Twitter has released its tweet recommendation algorithm code on GitHub, fulfilling the promise of Elon Musk earlier this month. It also posted a new overview of how its tweet recommendation algorithm works, providing new insights into the order in which tweets are displayed. While it aimed to provide the highest possible degree of transparency, Twitter has excluded any code that would compromise user safety and privacy or the ability to protect its platform from bad actors. Likes and retweets are the most significant indicators of interest and will help to boost tweet reach, while tweets that trigger negative user actions, like a block or unfollow, will limit performance. Twitter’s tweet ranking is conducted via a ‘~48M parameter neural network, which is continuously trained on tweet interactions to optimize for positive engagement (e.g. Likes, Retweets, and Replies)’
