Code Agents

Cybersecurity startup Cathedral used AI agents to breach Anthropic's internal systems through third-party software as part of a bug bounty program. Anthropic confirmed access to sales, HR, and IT data but disputes how deep the breach went, highlighting exposure through external software dependencies.

AI Safety • The Decoder

Google opened Android Studio to third-party AI agents via a 'Bring Your Own Agent' feature supporting more than 40 coding agents, including Anthropic's Claude Agent and OpenAI's Codex. Google says new terminal tools help agents finish UI tasks 77% faster while using up to three times fewer tokens.

Developer Tools • The Decoder

Mistral released the weights for Mistral Large 4, codenamed Le Chonk, claiming it rivals top open models like GLM-5.3 on the Artificial Analysis Coding Agent Index. The release includes benchmark scores the company says are competitive with leading open-weight models.

Model Release • Techmeme

Productivity

Jacob Sansbury describes Spawn's agents changing live code and shipping to production hourly, reading errors, bug reports and metrics themselves; about 60% of work reportedly starts from a production error, a Savi bug report, logs, or a feature request.

An AI agent built on a Claude model autonomously selected a gene therapy, chose a brain target and injection coordinates, and performed surgery on an epileptic mouse with no human decision-making, according to lead researcher Ocana-Santero. When a calibration error occurred mid-surgery, the agent refused to continue and could not be persuaded otherwise.

Research • The Decoder