Ginux Kernel Built by AI Agent in 11 Days
Gregory Kurtzer used an AI coding agent to build an experimental operating system that boots Rocky Linux 10 without the Linux kernel.
Updated on Nov. 2, 2026 in Artificial Intelligence
Gregory Kurtzer developed Ginux, an experimental operating system kernel, using an AI coding agent in an 11-day period. The project contains about 139,000 lines of implementation and test code, mostly written in Rust, and targets x86-64 processors.
Why it matters
The project illustrates how AI coding agents can produce substantial, working systems software under human-directed engineering. Ginux has significant limitations in security, compatibility and hardware support, so it remains an experiment rather than a production alternative to Linux.
Ginux is built primarily in Rust and implements the Linux application binary interface without using the Linux kernel. It boots directly through UEFI and executes selected Alpine Linux programs, though its security, compatibility and hardware support remain limited.
The players
Gregory Kurtzer
Kurtzer is a technologist known for founding the Rocky Linux and CentOS-related open-source communities, who directed the Ginux development.
OpenAI Codex
OpenAI Codex is a coding agent product from OpenAI, powered in this project by the GPT-5.6 Sol model.
The details
Kurtzer used OpenAI Codex, powered by GPT-5.6 Sol, through a method he calls Human-Directed Agentic Engineering. He set the architecture, requirements and acceptance criteria, while the agent wrote code, ran tests, identified problems and revised its work.
Timeline
The Ginux project was completed in an 11-day development period.
The Tech Race
Ginux is a test of whether OpenAI Codex can handle systems-level software that has traditionally required years of specialist engineering. The project positions AI coding agents as a potential accelerator for low-level operating system development.
For developers, Ginux shows that AI agents can produce large amounts of working code quickly, but its limits mean it is not yet a practical replacement for mainstream operating systems. Users should expect that experimental kernels of this kind will not offer the security or hardware support of established systems.
The takeaway
AI coding agents can now help build complex systems software, but human architects still define the goals and judge the results. Ginux demonstrates the approach's promise while also showing how far experimental systems remain from production use.
Further reading
Explore more Artificial Intelligence stories to see how coding agents are reshaping software development.
