X open sources its ranking algorithm and adds a shadowban transparency tool
X is releasing the source code for its For You recommendation feed on GitHub under the Apache v2 license, making the codebase roughly 10 to 15 times larger than its previous open source release. The company is also adding a transparency tool that lets users download a JSON file showing whether any labels were applied to their account or posts over the past calendar month.
The expansion covers the model configuration, the filter pipeline, and the core ranking system, including the parameters used to weight different signals that determine which posts appear in a user’s feed. Developers can also run parts of the ranker and scoring code outside X. The company framed the move as a response to long-running questions about how its algorithm shapes the spread of political content, misinformation, and elections on the platform.
What is being released?
The new GitHub repository contains the code that pulls posts, scores them, and assembles the For You timeline that opens when a user launches the app. It includes the configuration of the underlying models, the filters that screen rule-violating or potentially problematic content, and the core ranking engine that orders what each user sees.
According to the company, external researchers familiar with recommendation systems were given access ahead of launch and were able to train and run X’s Phoenix scoring system using the open source code. Researchers did not, however, get access to the per-post score. A subset of systems is intentionally excluded from the release, including the components that use Grok to predict whether a post could violate a rule. X said withholding those pieces is meant to keep bad actors from reverse engineering safeguards to flood the network with spam.
How users can check for shadowbans
X is rolling out a new section called Under the Hood inside the app’s settings. Users who have posted at least 10 times in the past month can download their aggregate stats as a JSON file, which surfaces any labels applied to their account or to individual posts over the previous calendar month.
Non-technical users can upload the file to a large language model of their choice, point it at X’s GitHub repository, and ask for an interpretation of what the labels mean. X is initially limiting the tool to a pilot group of accounts that are at least one year old, with a broader rollout planned after testing.
Developers can submit changes
Beyond reading the code, developers can file pull requests against the repository. X engineers will review submissions and decide which ones to incorporate into the live algorithm, though not every change will be accepted. The open effort mirrors the broader direction X has taken with Community Notes, its crowdsourced fact-checking system, by inviting outside participation in shaping the platform.
Why the transparency push now?
The release lands against a backdrop of years of scrutiny over how X’s algorithm ranks political content and whether it suppresses certain voices. Before its acquisition, Republicans in Congress alleged that the platform leaned too far left and claimed posts had been shadowbanned, a charge the company at the time consistently denied.
Since becoming private and later merging with SpaceX, X has been less forthcoming in other areas, including user metrics, growth figures, revenue, and government takedown requests, areas where public companies face disclosure requirements. Monthly active user numbers have since become public again.
FAQ
What did X open source?
X released the source code for its For You recommendation feed on GitHub under the Apache v2 license. The release includes the model configuration, filter logic, and core ranking engine, and is roughly 10 to 15 times larger than its previous open source contribution.
How can users see if they have been shadowbanned?
Users with at least 10 posts in the past month can open the new Under the Hood page in the app’s settings and download a JSON file showing any labels applied to their account or posts over the previous calendar month. The file can be uploaded to a large language model along with X’s GitHub repository for an interpretation.
Can developers change X’s algorithm?
Yes. Developers can submit pull requests to the GitHub repository, and X engineers will review them for possible inclusion in the live algorithm. X is also piloting the transparency tool with accounts at least one year old before opening it to a wider audience.
This article summarizes reporting from techcrunch.com.