🤖 Sashiko: Agentic Linux Kernel Code Review

⭐⭐⭐⭐⭐ 5 stars | Source: Hacker News | Date: 2026-03-22

Summary

Sashiko (刺し子, "little stabs") is an agentic Linux kernel code review system developed by Google. It monitors public mailing lists to thoroughly evaluate proposed Linux kernel changes using AI, acting like a team of specialized reviewers covering domains from high-level architecture verification to low-level concurrency analysis.

Key Features

Multi-domain review: Covers architecture verification, security audits, resource management, and concurrency analysis
Open-source prompts: Uses per-subsystem and generic prompts created by Chris Mason
Google-funded: All compute resources and LLM tokens provided by Google
Real-world results: Successfully identified 53.6% of bugs based on historical commits with "Fixed:" tags
Human complement: 100% of these historical bugs had previously made it through human-driven code reviews

        🔑 Key Insight: Sashiko proves AI can find bugs that human reviewers missed. It's designed to augment and assist human reviewers, not replace them.
    

Technical Details

The system relies on:

An open-source set of per-subsystem and generic prompts
A custom multi-stage review protocol to maximize accuracy and minimize false positives
Gemini 3.1 Pro as the underlying LLM

Related Projects

The review-prompts repository includes specialized prompts for:

Linux Kernel: /kreview, /kdebug, /kverify
systemd: /systemd-review, /systemd-debug, /systemd-verify
iproute: /iproute-review, /iproute-debug, /iproute-verify

Implications

This represents a significant step in AI-assisted software engineering. While AI won't replace human reviewers, it can catch bugs that slip through traditional review processes. The fact that 100% of identified bugs had passed human review highlights the potential value of AI augmentation.

探索于 2026-03-22 | Hacker News