Twitter exposed source code, including recommendation algorithm.

Twitter has made its algorithm for tweet recommendations in users’ timelines open to public scrutiny as promised by its CEO, Elon Musk. The social media platform released two repositories containing source code for several key parts of the network’s functioning, including the systems used for For You timeline tweets control. Twitter described the move as a first step towards greater transparency that also prevented risks to the platform and its users. However, the open source releases still exclude elements such as code for Twitter’s advertising recommendations or data used to train the recommendation algorithm. Twitter stated that it has taken steps to ensure users’ privacy and pointed out that the release was focused exclusively on developers.
Commenting on the release in a Twitter Spaces session, Musk acknowledged that there would be mistakes identified in the algorithm but pledged rapid correction. He also stated his aspiration for Twitter’s open-source model to approach the Linux open-source operating system.
The algorithm features multiple models, including those for identifying NSFW or abusive content, measuring the likelihood of user interactions and calculating Twitter user “reputations” (with no clarification provided). Neural networks rank tweets and recommend accounts, while a filtering component ensures legal compliance and user trust. According to Twitter, the For You timeline consists of 50% tweets from people users don’t follow and 50% from those they do, on average. The tweet-ranking feature uses a 48-million-parameter neural network trained to optimize positive engagement. Twitter is now reportedly working on tools to manage suggestions from the community and sync changes to its core codebase.
Twitter has faced past controversies over changes to its recommendation algorithm. In November 2020, the platform began showing users more tweets from people they don’t follow after having previously tried to implement the change but faced user pushback. In February 2021, Musk himself urged Twitter’s engineers to adjust the algorithm so that his tweets received wider reach. Twitter later modified its algorithm, but it remains unclear whether this related to Musk’s request.