X is making a much larger portion of its recommendation technology available to the public, while giving some users a new way to check whether the platform’s ranking systems have limited the visibility of their accounts or posts.
The social network is releasing the source code behind its “For You” feed under the Apache 2.0 open-source license. The expanded repository covers important parts of the recommendation pipeline, including ranking logic, filters, model configurations and the weighting applied to signals used when deciding which posts appear in a user’s feed.
According to X, the newly published codebase is roughly 10 to 15 times larger than its previous open-source release. Some of the ranking components can also be run independently by researchers and developers, potentially making it easier to study how the platform scores and distributes content rather than relying solely on X’s explanations of the process.
That additional visibility is important because recommendation algorithms effectively determine what much of a social network’s audience sees. Publishing code does not automatically make every ranking decision understandable, but access to signal weights, filters and scoring systems gives outside researchers considerably more material to examine.
X is pairing the code release with a user-facing transparency feature called “Under the Hood.” Eligible users will be able to download a JSON file containing aggregated information about labels that X’s systems applied to their account or posts during the previous calendar month.
The initial pilot is limited to a test group of accounts that are at least one year old. Users must also have published at least 10 posts during the previous month. X plans to expand availability later.
The feature could provide a more concrete answer to a question that has followed Twitter and X for years: whether an account has effectively been “shadowbanned,” a loosely defined term generally used when content becomes less visible without an obvious suspension or removal.
The raw JSON data will naturally be more useful to technically inclined users. X suggests people could also provide the file and its GitHub repository to a large language model and ask the AI to interpret how the ranking system affected their posts.
There are limits to the transparency push. X is withholding some systems that could reveal how the platform identifies potentially rule-breaking content. That includes certain tools involving Grok. The company argues that publishing those details could allow spammers and other bad actors to learn how to evade enforcement.
Developers can submit proposed changes to the open-source project through GitHub pull requests, although X retains control over whether any outside contributions become part of its production systems.
The release arrives amid longstanding scrutiny of how social media algorithms influence political discussion, misinformation and content visibility. Those arguments predate Elon Musk’s acquisition of Twitter, with the platform facing years of accusations from politicians and users who believed its ranking and moderation systems unfairly suppressed particular viewpoints.
Opening more of the recommendation engine could make those debates easier to test against actual code. It does not resolve X’s wider transparency issues, however. Since becoming a private company, the platform no longer provides the same level of public financial and operational reporting associated with its years as a listed business, while information covering areas such as user numbers and government requests has become less consistently available.
Algorithmic transparency is therefore only one piece of a much larger accountability question. X is giving researchers and users an unusually detailed look at how posts move through its recommendation machinery; what outsiders discover once they start examining that machinery may prove more significant than the code release itself.
