hn.today

Greg Kurtzer's Ginux is now a userspace kernel

github.com121 points81 comments
Screenshot of Greg Kurtzer's Ginux is now a userspace kernel

Ginux has shifted from an experimental standalone Rust kernel, which booted through UEFI and ran Rocky Linux, to a userspace kernel running as an ordinary Linux process. The earlier kernel work demonstrated that an AI coding agent, guided by an experienced developer and executable tests, could build a system capable of running real Linux software. It also showed why recreating kernel mechanisms cannot replicate Linux’s decades of compatibility, hardware support, and development experience. Ginux now takes a different approach: preserve Linux as the host, but give each managed workload a smaller, explicit, versioned interface rather than promising the full Linux ABI.

Before a managed program’s system call affects the host, Ginux can answer it using virtual state, validate and translate it into a limited host operation, or deny it. The implementation includes process and identity models, virtual filesystems and system views, assigned devices, mediated networking, and a pinned Rocky Linux 10 baseline that boots systemd to multi-user mode and supports login, updates, and clean shutdown. The project is also testing whether this boundary can contain AI agents, limiting access to capabilities such as files, processes, devices, and networking while recording policy decisions. That remains a research goal, not a claim of production-ready isolation. Compatibility and security claims are restricted to named configurations and tested workloads; the project emphasizes reproducible tests, performance checks, and explicit limitations.

Read on github.com81 comments on Hacker News

Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.

More in Programming

Ggml v1.0.0

Ggml v1.0.0

ggml 1.0.0 removes deprecated public APIs and legacy DIAG_MASK operations, while adding configurable YaRN truncation and FP8 support. It also improves backend performance with quantized matrix multiplication, fused MoE operations, and new attention and scheduling paths. (github.com)

The daily digest

Today's best Hacker News stories, summarized and screenshotted, one email a day.