What Was Shipped

The new GitHub repository (xai-org/x-algorithm) is roughly 10–15 times larger than X's previous open-source efforts, including model configuration, filters, and the weighting parameters that determine which posts appear in feeds . The source code for the default "For You" timeline is now public under the permissive Apache v2 license, allowing developers and researchers to inspect how posts are pulled, ranked, and assembled into a personalized feed .

The "Under the Hood" transparency tool appears as a new settings page. Users who have posted at least 10 times in the past month can download aggregate statistics as a JSON file detailing any account labels that limited visibility over the prior calendar month . According to Coleman, users can then drop the JSON into an LLM to interpret how the algorithm treated them . The feature initially rolled out as a pilot to accounts that are at least one year old, with plans for broader availability .

Developers can also submit pull requests to the public GitHub repository, study the ranking mechanics, and run parts of the ranker themselves .

What Was Excluded (and Why)

The release notably omits the Grok-based moderation systems that detect and filter problematic content. Coleman indicated that these systems were intentionally held back to prevent bad actors from reverse-engineering them to evade detection .

Broader Context and Rationale

The move directly responds to years of user and regulator complaints about opaque shadowbanning, alleged political bias in content curation, and potential influence on elections . By making the ranking logic auditable and providing per-account visibility data, X aims to demonstrate that its systems apply labels based on rule violations rather than political viewpoint. Coleman told TechCrunch the initiative was intended to build trust through radical transparency, saying the company is releasing things "people will be fairly shocked that we are releasing" . He positioned the tool as a way for users to independently verify whether they have been impacted by ranking systems, rather than relying solely on X's public statements .

What Was Shipped

The new GitHub repository (xai-org/x-algorithm) is roughly 10–15 times larger than X's previous open-source efforts, including model configuration, filters, and the weighting parameters that determine which posts appear in feeds . The source code for the default "For You" timeline is now public under the permissive Apache v2 license, allowing developers and researchers to inspect how posts are pulled, ranked, and assembled into a personalized feed .

The "Under the Hood" transparency tool appears as a new settings page. Users who have posted at least 10 times in the past month can download aggregate statistics as a JSON file detailing any account labels that limited visibility over the prior calendar month . According to Coleman, users can then drop the JSON into an LLM to interpret how the algorithm treated them . The feature initially rolled out as a pilot to accounts that are at least one year old, with plans for broader availability .

Developers can also submit pull requests to the public GitHub repository, study the ranking mechanics, and run parts of the ranker themselves .

What Was Excluded (and Why)

The release notably omits the Grok-based moderation systems that detect and filter problematic content. Coleman indicated that these systems were intentionally held back to prevent bad actors from reverse-engineering them to evade detection .

Broader Context and Rationale

The move directly responds to years of user and regulator complaints about opaque shadowbanning, alleged political bias in content curation, and potential influence on elections . By making the ranking logic auditable and providing per-account visibility data, X aims to demonstrate that its systems apply labels based on rule violations rather than political viewpoint. Coleman told TechCrunch the initiative was intended to build trust through radical transparency, saying the company is releasing things "people will be fairly shocked that we are releasing" . He positioned the tool as a way for users to independently verify whether they have been impacted by ranking systems, rather than relying solely on X's public statements .