The rise of AI coding tools is reshaping software development, with OpenAI at the forefront, utilizing its AI coding agent, Codex, to enhance the very tool itself.
OpenAI employees detailed the company's reliance on Codex in a recent discussion with Ars Technica, revealing that the AI contributes significantly to its own development. 'I think the vast majority of Codex is built by Codex, so it’s almost entirely just being used to improve itself,' said Alexander Embiricos, the product lead for Codex at OpenAI.
Launched in May 2025 as a research preview, Codex acts as a cloud-based software engineering agent capable of writing features, fixing bugs, and proposing pull requests. It operates in secure environments tied to users' code repositories, managing multiple tasks simultaneously. Codex is accessible through ChatGPT’s web interface, a command-line interface (CLI), and extensions for popular IDEs such as VS Code, Cursor, and Windsurf.
The name 'Codex' stems from a 2021 OpenAI model based on GPT-3, which powered GitHub Copilot’s tab completion feature. Embiricos mentioned the name’s alleged meaning, 'code execution,' and the desire to link the new tool to the earlier impact of AI on code development. 'For many people, that model powering GitHub Copilot was the first ‘wow’ moment for AI,' Embiricos reflected. 'It showcased the potential of AI to understand and accelerate user objectives.'
Codex’s current CLI version has parallels to Claude Code, Anthropic’s February 2025-launched tool, drawing comparisons in their functionality. Though Embiricos avoided direct comparisons, he acknowledged the vibrant competitive environment, highlighting OpenAI’s lead in web-based features before Codex's CLI debut.
Credit: OpenAI - The interface for OpenAI’s Codex in ChatGPT.