OpenAI recently introduced a revamped version of Codex—powered by the cloud, it can AI handle multiple programming tasks simultaneously without compromising your laptop’s performance. This comes just a month after the launch of the free, open-source version, “Codex CLI.”
Unlike the usual code completion and code generation tools we know from our coding sessions with AI chatbots, Codex operates as a semi-autonomous agent with its own computing power. It is capable of handling everything from bug hunting to complex refactoring.
During an official presentation, OpenAI President Greg Brockman made it clear that software engineering is accelerating and that its appearance will fundamentally change by the end of 2025. He emphasized that the new Codex is a step towards the future of software engineering: a remote software agent that can perform many tasks simultaneously.
In a demonstration, the team shows how multiple AI agents can work on different parts of a codebase simultaneously, each in its own isolated environment and with minimal human guidance. This kind of efficiency is impressive; one team member reported previously spending at least half an hour or even hours on a problem, only to have the AI solve it in the background in no time.
The AI is perceived as equally, if not more, trustworthy than human colleagues. This trust in Codex is an important step in creating an advanced collaboration between humans and machines.
The new Codex runs on OpenAI’s infrastructure instead of locally. Its agentic architecture allows the AI to self-evaluate and improve. Developers can run multiple programming tasks simultaneously and view the results later, as the agent is able to evaluate its own output and even fix bugs.
With the new Codex One, OpenAI's most advanced coding model to date, they have managed to optimize performance for the types of code that are actually integrated into applications.
Currently, Codex is available to ChatGPT Pro, Enterprise, and Teams users, with plans to expand to Plus and Edu users later. OpenAI has said that they will start with “very generous rate limits” and no additional fees, but that this may change as they gather more usage data and feedback.
Future integrations of Codex with issue trackers and CLI systems could further streamline the automation of the development process. OpenAI also continues to work on Codex CLI, with the vision that local and remote versions can work together seamlessly.
Brockman put it well: What you really want is a remote colleague with their own computer who is also looking over your shoulder. Imagine you’re typing and you want to take a break—you say to Codex, “Can you finish this?” The AI takes over without a hitch and runs it in the cloud.
A twist on the future of software engineering: it may seem like science fiction, but here we are.
What is Codex and what makes it different?
Codex is a cloud-based system of AI agents that executes multiple programming tasks simultaneously without taxing local computer resources, making it truly semi-autonomous and capable of self-evaluation and improvement.
Who is Codex currently available to?
Initially, Codex will be accessible to ChatGPT Pro, Enterprise, and Teams users, with plans to expand roles for Plus and Edu users.
What does the future of Codex look like?
OpenAI plans to integrate Codex with issue trackers and CLI systems, further automating the development process. They will also continue to work on the open-source version, with the vision of on-premises and cloud coming together.