X Publishes AI-Powered Algorithm Code

Why It Matters

The code release offers unprecedented insight into X’s content curation, affecting user retention, creator strategy, and competitive dynamics with platforms like Threads and TikTok.

Key Takeaways

• X released its “For You” feed code on GitHub
• AI now drives ranking via Thunder and Phoenix components
• Creator diversity limits post visibility, penalizes overposting
• Video dwell time weighted higher; external links de‑boosted
• Engagement gains disputed as user growth stalls

Pulse Analysis

The decision to open‑source X’s feed algorithm marks a rare moment of transparency in the social‑media arena. After a 2023 debut that never saw updates, the new repository shows how xAI’s servers now power the “For You” timeline, replacing hand‑crafted weighting with two distinct pipelines: Thunder, which ranks content from accounts you follow, and Phoenix, which surfaces AI‑curated posts based on the last 128 engagements. This shift reflects a broader industry trend toward machine‑learning‑first recommendation engines, aiming to keep users glued to ever‑shorter scroll sessions.

From a creator’s perspective, the algorithm introduces a diversity safeguard that throttles the number of times a single user appears in the feed each day. While this reduces echo‑chamber effects, it also diminishes the payoff for high‑frequency posting, nudging creators toward spaced, high‑quality content—especially video, which now enjoys elevated dwell‑time weighting. External links are penalized, and duplicate posts are filtered, reinforcing a preference for native, engaging media. These mechanics compel marketers to rethink cadence, prioritize long‑form or video assets, and engage directly through comments to boost signal strength.

Business implications are immediate. By tweaking the feed to prioritize AI‑predicted relevance, X hopes to lift average session length, a metric the company claims is rising among new users. Yet overall platform growth remains under pressure from rivals like Meta’s Threads, which are gaining traction. The public code may invite external audits and third‑party innovations, but Musk’s own admission that the algorithm “sucks” signals further iterations are on the horizon. How effectively X can translate these technical tweaks into sustained user growth will be a key barometer for the platform’s long‑term viability.