Last updated: May 18, 2025

Justin Varghese, Your Money Editor

Dubai: By launching a research preview of Codex, OpenAI reveals its most capable AI coding agent yet on ChatGPT.

OpenAI released a new AI-powered tool called Codex, designed to help software engineers write and fix code much faster. The tool, now available in preview form to some ChatGPT users, is being described as a potential game-changer for the world of programming.

Virtual coder, not just a tool

Codex functions like a smart coworker that operates inside the ChatGPT interface. Rather than replacing your computer setup, it runs in the cloud and can handle multiple programming tasks at once.

Users simply type what they want it to do — such as writing a feature or fixing an error — and Codex takes over from there. Depending on how complex the job is, it might complete the task in a few minutes or take up to half an hour.

During its work, Codex can read code files, run tests, and even edit existing codebases, all without needing internet access. It connects directly to GitHub repositories you provide, keeping its work limited to the environment you set. Once done, it shows a record of its actions, including any challenges it faced during the task.

How everyday users can try it

To use Codex, subscribers to ChatGPT’s Pro, Team, or Enterprise plans can find it in the app’s sidebar. From there, it’s a matter of assigning a task using a natural language prompt — no technical instructions needed. For example, you might type “Add a login function to my website,” and Codex will begin work. Users can also ask questions about their code and get step-by-step help in return.

If the tool runs into a problem — such as a failed test or unclear instructions — it will pause and explain what went wrong. This allows users to make decisions or adjust the task without needing deep coding expertise.

Coders, developers, take note

For developers worldwide, tools like Codex signal a shift in how software projects might be built in the near future. Companies looking to reduce costs or speed up delivery times may increasingly rely on AI to take care of routine programming jobs. This also opens up opportunities for individuals without deep technical backgrounds to begin experimenting with software development.

While Codex is impressive, it’s part of a much larger trend. AI tools that write code are becoming more common in the tech world. Google and Microsoft both estimate that about 30% of the code within their companies is already generated by AI systems. This figure is expected to grow as AI models become faster, more accurate, and better at understanding complex instructions.

OpenAI is also using Codex internally to handle repetitive tasks, like testing and documentation, freeing up human engineers to focus on more creative or strategic work. According to OpenAI’s leadership, the long-term goal is for AI tools to work as reliable teammates — not just assistants — within software teams.

Also Read:OpenAI says it raised $40 billion at valuation of $300 billion

Codex still has limits

Despite its strengths, Codex isn’t perfect. Like all AI tools, it can make mistakes, especially when faced with vague or poorly written instructions. Studies have shown that even the most advanced coding models sometimes struggle with debugging or fail to catch subtle errors. So, human oversight is still essential, especially for critical applications.

OpenAI has built in safeguards to reduce risks. For example, Codex refuses to create harmful or malicious software and operates in a tightly controlled environment with no access to the wider internet or external APIs. These measures help prevent misuse, but they can also limit the tool’s flexibility in certain advanced projects.

AI coding: A tool, not a threat

For software developers, startups, and tech students, tools like Codex offer a chance to boost productivity and explore new workflows. But they also highlight the growing importance of learning how to work alongside AI — not against it. As AI handles more of the grunt work, the role of the developer could shift from writing code line-by-line to reviewing, guiding, and refining what AI produces.

Staying updated on tools like Codex may not just be helpful — it could be critical for anyone planning to remain competitive in the job market over the next decade.