OpenAI just dropped a major update for Codex—its coding agent—by launching a new desktop app for macOS (with Windows already available). Think of it as a command center for software development. Instead of one agent making edits, you can now run multiple agents in parallel, each working on different parts of a project. The app organizes agents into separate threads by project, so you can jump between tasks without losing your place. You can review code changes in the app, comment on diff lines, and even open edits in your editor for manual tweaks. It also supports worktrees, meaning multiple agents can work on the same repo simultaneously without stepping on each other’s toes. And for a limited time, Codex is included with ChatGPT Free and Go plans, with doubled rate limits across all tiers. The bigger picture here: we’re shifting from asking “what can agents do?” to “how do we collaborate with them at scale?”—and that requires a new kind of tool. Why it matters: This signals a move toward AI agents as true teammates that manage entire software lifecycles, not just write snippets of code.