X Algorithm Pledge Draws Criticism
Elon Musk’s promise to make X’s algorithm more transparent is facing fresh pushback from crypto users. The May 15 update to the platform’s code repository has not fully addressed concerns about how feeds rank posts. Crypto users say they expected more openness after Musk pledged monthly updates in January.
In January, Musk said the algorithm code would be published within a week and refreshed every four weeks with detailed developer notes. He also said he welcomed criticism and pointed users to the Following tab for a non-algorithmic feed. But the official xai-org repository has only received one commit since its launch on January 17. That has left many users frustrated.
The repository itself contains four components written mostly in Rust and Python. The promised developer notes explaining ranking changes have not appeared. Without those notes, users say they cannot understand why some posts get more reach than others. A similar situation occurred in 2023 with the older Twitter algorithm repository, which also saw activity slow after initial promise.
Reach Issues for Crypto Content
Crypto users have reported weaker reach for their posts on X. Several users say crypto-related content appears less often in their feeds. Market watcher Ethan noted that the feed now shows more politics, rage bait, and engagement bait. He said crypto content appears far less often than before, and that X is losing the topic-based community structure that once made it useful.
Ethereum co-founder Vitalik Buterin had questioned the transparency standard before the repository was released. He wondered if X could provide enough detail for meaningful public review. The published code does show the final score formula, but it does not include the weights for each predicted action. That missing detail limits outside analysis, making it hard to assess how posts are ranked or why certain content gains visibility.
Critics Point to Missing Details
The Phoenix module README says its transformer is representative of the internal model, except for specific scaling optimizations. Critics say this shows the public code differs from the deployed system. There are also concerns about negative signals. The model could learn from reports and blocks, which critics say could make coordinated bot activity a suppression tool.
Some have pointed to Farcaster as a contrast, since it publishes forkable protocols instead of limited sample code. For now, crypto users remain skeptical that X’s algorithm transparency efforts will improve anytime soon.
