The X ranking algorithm has been made open source alongside the launch of new transparency tools, according to a report by TechCrunch. The social network announced these updates on Thursday, allowing users to see how the platform’s ranking systems affect their accounts and posts.
Details of the X Ranking Algorithm Release
The company is making the source code for the “For You” timeline available on GitHub under the Apache v2 license. This timeline serves as the default feed when users open the social media application. By expanding its open-source codebase, the company has made the X ranking algorithm roughly 10 to 15 times larger on GitHub than its previous releases.
The release includes model configurations, filters, and core ranking system details. It also details the parameters used to weight different signals, which determine which posts are displayed to users. Keith Coleman, VP of Product at X, stated that developers can now access the core ranking code that pulls and ranks posts for any given user.
New Transparency Tools for Users
In addition to the code repository, X is introducing tools to help users determine if ranking systems have impacted their accounts. A new transparency feature is rolling out to an “Under the Hood” page within the apps settings menu. This tool allows users who have posted 10 or more times over the past month to download their aggregate statistics as a JSON file.
The downloaded file shows if any labels have been applied to the account or posts over the past calendar month. Non-technical users can analyze this information by uploading the file to a large language model and asking for an interpretation based on the GitHub repository. The tool is initially available to a test group of accounts that are at least one year old.
Code Exclusions and Security Measures
Certain systems are excluded from this open-source release to maintain platform security. Specifically, systems that use the Grok model to predict whether a post violates platform rules remain private. This exclusion prevents bad actors from using the code to bypass rules or distribute spam across the network.
External researchers familiar with recommendation systems previewed the codebase prior to the official launch. These researchers successfully trained and ran X’s Phoenix scoring system using the open-source code. Developers can now submit pull requests on GitHub, which X engineers will review for potential integration.
Addressing Historical Platform Concerns
These changes aim to address ongoing concerns regarding how the platform’s systems influence public discussions and elections. Historically, users and political figures have raised concerns about “shadowbanning,” where posts are made invisible without the user’s knowledge. This open approach to the X ranking algorithm aims to build trust among the platform’s global user base.
While X is increasing transparency around its code, its overall corporate transparency has changed since becoming a private company. The platform is no longer required to report regular financial metrics or user growth to regulatory bodies. However, following its merger with SpaceX, the company has started sharing its monthly active user numbers publicly once again.
