X Releases For You Ranking Code and Tests Visibility Labels
X has released the source code for its default For You timeline and core ranking engine on GitHub under the Apache v2 licence. The social network also announced a pilot transparency feature designed to show users whether its ranking systems have applied labels to their accounts or posts.
The release expands X’s previously open-sourced materials with model configuration, filtering information and core ranking-system details, including parameters used to weight different signals. X says the newly available codebase is roughly 10 to 15 times larger than the earlier release.
More of the feed assembly process is available
Keith Coleman, X’s vice-president of product, said the repository includes the core code used to retrieve posts, rank them for an individual user and assemble a feed. It also exposes systems that filter potentially problematic or rule-violating material, while some ranking and scoring components can be run outside the company.
External researchers familiar with recommendation systems previewed the code before launch and were able to run the score, the numerical value calculated for every post, outside X. Developers can also submit pull requests to the repository. X engineers will consider proposed updates, but the company has not said that all contributions will be accepted.
Account-level information arrives in a limited pilot
The new tool will appear on an Under the Hood page in the app’s settings. Users who posted at least 10 times during the previous month will be able to download aggregate statistics in a JSON file. That file will indicate whether labels were applied to their account or posts during the past calendar month.
X says the feature will initially be available to a test group of accounts that are at least one year old, before a broader rollout. The company suggests that non-technical users could provide the exported file and the GitHub repository to a large language model for interpretation.
Transparency has defined limits
Not every relevant system is included. X is withholding tools that use Grok to predict whether a post may violate a rule, saying that publishing those details could enable bad actors to circumvent enforcement and send spam through the network.
The initiative responds to long-running debate over how ranking affects political discussion, elections and misinformation, as well as claims that posts may have been made less discoverable without users’ knowledge. X says public access will allow people to audit and critique post distribution and assess whether the platform operates fairly.
For organisations that depend on X for audience reach, the practical implication is to retain the exported label data, compare it with distribution changes and recognise that the published repository does not cover every moderation or ranking mechanism.