OpenAI has launched a new cloud based software engineering agent on ChatGPT that can work on various coding related tasks at a time. As per OpenAI, the new tool is capable to writing features, fixing bugs and answering questions about the user’s codebase while running each task in its own sanbox (private coding environment).
Codex is powered by a version of OpenAI’s latest reasoning o3 which was optimized for software engineering related tasks. OpenAI says this model was trained using reinforcement learning on real-world coding tasks in a variety of environments to generate code that “closely mirrors human style and PR preferences, adheres precisely to instructions, and can iteratively run tests until it receives a passing result”.
OpenAI CEO Sam Altman while announcing the Codex release in a post on X (formerly Twitter) wrote, “It is amazing and exciting how much software one person is going to be able to create with tools like this.”
OpenAI is rolling out Codex to ChatGPT Pro, Enterprise and Team users from today and there are also plans to add support for Plus and Edu users in the coming weeks.
In order to use Codex users can go to the sidebar on ChatGPT and assign the AI agent a new coding task by typing a prompt and clicking on ‘Code’. To ask questions to Codex about their codebase, they can just click on ‘Ask’ after the text prompt.
“Codex can read and edit files, as well as run commands including test harnesses, linters, and type checkers.” OpenAI said in its blogpost.
That ChatGPT maker said that each task completion from Codex takes typically anywhere between 1 to 30 minutes, depending on its complexity and useres can monitor the AI agent’s performance in real time.
After the completion of an assigned task, Codex provides users of verifiable evidence of its actions via citatins of terminal logs and test outputs to allow users to trace each step taken and ensure that the model didn’t hallucinate (make stuff up).
Source link