X Opens More of Its Ranking Algorithm Code to The Public

X is expanding access to the source code behind its recommendation and ranking systems, giving developers and researchers more visibility into how posts are selected for users and introducing a tool designed to show whether accounts or posts have been affected by ranking interventions.

The company is making the code for its “For You” timeline available on GitHub under the Apache 2.0 license. The release also includes additional details on model configurations, filters and the core ranking system, including parameters used to weight different signals when determining which posts appear in users’ feeds.

According to X, the expanded codebase is roughly 10 to 15 times larger than its previous open-source release. Some components, including parts of the ranking system, can also be run outside the company, potentially allowing researchers and developers to examine how the systems work independently.

For users, the more notable change is a new transparency feature being introduced through X’s “Under the Hood” settings. During an initial pilot, eligible accounts that have posted at least 10 times during the previous month will be able to download aggregate account data as a JSON file. The information will indicate whether labels have been applied to their accounts or posts during the previous month.

The tool is intended to provide greater clarity around claims of “shadowbanning,” a term used to describe posts or accounts being made less visible without the user’s knowledge. X has said the new data should allow users to examine whether its ranking systems have affected distribution.

The company is also inviting developers to submit changes to the open-source code through GitHub, although X engineers will decide which contributions, if any, are incorporated into the platform.

Not every system is being disclosed. X is withholding certain tools, including those using Grok to identify potentially rule-breaking content, citing concerns that detailed disclosure could help bad actors evade moderation systems.