OpenAI Releases Code Agent to Open the Era of Automated Programming
OpenAI has released a new cloud-based code agent, Codex, on its website. Codex can automate a variety of tasks for developers, such as writing functional code, answering questions about the codebase, fixing bugs, and making pull requests for review. Each task runs in its own cloud sandbox environment, while preloading the user's codebase.
It is reported that Codex is powered by the Codex-1 model, a fine-tuned version of OpenAI-o3 that is optimized for software engineering. Trained on real-world coding tasks in a variety of environments through reinforcement learning, the generated code fits the programmer's coding style and pull request preferences, and strictly follows instructions. (Jin Ten)