Microsoft’s Claude Code Study: 24% More PRs, Then Canceled
Microsoft's largest Claude Code field study found 24% more merged PRs across tens of thousands of engineers — then they cancelled the tool. ...
ChatGPT Voice in Codex and Work: GPT-Live Desktop Guide
OpenAI brought GPT-Live's full-duplex voice to the ChatGPT desktop app for Codex and Work. Here's ...
GuardFall: Shell Injection Hits 10 of 11 AI Coding Agents
GuardFall exposes shell injection flaws in Aider, Cline, Goose, and 7 more AI coding agents. ...
Certighost: Any Domain User Can Compromise Active Directory
CVE-2026-54121 (Certighost) gives any domain user full Active Directory access. Working exploit is public. Patch ...
Big Tech Signed an Open-Weight AI Letter. Check the Guest List.
Twenty-five companies — including Nvidia, Meta, and Microsoft — published a letter defending open-weight AI. ...
Claude Opus 5 Beats Fable 5 on Coding at Half the Price
Anthropic's Claude Opus 5 scored 30.2% on ARC-AGI-3 yesterday, tripling any prior record, while beating Fable 5 on coding benchmarks at half the ...
FrontierCode: AI Can’t Write Production Code Yet
Cognition's FrontierCode benchmark tests whether AI-generated code is actually mergeable. The best model scores 13.4% ...
GPT-5.6 Sol, Terra, Luna: Which Tier Fits Your Workflow
OpenAI’s GPT-5.6 ships as three models with a 5x price gap. Sol, Terra, and Luna ...
Big Tech’s $1.65T Hidden AI Debt: What Developers Risk
$1.65 trillion in hidden AI debt sits off Big Tech balance sheets. Nikkei Asia just ...
VS Code 1.130: Agent Host, Assisted Approvals, TypeScript 7
VS Code 1.130 ships the Agent Host Protocol, AI-gated tool approvals, and TypeScript 7 as ...