As AI coding tools gain traction among software developers, their influence is extending to the very process of improving these tools. OpenAI is increasingly relying on its own AI coding agent, Codex, to develop and enhance itself. According to OpenAI employees, Codex is instrumental in its own evolution. In fact, “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, product lead for Codex at OpenAI. OpenAI built an AI coding agent and uses it to improve the agent itself.
Background Context
Codex, initially launched as a research preview in May 2025, functions as a cloud-based software engineering agent. It is capable of executing tasks such as writing new features, identifying and fixing bugs, and suggesting pull requests. The tool operates within sandboxed environments connected to a user’s code repository, enabling it to perform multiple tasks simultaneously. OpenAI provides access to Codex through ChatGPT’s web interface, a command-line interface (CLI), and IDE extensions for VS Code, Cursor, and Windsurf.
Table of contents
Official guidance: W3C — official guidance for OpenAI built an AI coding agent and uses it to improve the
The name “Codex” originates from a 2021 OpenAI model based on GPT-3, which powered GitHub Copilot’s tab completion feature. The name is rumored to stand for “code execution.” OpenAI wanted to connect the new agent to that earlier technology. OpenAI built an AI coding agent and uses it to improve the efficiency and effectiveness of the code generation process, echoing the impact of the earlier model. The initial model was a pivotal moment for AI in coding. “For many people, that model powering GitHub Copilot was the first ‘wow’ moment for AI,” Embiricos said. “It showed people the potential of what it can mean when AI is able to understand your context and what you’re trying to do and accelerate you in doing that.”
Competitive Landscape and Adoption
The command-line version of Codex shares similarities with Claude Code, Anthropic’s agentic coding tool launched in February 2025. While Embiricos acknowledged the competitive environment, he noted that OpenAI had been developing web-based Codex features internally before releasing the CLI version. OpenAI built an AI coding agent and uses it to improve its offerings in a competitive marketplace.
The command line version of Codex has seen significant adoption. Codex usage among external developers increased twentyfold after OpenAI launched the interactive CLI extension alongside GPT-5 in August 2025. Further accelerating adoption, OpenAI released GPT-5 Codex on September 15, a specialized version of GPT-5 optimized for agentic coding. OpenAI built an AI coding agent and uses it to improve the coding experience and performance for developers.
Internal Use and Open Source Integration
A significant portion of OpenAI’s engineers now regularly use Codex. The company utilizes the same open-source version of the CLI available to external developers. This allows developers to suggest additions and modify the tool. “I really love this about our team,” Embiricos said. “The version of Codex that we use is literally the open source repo. We don’t have a different repo that features go in.” OpenAI built an AI coding agent and uses it to improve the platform through internal use and external contributions.
The self-improving nature of Codex extends beyond code generation. Codex monitors its own training runs and analyzes user feedback to determine its next development steps. “We have places where we’ll ask Codex to look at the feedback and then decide what to do,” he said. “Codex is writing a lot of the research harness for its own training runs, and we’re experimenting with having Codex monitoring its own training runs.” OpenAI employees can also assign tasks to Codex through project management tools like Linear, similar to assigning tasks to human colleagues. OpenAI built an AI coding agent and uses it to improve its capabilities through continuous monitoring and feedback integration.
Historical Context and Future Implications
The concept of using tools to build better tools has historical precedents in computing. The creation of integrated circuits in the 1960s, which were initially designed by hand, led to the development of electronic design automation (EDA) software. This software enabled the design of more complex circuits than humans could manually create. Modern processors, with billions of transistors, exist because of software-driven design. OpenAI’s approach with Codex follows this pattern, where each generation of the tool enhances the capabilities of the next. OpenAI built an AI coding agent and uses it to improve the agent, which mirrors the historical progression of technology development.
OpenAI’s reliance on Codex to develop and improve itself highlights the increasing sophistication and potential of AI coding tools. By using its own AI to enhance its development processes, OpenAI is pushing the boundaries of what’s possible in software engineering. This recursive development model could lead to further advancements in AI coding agents and their role in the software development lifecycle. OpenAI built an AI coding agent and uses it to improve the future of software development.
Technology Disclaimer: Product specifications and features may change. Always verify current information with official sources before making purchase decisions.
Explore more: related articles.