OpenAI has just rolled out Codex CLI v0.105, bringing syntax highlighting, faster prompt input, and improved multi-agent workflows. Voice dictation via spacebar is the buzziest addition, though early testers report gaps across setups and platforms.
Claude Code has just rolled out auto-memory, letting it retain project context, debugging patterns, and preferred approaches between sessions. The update also introduces Memory.MD alongside Claude.MD, with a /memory toggle to disable it.
A recent write-up by Vercel takes a closer look at how many AI coding agents run generated code with the same access as real credentials. It outlines practical boundary patterns—plus what still fails—to reduce prompt-injection and secret-leak risk.
In a newly published post, David Heinemeier Hansson revisits an Etsy rewrite that stalled for years, arguing the real work isn’t code—it’s the system around it. He connects today’s Claude Code moment to shifting team “social contracts” as output gets near-free.
Cursor has just rolled out an upgraded cloud agent system that runs each task in an isolated VM. The agents can build, test, and attach artifacts like videos and logs before opening merge-ready PRs. Cursor says over 30% of its merged PRs now come from these agents.