresses long-running concerns about how X’s algorithm distributes posts and whether users are being quietly limited in reach. By making more ranking logic visible and giving users access to account-level label data, X is creating more ways for the public to scrutinize feed decisions.

The practical takeaway: technical users can inspect the repository, while non-technical users may be able to use the downloaded JSON data to better understand whether ranking labels affected them. The broader impact will depend on how widely the tools roll out and how X responds to critiques, audits, and developer contributions.