Executive Summary

I learned that Ginux is an independent Rust-based kernel for x86-64, not a fork of the Linux kernel. The project boots Rocky Linux 10 with systemd and refers to Linux as an "offered personality," while its memory management, scheduler, process model, filesystem, networking, and device support are written from scratch.

The Problem It Solves

What is interesting about Ginux is its dependency on Linux at the ABI level, rather than on the Linux kernel implementation. Existing compiled Linux binaries can run without modification because Ginux implements the Linux ABI; therefore, a Rocky Linux 10 user space can be used as the testing basis.

This is also an experiment on the limits of assisted coding for software systems. Gregory Kurtzer—the creator of CentOS and Rocky Linux—clearly divides the roles: humans determine the architecture, constraints, milestones, acceptance criteria, and stop conditions, while an AI agent runs cycles of writing code, booting, testing, diagnosing failures, and then fixing them.

Key Features and Code Examples

Ginux is packaged as a single GINUX.EFI image that boots directly through UEFI without a host Linux kernel. Ginux runs isolated "domains," each with its own processes, address space, filesystem, and lifecycle; these boundaries are portrayed as stronger than those of OCI containers or Linux namespaces.

To form a domain's root filesystem, Ginux can pull and verify OCI container images. Rocky Linux 10 can then boot up to systemd's multi-user.target, and several Alpine Linux programs such as musl and BusyBox also run.

The name Ginux refers to GPT: GPT-5.6 Sol was used to write the code. In one 11-day OpenAI Codex task, the agent generated approximately 139,000 lines of implementation and test code, maintained across more than 1,000 commits.

Example of using an image and domain:

# Illustration: forming an image that can be used by Ginux
podman pull docker.io/library/rockylinux:10
podman verify docker.io/library/rockylinux:10

The commands above only demonstrate the image-handling flow described; the article does not provide a complete Ginux API or CLI.

Impact and Improvement Notes

Ginux is still at an early stage. Currently, the project only runs in a QEMU virtual machine and does not yet have a physical hardware profile, GPU or NVMe support, nor has it undergone an independent security audit.

I see it waking up; this is not yet a replacement for Linux. Compatibility with systemd, hardware support, security review, standards, community, and long-term maintenance still require optimization. The Linux kernel has been built through decades of hardware enablement and code review, while Ginux has just shown that an architect who understands kernels can use AI agents to expand a prototype much faster.

For me, Ginux's main value lies not in the claim that "AI replaces kernel engineers," but in the demonstration of a reproducible engineering loop: architecture and acceptance criteria remain human decisions, while the agent handles implementation and repeated testing.

---

First announced in It's FOSS News. Read the original source for complete details.

Read also