X Open-Sources Core Ranking Code and Adds Account Visibility Tracking Tools

Social network X announced Thursday that it is publishing an expanded version of its core recommendation algorithms to GitHub while launching user-facing transparency features designed to reveal how platform systems restrict or boost account visibility, first reported by TechCrunch.

The platform released the underlying source code for its default "For You" timeline under the Apache v2 license. The repository contains model configurations, content filtering rules, and ranking parameters that assign weight to various user interaction signals. Company officials stated that the newly published code footprint is approximately 10 to 15 times larger than X's previous open-source releases.

In an interview with TechCrunch prior to the rollout, X Vice President of Product Keith Coleman said the release gives external observers access to the foundational pipeline responsible for selecting, scoring, and organizing posts for individual user feeds. Coleman noted that specific components, including the internal ranking and scoring infrastructure, can now be executed independently outside the company's internal infrastructure.

Alongside the code repository, X is testing an in-app transparency feature within an "Under the Hood" section of user settings. Initially restricted to a pilot group of accounts that are at least one year old and have generated 10 or more posts over the past month, the tool allows individuals to download a JSON file detailing any account-level or post-level labels applied during the preceding calendar month. Users can supply this data file alongside the GitHub repository to third-party large language models to evaluate whether their reach has been impacted by platform algorithms.

Prior to the public launch, X granted access to external recommendation system researchers, who successfully trained and executed the platform's Phoenix scoring system using the open-source repository. X is also enabling public software developers to submit pull requests on GitHub, allowing company engineers to review community-submitted modifications for potential inclusion in the production algorithm.

X deliberately excluded certain systems from the open-source release, including models that leverage the Grok artificial intelligence system to identify potential rule violations. Company representatives explained that withholding these automated moderation mechanisms prevents bad actors and spammers from studying platform defenses to bypass enforcement protocols.

The updates aim to address long-standing disputes regarding how algorithmic feed curation impacts political discourse, election integrity, and content moderation on the network. Conservative figures and lawmakers previously alleged that pre-acquisition Twitter engaged in covert reach reduction against right-leaning accounts, assertions that the social network repeatedly rejected.

The code release marks a shift in disclosure practices for X, which reduced public reporting regarding operational metrics, revenue, and government takedown requests following its acquisition by Elon Musk and transition to a private entity. The company, which recently merged with SpaceX, has since resumed disclosing monthly active user statistics as part of its updated reporting.