OpenAI's Codex update is the clearest signal yet that AI coding tools are expanding beyond the editor and into full desktop control. Codex can deploy multiple agents that operate independently on a user's Mac, each working on different apps while the user continues other work. Operations include clicking buttons, typing into forms, and managing multi-app workflows.
The feature directly mirrors Anthropic's Computer Use API and Claude Code's local orchestration capabilities, signalling a new battleground between OpenAI and Anthropic. The Pragmatic Engineer survey from last week showed Claude Code at 46% most-loved rating among coding agents - OpenAI appears to be catching up fast on capability.
Desktop-level AI agents are no longer a Claude-exclusive. If you're building workflows around Computer Use or Claude Code, plan for OpenAI Codex to be a drop-in alternative within weeks.
