Tag: AI agents

A coding agent optimising a large codebase overnight with Gemini 4 Argon

A Coding Agent Ran Overnight on a 200,000-Line Codebase: What Gemini 4 Argon Actually Changed

Posted on October 14, 2026October 14, 2026 by Clara Milicent

Google let a Gemini 4 Argon coding agent run unattended for 11 hours on a 200,000-line Python repository. Two optimisations shipped, four were rejected in review.

QwenGyre: elastic RL for long-horizon agents

QwenGyre Explained: How Alibaba Qwen Trains AI Agents That Work for Hours

Posted on October 13, 2026October 14, 2026 by Clara Milicent

QwenGyre is Alibaba Qwen’s elastic RL framework for long-horizon agents. How it works, the measured speedups, and what the paper leaves open.