AI for Kernel Engineers
summary
AI generates a full patch that replaces open‑coded hash tables with hashtable.h and lands upstream after seven lines change and a human review, and it writes commit messages as well. A Rust rewrite converts fragile bash CV tooling into tested, faster tools and a git-resolve tool with comprehensive tests ships upstream. A retrieval‑augmented autoell backport classifier builds a vector DB from decades of decisions and reduces the 100‑patch/day review burden to a tractable queue. A sibling CV classifier and commit‑classifier library runs git research, explains its reasoning, and is already used daily.
AI-written summary · about
chapters (12)
- 0:00AI for kernel engineers
- 3:51LLMs as compilers
- 6:18Kernel code patches with AI
- 8:44AI for kernel tooling
- 11:08Semantic code search
- 13:11RAG for AI hallucinations
- 15:38AI agents for git history
- 17:26AI-powered stable backports
- 24:51Bash to Rust rewrite
- 28:45AI for CVE assignment
- 32:15Summary and resources
- 32:59Q&A
related talks
14mAI's impact on open-source contributions and mentoring
25mHow AI Is Changing Open Source Communities: Lessons From openEuler
51mAI for the People
15mAI's Impact on Development, Post-Production & Support Journeys
41mSkill Degradation: An Empirical Analysis of 400+ AI‑Generated Security Fixes
21mHow To Make AI-Assisted OSS Contributions Review-Ready
15mUsing AI effectively in large Java codebases
20mAI inside the IDE