OpenAI launched Codex, a cloud-based software engineering agent designed to run multiple coding tasks in parallel. It operates in isolated cloud sandboxes, reads your codebase, writes code, runs tests, and submits pull requests without requiring you to babysit each step.
The pitch is raw throughput. Instead of one task at a time, Codex handles a queue. You assign work across projects simultaneously and check results when they are ready. This is a structural change in how a single developer manages workload, not a smarter autocomplete.
The full product page at openai.com/codex details the actual task scope, supported repositories, and access tiers. If you are evaluating AI coding tools for a team, the specifics there matter more than this launch video.
[WATCH ON YOUTUBE →]