X Opens For You Algorithm, Visibility Data

This is X’s most meaningful algorithm-transparency release yet, but it still exposes architecture more clearly than production behavior.

• –Developers can inspect candidate retrieval, transformer ranking, scoring weights, filtering, and label-driven visibility decisions.
• –The Phoenix code now includes training and serving paths, making small-scale experimentation possible with synthetic data.
• –X separates ranking from visibility filtering, clarifying that reduced reach can result from eligibility labels rather than ranking alone.
• –The transparency tool gives users aggregate diagnostics, but not a post-level explanation or a guaranteed appeal path.
• –Some anti-gaming rules remain withheld, so the repository is useful for research without becoming a complete reach-optimization manual.