Simon Willison gave the same one-shot prompt that had Claude Fable 5 build a little raccoon game to Codex running GPT-5.6 Sol Ultra. The result: a noticeably better game — with a very funny bug.
Free and Go users get unlimited text chats and GPT-5.6 Luna as the new default model. Add a Think button for hard questions - and Sol gets sharpened for paying users.
At Black Hat, OpenAI walked through — for the first time in detail — how autonomous agents accidentally attacked Hugging Face during an internal cyber evaluation. The timeline is sobering, and instructive.
Two new Claude Code versions in quick succession. The focus is clearly on security: worktree isolation now applies everywhere, and several permission gaps are closed. Plus a handy /teleport hint for cloud sessions.
The new release brings self-hosted environments: Claude Code's web, mobile, and desktop sessions now run on your own hardware. Plus cross-session messaging — agents can now talk to each other across multiple machines.
Anthropic reworked the biology safety classifiers behind Claude Fable 5. The result: roughly 85 percent fewer false alarms on harmless biology questions. Everyday topics like lab results and symptoms now hit the fallback far less often.
First Anthropic and OpenAI, now Meta: another AI model attacked an outside company during safety testing. And the UK's AI Security Institute reports agents that went after real people on the open internet.
Meta unveiled Muse Code — an AI agent for large codebases that handles complete engineering tasks. It runs on Muse Spark 1.2 and fans big jobs out into parallel sub-agents. The pitch: cheaper than Codex and Claude Code.