OpenAI launches Codex for AI coding: Is this the end of software engineers?

OpenAI has unveiled Codex, its most capable AI coding agent to date, currently available as a research preview for ChatGPT Pro, Enterprise, and Team subscribers. Powered by the latest o3 AI reasoning model fine-tuned specifically for software engineering, Codex is designed to act as a "virtual coworker," helping developers write code, fix bugs, run tests, and perform multiple software engineering tasks simultaneously, all in the cloud.

Codex operates within a sandboxed virtual environment preloaded with users’ code repositories via GitHub integration, ensuring it accurately reflects individual development environments. Users interact with Codex through the ChatGPT sidebar by entering prompts and choosing to either generate code or ask questions about their codebase. The agent executes tasks in isolated containers, with execution times ranging from one to 30 minutes depending on complexity.

A notable feature of Codex is its transparent workflow: it shows its thinking process step-by-step, provides citations of terminal logs as verifiable proof of its actions, and openly communicates any uncertainties or test failures. This design aims to build trust and improve code reliability, addressing common criticisms of earlier AI coding tools that sometimes generated opaque or non-compliant scripts.

Developers can further customise Codex’s behaviour by including an "AGENTS.md" file in their repositories, which gives the AI specific instructions about code context, style guides, and standard practices, essentially serving as a README for AI agents.

OpenAI CEO Sam Altman highlighted Codex’s potential on social media, emphasising its ability to run multiple tasks in parallel and speed up engineering workflows significantly. The agent can read and edit files and execute commands such as test harnesses, linters, and type checkers.

Currently, Codex is accessible to eligible ChatGPT users with generous free usage to explore its capabilities. However, OpenAI plans to introduce rate limits and a credit-based pricing model in the coming weeks. The company also intends to expand access to ChatGPT Plus and educational users soon.

While Codex automates many programming tasks, OpenAI stresses that manual review and validation by human developers remain essential to ensure code quality and security.