X ranking algorithm open source release adds account-label transparency tool
X says it has released more of its For You ranking code and will show eligible users labels applied to posts or accounts.
By Jordan Bell · Startups & Deals Reporter
· 3 min read
X ranking algorithm open source efforts are expanding, with the company saying it will publish more code behind its For You feed and introduce a feature that shows eligible users labels applied to their posts or accounts. For investors watching platform companies, the move offers a clearer view of how X says it sorts content, though the new user tool does not deliver a definitive finding that someone was “shadowbanned.”
According to TechCrunch, X is making the source code for its For You timeline available through a GitHub repository under the Apache v2 license. The For You timeline is the default feed users see on opening the app.
X said the broader release includes its model configuration, filters, core ranking details and the settings that determine how heavily different signals affect a post’s placement. TechCrunch reported that the codebase is roughly 10 to 15 times the size of X’s earlier release.
Can X users see whether ranking systems affected their posts?
X is rolling out an in-app settings page called Under the Hood, TechCrunch reported. Users who made at least 10 posts during the prior month will be able to download aggregate statistics in JSON, a structured data-file format.
That file is expected to list labels applied to the user’s account or posts during the preceding calendar month. The information may help explain whether X’s ranking systems affected distribution, but X has not described the download as a standalone verdict that an account was shadowbanned.
Access will begin with a pilot group of accounts that are at least one year old, before a wider release, according to TechCrunch. That means the feature will not initially be available to every account that meets the posting threshold.
What is included in X’s code release?
Keith Coleman, X’s vice president of product, told TechCrunch that the release contains code used to retrieve posts, rank them for a particular user and build the feed. He said it also includes systems that screen potentially problematic or rule-breaking content.
Coleman said some ranking and scoring systems can be run outside X. TechCrunch also reported that external researchers who previewed the code were able to run the numerical score calculated for each post outside the company.
Developers may submit proposed code changes, known as pull requests, through the GitHub repository. X engineers will decide whether any submissions become part of the company’s algorithm. TechCrunch reported that some systems remain outside this release, without identifying them in the available report.
The announcement provides more material for researchers and developers to examine the mechanics behind X’s recommendations. For ordinary users, the near-term practical feature is narrower: qualifying pilot participants can review a recent history of applied labels, rather than receive a complete explanation for every shift in reach.
This story draws on original reporting from TechCrunch.
