In its most aggressive transparency push to date, X is open-sourcing the core code that powers its “For You” feed, publishing the full ranking engine and model configurations on GitHub. At the same time, it is rolling out a tool that will let users see exactly how its systems have impacted their account’s reach, effectively giving them a way to check if they have been “shadowbanned.”
This is a major departure from the playbook of other social networks, which guard their proprietary algorithms like state secrets. By making the codebase available under an Apache v2 license, X is handing competitors, researchers, and hobbyists the keys to its kingdom.
However, while this move offers unprecedented technical transparency, the company’s status as a private entity means it is still a black box in other critical areas, leaving a complex picture of what “transparency” actually means in the Musk era.
A New Standard for Algorithmic Transparency
The social network announced on Thursday that its GitHub repository has expanded significantly to include the “For You” timeline and its core ranking engine. According to X’s VP of Product Keith Coleman, this release makes the codebase “roughly 10 to 15 times larger than it was before.”
Previously, X had open-sourced some recommendations code, but this new release dives much deeper. It includes the model configuration, filters, and the core ranking parameters used to weight different signals. In simple terms, you can now see the formulas that decide which posts you see and which you do not.
“We’re releasing the code for the core ranking that pulls posts and ranks them for any given user and assembles the feed,” Coleman told TechCrunch. The company is even providing tools that allow developers to run the ranker and score systems “outside the company,” meaning external parties can replicate the logic that powers X’s primary feed.
This move effectively throws down the gauntlet to competitors like Meta and TikTok, which have historically been opaque about how their algorithms work. For years, critics have argued that algorithmic opacity allows platforms to manipulate public discourse with impunity. X is now offering the data to prove—or disprove—those concerns.
The “Under the Hood” Tool: Checking Your Shadowban Status
While developers will be diving into the code on GitHub, the average X user gets a new tool that could be equally impactful. A new “Under the Hood” page is rolling out in the app’s settings.
This feature allows users who have posted at least ten times in the past month to download their aggregate stats as a JSON file. The file reveals any labels that have been applied to their account or posts by X’s ranking systems. This effectively provides a technical answer to the persistent question of whether an account has been shadowbanned.
Crucially, X is acknowledging that the data is complex. For non-technical users, the solution is to drop the JSON file into an LLM (like ChatGPT or Grok) and ask it to interpret the results based on X’s GitHub repo.
This is a unique approach to UX design—wielding AI as a translation layer for technical data. It suggests that X is prioritizing radical data access over building a polished user dashboard, betting that generative AI will bridge the gap for the average user.
The tool is currently rolling out to a test group of accounts that are at least a year old before a wider launch.
True Transparency vs. Private Control
Here is where the narrative gets complicated. The company is making a massive bet that algorithmic transparency is its path back to public trust. Yet, it is doing so while being arguably less transparent than it was as a public company.
This creates a paradox that is central to understanding X’s current strategy: The company is willing to reveal the exact mathematical logic of its product, but refuses to reveal the basic operational metrics of its business.
X no longer reports user metrics, revenue, or even government takedown requests with the frequency it did when it was a public company. It operates as a private entity with less regulatory oversight.
By focusing exclusively on algorithmic transparency, X creates a “controlled narrative.” The debate about the platform’s fairness becomes a technical debate about code parameters rather than a political debate about content moderation volume or business pressures. It shifts the accountability from the CEO’s decision-making (which remains opaque) to the engineering team’s logic (which is now open).
This is a brilliant deflection mechanism. If users complain about reach, X can simply point to the code and the “Under the Hood” data. If the data shows no labels, the implication is that the user’s content is simply unpopular. The burden of proof shifts from the company to the user.
Why This Matters for the Free Speech Debate
This is the most significant development for the “shadowbanning” debate since it first became a mainstream political talking point.
For years, critics on the political right claimed that social media algorithms intentionally throttled their content. X’s new tools provide a mechanism for these claims to be tested empirically. If a user downloads their stats and sees a heavy label being applied, they have evidence. If they see no label, they have to accept that their reach is dictated by engagement, not political bias.
This is the ultimate “put your money where your mouth is” moment for the platform. The company is essentially daring its critics to audit the code and prove the system is unfair.
The Developer Community’s Role
X is also opening the door to external contribution. Developers will be able to submit “pull requests”—essentially suggested code changes—to X’s algorithm.
“Imagine if the X algorithm is not just visible to the public, but is built by the public,” Coleman mused.
This is an ambitious vision that borrows from the open-source ethos of Linux. However, it remains to be seen whether X will actually merge external changes. The company has final veto power, and one has to wonder how eager a trillionaire-owned company is to accept code from anonymous developers on the internet. Still, even the offer to consider suggestions provides a veneer of community control.
What is Still Hidden?
Not everything is on the table. X is keeping proprietary certain safety systems, specifically those that use Grok AI to predict if a post might break the rules.
“The reason those are not included is that those include some of the details about some of the automated enforcement that would be useful for a spammer, for instance,” Coleman explained.
This is a legitimate security concern. Publishing the exact logic used to catch spammers would allow bad actors to bypass the defenses. It is a necessary limit to transparency—one must protect the system from gaming, or the algorithm becomes useless.
X has redefined the transparency landscape. By open-sourcing its ranking algorithm and offering a shadowban audit tool, it has raised the bar for how social media platforms are held accountable. The company has moved the discussion from vague accusations of bias to technical scrutiny of code.
However, the broader implication is a warning: transparency on one front does not mean transparency on all fronts. X is choosing a very specific, technical version of transparency that enhances its public image while leaving its core business practices hidden.
For users, the message is clear: The algorithm is no longer a secret. Whether you are a politician, a journalist, or a casual poster, you can now check the code and your stats to understand why you are being seen—or ignored.
What happens next depends on the community. If developers and researchers take advantage of this repository, we may see significant improvements—or exposure of flaws—in the X algorithm. But if the tool is ignored, this will remain a powerful symbol of transparency rather than a functional shift in power.
One thing is certain: The era of the black-box social media algorithm is ending, at least for X. The question is whether other platforms will be forced to follow.
