Understanding Twitter's Algorithm: Key Insights and Implications

0xTodd posted on X. Elon Musk has fulfilled his promise to open-source Twitter's algorithm, providing users with insights into how content is curated and displayed on the platform. This move allows users to better understand what content they are likely to see and how they can effectively express their viewpoints to reach a wider audience. The algorithm introduces two key concepts: 'Thunder,' representing accounts a user follows, and 'Phoenix Retrieval,' encompassing all Twitter content, including accounts not followed by the user. Additionally, each user has a hidden 'Grok' assistant that predicts engagement with posts and determines whether to display content from Thunder or Phoenix Retrieval. Twitter's algorithm has evolved from hard-coded rules to a more dynamic system. For instance, in 2023, posts with videos were scored higher, while those with links were penalized. Now, the scoring is less mechanical, focusing on user preferences. Several overlooked rules impact content visibility. An invisible reputation score, ranging from -128 to +100, is assigned to each user. New accounts start with a low score, limiting their content's reach until they interact positively with high-reputation users. Engaging with low-quality accounts can negatively affect one's score. The algorithm also addresses 'topic fatigue,' reducing the ranking of multiple posts from the same author within a short period. Repetitive content on popular topics is penalized, encouraging timely posting. Negative feedback, such as muting or blocking, carries more weight than positive interactions like likes. Extreme behavior can lead to reduced visibility, as one block can outweigh numerous loyal followers. Verified accounts (Blue V) are favored in the algorithm, making it easier for their content to be retrieved by Phoenix Retrieval. Unverified accounts struggle to reach the initial candidate pool unless their engagement grows rapidly. Other factors include dwell time, which adds points even if users don't interact with posts, and candidate isolation, where each tweet is evaluated independently. The algorithm avoids recommending repetitive or outdated content and emphasizes the importance of video completion rates. These insights provide a deeper understanding of Twitter's algorithm, inviting further discussion and exploration.