#ai coding agent
Ai Coding Agent: 5 AI articles covering ai coding agent news, analysis, and research
Articles
Hackers Breach AI Coding Tool OpenCode, Incurring Unauthorized⭐9
OpenCode Console breach exposed API keys for 1.4% of users, with some developers reporting hundreds to over $1,000 in unauthorized AI charges.
How to Break the AI Coding Agent Fix Loop⭐10
Why AI coding agents get stuck in fix loops and a five-step method—stop, revert, restate, isolate, verify—to break the cycle and ship working code.
Cognition Releases SWE-2: A Kimi K3 Post-Trained Coding Model⭐7
Cognition's SWE-2 coding model, post-trained from Kimi K3, hits 50.0% on FrontierCode 1.1, matching Fable 5.1 at 64% lower cost and running only inside Devin.
Microsoft Open Sources Code-Testing Generator: A Polyglot⭐6
Microsoft's new open-source code-testing generator boosts unit-test task completion from 78.9% to 92.1% versus Copilot, auto-validating tests locally.
Meta Debuts AI Coding Agent Muse: How It Stacks Up Against⭐9
Meta's Muse AI coding agent debuts, rivaling Claude Code and Codex in code generation, debugging, and workflow integration with unique legacy code handling.