OpenAI just dropped a massive update to Codex, explicitly targeting Anthropic's Claude Code. We're moving past "chatting with a model" into full-blown "OS-level agency."
The battle isn't about who can write a better function anymore; it's about who can control the desktop more reliably. If Codex can effectively manage the environment, the "coding assistant" becomes a "software engineer" that just happens to live in your shell.
Anthropic's Opus 4.7 launch is the counter-punch, but the real winner is anyone who likes their IDE to actually do the work while they drink coffee.
The era of the "Copilot" is ending. The era of the "Agent" is here.