X opens algorithm to public that determines how posts are ranked, filtered

X is opening more of the code behind its 'For You' feed and adding a tool that lets eligible users check whether labels on their accounts or posts could be limiting their visibility

Elon Musk-owned X, formerly Twitter, is making more of its recommendation system public, giving users a closer look at how posts are selected and their visibility determined. The company has open-sourced code used to rank and filter posts on its ‘For You’ timeline and introduced a tool that lets eligible users check whether labels on their accounts or posts could limit their reach.

 

The changes are aimed at giving users greater clarity on why certain posts appear in their feeds and whether their content is subject to visibility restrictions.

For You algorithm: Details

X is releasing the code that determines how posts are ranked and filtered in the For You timeline. The release expands on the company's earlier open-source work and includes details about model configuration, filters and the core ranking system.

 

 

For context, the For You timeline is X's default feed that appears when users open the app. Instead of showing posts only in chronological order, X's recommendation system selects, ranks and filters posts based on signals such as what it thinks users may be interested in.

 

According to TechCrunch, the expanded codebase is around 10 to 15 times larger than X's previous release. It also includes parameters used to weigh different signals, offering more information about how the system determines which posts appear in users' feeds.

 

X also allowed external researchers specialising in recommendation systems to review the code before its public release. 

 

Users can check visibility labels

 

X is also introducing a new Under the Hood page that will provide information about labels that may affect the visibility of an account or its posts.

 

Users who have posted at least 10 times during the previous month will be able to check labels applied to their accounts and posts during that period. They will also be able to download the information.

 

The feature is being introduced as a pilot. Initially, it will be available to a randomly selected group of eligible accounts that are at least one year old. X said it plans to use feedback from the pilot to determine how the feature should be expanded.

 

X, however, is not releasing every part of its recommendation and moderation systems. According to TechCrunch, systems that use Grok to predict whether a post could violate X's rules are not part of the release. The company said keeping these systems private is intended to prevent bad actors from using the information to bypass its rules or create spam.

 

X said the open-source code and transparency tool are designed to work together. The code shows how the For You feed is generated and which labels can affect visibility, while the tool lets users check whether those labels have been applied to their accounts or posts.

 

X plans regular updates

 

X said the system has been designed to support updates as its recommendation algorithm changes. The company plans to make further updates to the open-source code over time.

 

The announcement comes amid wider questions about how social media recommendation systems influence what users see and how much reach individual posts receive. X's changes are also relevant to long-running concerns around the platform's algorithm, including claims of political bias and shadowbanning.

 

X said it wants the new tools to give users and outside researchers more information about how its recommendation system works, while allowing the public to examine parts of the technology behind the For You feed.