New X/Twitter Algorithm Explained: Phoenix, Grok, & Scoring (2026)

The new X (Twitter) algorithm just dropped. I dug into the open-source code to understand exactly how the new "Phoenix" LLM (based on Grok-1) ranks content, calculates engagement scores, and filters your feed. In this video, I break down the technical pipeline—from candidate sourcing to the final weighted score—and give you actionable strategies to grow your account based on the actual code. IN THIS VIDEO: - Key Takeaways: Engagement is king, "In-Network" priority, and Video boosts. - The Architecture: Thunder vs. Phoenix (The AI Layer). - The Scoring System: Exact positive & negative weights (Likes, Replies, Dwell Time). - The "Diversity Penalty" and how to avoid it. - Actionable Advice: How to optimize for dwell time and replies. THE LOGIC: - Model: Grok-1 Open Source (Transformer Architecture) - Sourcing: Two-Tower Candidate retrieval - Signals: High penalties for blocks/mutes; high rewards for profile clicks & shares. TIMESTAMPS: 0:00 - Intro & Key Takeaways 1:40 - The Technical Flow (Hardware Stack) 2:20 - Positive Signals (Likes, Dwell Time, Video) 3:35 - Negative Signals (The "Kneecap" Killers) 4:20 - The Scoring System & Phoenix Prediction 6:40 - Network Logic (In-Network vs. Out-of-Network) 8:00 - Grok-1 & Machine Learning Context 9:30 - Actionable Strategy for Creators RESOURCES: Follow me on X ( https://x.com/alphastack1 ) and DM me if you want the breakdown document and code snippets used in this video.