Openai has recently unveiled its new AI coding agent, Codex at Chattgpt. The company describes it as a cloud -based software engineering agent that can work on several tasks in parallel and operate by Codex -1. Openai says that Codex-1, a version of its O3 AI Reasoning Model, is quite well-organized. Designed to assist with software development, Codex can handle tasks such as feature implementation, bug fixing, codbase querry and even generating bridge requests for reviews. Each task is executed within a dedicated cloud-based sandbox, already set with the repository of your project. Let’s look at the whole story in 5 digits.

Openai’s new AI Agent, Codex launched: Story in 5 digits

– What is Codex AI agent by Openai? On Friday, Openi organized a live stream to announce the roll-out of its new AI coding agent, Codex. It is a coding AI agent built in Chatgpt and is capable of managing several development works at once. Openai states that it can generate features, fix bugs, and answer technical questions about the user’s codebase, each task is running safely within its own sandbox coding environment.

Powered by an optimized version of O3 Reasoning model of Openai, Codex has been adapted to software engineering specifically. The company says that this model was trained through learning reinforcement on real -world growth scenarios to produce the code that “the human style and PR preferences closely reflects the instructions, compliance with the instructions, and until it does not receive a passing results, can conduct a repetition testing”.

– Availability: Openai has slowly started rolling the Codex, subscribing to the Chatgpt Pro, Enterprise and team plans starting today. This phased release means that the first of those on high -level plans will be the first to achieve access to the new software development assistant.

The support for users on Chatgpt Plus and EDU schemes is expected to follow soon, although no specific timelines have been announced. This broad rollout will eventually create the advanced coding capabilities of Codex – such as feature implementation, bug fixing, and automated tests – are available to widespread audiences in both professional and educational settings.

— how to use? Codex Chatgpt is accessible through the sidebar, where you can assign new programming tasks by entering a prompt and selecting “code”. If you need an explanation or insight about your codebase, just click “Ask”.

– logic: Depending on the complexity, the tasks are usually completed within 1 to 30 minutes, and you can reveal the process in real time. Each task is handled separately in its own separate workspace, which automatically loaded with your codebase. Codex is capable of reading and modifying files, as well as executing commands such as test suits, lines and type checkers.

– After a given work: After completing a task, Codex presents its changes within its environment. This terminal provides transparent, verification qualified evidence of its functions by referring to logs and test results, so what you really were did. From there, you are able to review the result, request changes, increase a github bridge request, or apply the update directly into your own development environment. The tool allows you to configure the workspace of Codex to mirrors as closely as possible to your actual Dev setup.