X Censorship Exposed: 'The Parzival' Reveals Code Employed for Account Censorship Tied to Former Twitter CEO/CTO Parag Agrawal
"This function generates a TSV File for User Mass Scores which was used to censor conservative Americans and dates back as far as 2009."
One user on the social media site 𝕏, formerly known as Twitter, recently linked a line of code from the social media site’s algorithm directly to ex-CEO/CTO of Twitter, Parag Agrawal. Agrawal was ousted along with thousands of other employees of Twitter when Elon Musk purchased the company for $44 billion.
𝕏 user The Parzival posted on Tuesday, “BREAKING: Special Coding from the Twitter algorithm called ‘Twadoop’ has now been definitively linked to .@paraga, Ex CEO/CTO of Twitter. This function generates a TSV File for User Mass Scores which was used to censor conservative Americans and dates back as far as 2009.”
User Eduard Brichuk added context to The Parzival’s post saying, “The ‘doop’ is the main root of the word and prefixes have its own meaning to the structure. Look at the Apache forum where Hadoop and twadoop is mentioned.” Brichuk added a link showing a data analytics structure used on Twitter published in 2012 by Bill Graham, then-Data Systems Engineer, Analytics Infrastructure.
Parzival posted a subsequent thread the following day detailing exactly what “Twadoop” is and how it is deployed on certain accounts on 𝕏:
“To begin let's first look at what a User Mass Score means and how it affects you. The User Mass is a Sub-Algorithm that assigns a Score to the account based on a variety of characteristics, many of which the values are hidden. The characteristics and code are shown here.”
The Parzival continues, “The User Mass is part of the ‘Tweepcred’ group of Scoring Metrics used to determine an account’s visibility on the platform. The User Mass has a TSV File that is generated by the Twadoop function. A TSV File is simply a plain text data table that is native to many applications.”
“This TSV File acted as a pseudo list of each account’s Social Credit Score on the platform. This Score could then be easily manipulated by changing values in the tables of the files manually or adding a weight multiplier to scores.”
”Here is where we begin to connect previous employees that had a hand in the development of Twadoop, Tweepcred, and this Social Media Social Credit System. There is a correspondence regarding Twadoop Files between Parag and Kevin Weil in 2012, both very high-level employees.”
Former Twitter Product Head Kevin Weil
Continuing in Wednesday’s thread, The Parzival explains who Kevin Weil is, and how he fits into this equation:
“So who is Kevin Weil? He was Twitter's Product Head. He later went on in 2016 to join Facebook's Instagram. I'm very curious to know what exactly was in the files Weil was asking Parag about? This post from Parag just so happens to be deleted now.”
“Now remember this interaction between them was in 2012, but Twadoop actually goes back even further, all the way to 2009. We can see references to Twadoop from another Ex Twitter Employee that ‘Helped Create the Data Platform Twitter’. Notice he said Data not Social Media!”
Parzival continues, “It should be clear that Twadoop has been around for quite some time specifically since 2009. Now let's put some focus back on Parag who was there since 2011 through his reign as CEO to 2022 when Elon Musk took over and fired him.”
Former CEO/CTO Parag Agrawal
Agrawal studied computer science at the Indian Institute of Technology as well as Stanford University before joining the social media site where he worked in advertising, later being named as chief technology officer of Twitter in 2017.
Parag was a major proponent of blockchain and cryptocurrency and was heavily involved in Twitter’s Project Bluesky initiative which sought to develop a decentralized social network protocol that would eventually see several platforms utilized.
Agrawal was announced in November 2021 as Twitter’s new CEO, when Dorsey unexpectedly resigned as CEO of the social media platform. Dorsey sent an email to the then-Twitter team where he announced Agrawal as his replacement which read in part, “The board ran a rigorous process considering all options and unanimously appointed Parag. He’s been my choice for some time given how deeply he understands the company and its needs. Parag has been behind every critical decision that helped turn this company around. He’s curious, probing, rational, creative, demanding, self-aware, and humble. He leads with heart and soul and is someone I learn from daily. My trust in him as our CEO is bone deep.”
The Parzival also highlighted in his thread on Wednesday that Agrawal had made a statement during an interview with MIT Technology Review in November of 2020, while he was still Twitter’s CTO that when it comes to protecting freedom of speech on the platform, “Our role is not to be bound by the First Amendment, but our role is to serve a healthy public conversation ... [and to] focus less on thinking about free speech, but thinking about how the times have changed."
Wednesday’s thread concludes with some of Parzival’s thoughts on the former statement, “It's very disturbing that someone can come from another country, obtain US Citizenship, reap the benefits of America, then go on to destroy the Constitution, the very thing that makes America great. My God-given rights are not debatable and they will never take our freedom!”
Twitter Releases Open-Source Algorithm
Parzival began tracking different lines of code found within the site’s algorithm after CEO Elon Musk decided to “open-source” Twitter’s algorithm on GitHub. Musk intended for the public to sift through the millions of lines of code used on the platform, and allow the community to submit their own code for the platform’s consideration. Twitter’s algorithm determines how users’ posts are filtered and ranked for different timelines on the site.
The Parzival’s Twitter Algo Analysis
Shortly after the algorithm was open-sourced, Twitter user The Parzival began examining the lines of code and found several troubling aspects that were previously unavailable to the general public. In a series of Twitter threads beginning in May of 2023, The Parzival revealed what was written in the code as the “Tombstone Generator”. The Tombstone Generator was often found directly under lines of code that determined an account’s visibility. “If you have ever been censored on Twitter or other platforms you will want to read this. This Thread will Expose the ALGORITHM LIKELY USED BY TWITTER FOR MASS CENSORSHIP. Let me introduce you to the Tombstone Generator!”
The lines of code continue and get more detailed as the algorithm is laid out, and The Parzival tells users they can follow the path to find the code for themselves using: the-algorithm/ visibilitylib / src / main / scala / com / twitter / visibility / generators / TombstoneGenerator.scala.
When The Parzival asked ChatGPT how the Tombstone Generator code was used to censor users on Twitter, the Open AI chatbot stated:
“…the ‘TombstoneGenerator’ algorithm is censoring users by hiding or removing their tweets based on a set of predefined rules. The algorithm receives a ‘VisbilityResult’ object that contains information about a tweet, such as its language and country codes, and returns a modified ‘VisibilityResult’ object with the appropriate ‘Tombstone’ object based on the tweet’s content and context.
“The ‘Tombstone’ object is a data structure that represents the fact that a tweet has been hidden or deleted. The ‘TombstoneGenerator’ algorithm applies various rules to determine whether a tweet should be tombstoned, and if so, which ‘Epitaph’ to apply, which in turn determines the reason for tombstoning the tweet.”
ChatGPT gives an example of how the code is applied and then summarizes its use, “Overall, the ‘TombstoneGenerator’ algorithm is designed to remove or hide tweets that violate specific rules, such as containing inappropriate content or violating local laws.”
The Parzival followed the May 10th thread with another thread on May 15, 2023, showing how Twitter uses algorithms to both rank, boost, and deboost posts. This included a line of code containing the term “OutOfNetworkReplyPenalty” that The Parzival says, “makes you VIRTUALLY INVISIBLE even with your Blue Checkmark. It also keeps Retweets from getting any traction, even if Retweeted by large accounts of 100k+ Followers.” The Parzival says this particular line of code is shown to weigh 10x more than the boost Twitter Blue assigns.
“FreedomOfSpeechNotReach” is another Twitter code that Elon Musk posted that moderates certain content on the platform. Parzival also studied this code and found this was a newly added feature in April 2023 used to moderate the visibility of users and their content.
A temporary remedy was recommended by The Parzival that would increase the boost given to Twitter Blue accounts he says, “should be an incredibly easy modification that can be made to increase this BOOST FACTOR. I recommend adjusting the BOOST FACTOR for Verified Twitter Blue Users to be SET TO 10. This at the very least would counter the Out of Network Tweet Deboost making it much more BALANCED. This could be adjusted further based on the results it produces.”
Parzival has continued his studying of the Twitter open-source algorithm sets, and has discovered code that essentially blacklists users based on a reputation score called “Tweepcred”.
Moving into June 2023, one term, defined as “breadcrumb” within the “StormAggregate” code has counters that track the activity of accounts deemed “Right” or, “Left”.
Parzival describes how this is applied on Twitter accounts saying, “As you post Right leaning material ‘AuthorRight’ and interact with others Right leaning material you get counters. These counters then become part of the users stats and are then subsequently used when importing the ‘StatsReceiver’ for future tweets or retweets.”
In July, Parzival suggested Twitter display all users’ Mass Score privately on their profile, essentially giving the user an idea of where they rank based on the site’s social credit score.
Also in July, The Parzival showed evidence of how certain accounts can be deboosted for interacting with other accounts they’ve deemed “low-quality” or that have been flagged previously with another label.
“BREAKING: TWITTER ALGO HAS CODING TO PENALIZE CERTAIN GROUPS, KNOWN AS CLUSTERS, WHERE CLUSTERS GET PENALIZED BASED ON LOTS OF INTERESTS AND POPULAR USERS! CONSERVATIVE CLUSTERS ARE BEING PENALIZED, WHICH LIMITS VISIBILITY FOR ALL THAT ARE IN IT!!”
Yet another massive thread from Parzival was posted on November 12th, 2023, where he addresses some of the details laid out during the February 2023 House Oversight Committee testimony where Twitter 1.0 executives were questioned about the response to the Hunter Biden laptop story and the censorship applied to conservative accounts on the platform. It’s important to note these execs (including Yoel Roth, Vijaya Gadde, Anika Collier Navaroli, and former Twitter Deputy General, James “Jim” Baker) claimed there was “no conspiracy” to silence the story itself, nor to muzzle conservative voices speaking out on the subject by those at Twitter 1.0. The full hearing can be viewed here.
Parzival states, “This thread will serve the purpose of providing a mountain of evidence to @realannapaulina (Anna Pauline Luna) for her congressional hearing against Twitter 1.0 employees for their role in mass censoring Conservative America,” followed by tons of codes gathered from the December 8th, 2022 Twitter Files dump by Bari Weiss, one of several batches released to the public by hand-chosen journalists. Other batches were analyzed by Alex Berenson, Lee Fang, Matt Taibbi, David Zweig, and Michael Schellenberg, and released on Twitter through March 2023.
RELATED: Upgraded Elements of 𝕏: Introducing a Tiered Structure, 'The Everything Finance App,' Social Credit Ratings, and Cryptocurrency Rewards for Content Creators
Parzival continues to make amazing discoveries in the Twitter algorithm, but with every jaw-dropping find, you can rest assured Parzival’s work will be plagiarized by larger accounts on the platform. Time and time again accounts with 100k+ followers, and in one case, even T(w)itter Daily News copied his work and posted it as their own. In turn, these larger accounts not only receive the traffic of clicks and shares on the posts but also financially benefit from the increased engagement with the platform’s monetization platform by way of ad revenue. Plagiarizing another user’s work is bad enough in and of itself, but turning other people’s work into revenue is a different ballgame altogether, and one that cannot continue to be overlooked. It’s an ongoing issue that many 𝕏 users have attempted to bring to the attention of Elon Musk, 𝕏 Premium, and 𝕏 Support, but is yet to be addressed.
Unlike the aforementioned 𝕏 users, The Parzival understands the importance of the information he’s providing others on the platform and continues to release his findings in good faith. The question is, will any of these issues ever be addressed by Musk, or engineers on 𝕏? Time will tell.