[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-torvalds-first-ai-assisted-patch-claims-23x-test-gain":10,"sections":35},{"siteName":4,"siteTagline":5,"publisherName":4,"contactEmail":6},"The Revision","Tech news, decoded.","editor@therevision.news",{"gaMeasurementId":8,"adsenseClientId":9},"G-ZW2MV82GYR","ca-pub-8533917693782264",{"article":11},{"id":12,"slug":13,"title":14,"dek":15,"body_md":16,"tags_json":17,"published_at":18,"created_at":19,"updated_at":20,"status":21,"review_note":22,"review_notes":23,"image_url":22,"persona_id":22,"persona_name":22,"section":24,"tags":25,"sources":30,"feedback":34,"feedback_at":22,"cost_usd":34,"total_tokens":34},11756,"torvalds-first-ai-assisted-patch-claims-23x-test-gain","Torvalds' First AI-Assisted Patch Claims 23x Test Gain","Linus Torvalds used an LLM to fix a long-running file-read bottleneck in Linux, though he says no real workload will see the 23x benchmark gain.","Linus Torvalds has posted his first Linux kernel patch written with help from a large language model (LLM), and he says a synthetic benchmark ran 23x faster with it.\n\nThe patch sits in a development branch, so it is not part of a released kernel. It targets an old problem: when many CPUs read the same file at once, they contend over shared cache data and slow each other down. Shared libraries, databases and web servers can all trigger that pattern. Several developers have tried fixes over the years, with varied results, and none was an all-encompassing solution.\n\nTorvalds posted his own draft in early 2024 and has run it on his laptop since. Recently he decided to \"try to actually solve this thing using an LLM.\" He also tempered the headline number: \"no real load\" will see anything close to 23x. A synthetic benchmark is built to hammer one weakness, so treat it as a ceiling, not a forecast.\n\nThe more interesting part is the pedigree. Torvalds already had a working draft and years of familiarity with the problem, so he could judge the machine's output rather than trust it. That is the use he has been defending. In July he said he was \"willing to absolutely put [his] foot down\" on AI skeptics, and that an automated review tool the kernel team uses outperforms humans and catches basic mistakes. The team is also dropping its requirement that authors disclose AI use. His position is that LLMs are tools like compilers or editors, and that people should \"understand what they are doing.\"\n\nThe same tools cut the other way. AI agents have become good at spotting common security flaw patterns, and the Linux security mailing list became \"unmanageable\" earlier this year. The kernel fixes around 2,000 CVEs (publicly catalogued vulnerabilities) per release. The flood of reports is pushing the team to drop older hardware drivers with a negligible install base. One tool speeds up the fix and the other multiplies the bug reports, and the open question is which grows faster.","[\"linux\",\"ai-coding\",\"kernel\",\"linus-torvalds\"]","2026-10-16T10:30:00.000Z","2026-10-18T11:08:24.158Z","2026-10-18T11:08:26.579Z","published",null,[],"dev-tools",[26,27,28,29],"linux","ai-coding","kernel","linus-torvalds",[31],{"name":32,"url":33},"Tom's Hardware","https:\u002F\u002Fwww.tomshardware.com\u002Fsoftware\u002Flinux\u002Flinus-torvalds-posts-his-first-llm-assisted-patch-to-the-linux-kernel-after-issuing-clear-pro-ai-directives-the-bossman-makes-use-of-ai-tooling-himself",0,{"sections":36},[37,42,47,52,57,62,67,72,77,82,86,91,96,101],{"name":38,"slug":39,"count":40,"latest_published_at":41},"AI","ai",7040,"2026-10-18T11:00:00.000Z",{"name":43,"slug":44,"count":45,"latest_published_at":46},"Security","security",1004,"2026-10-18T12:05:00.000Z",{"name":48,"slug":49,"count":50,"latest_published_at":51},"Policy","policy",539,"2026-10-17T19:38:12.000Z",{"name":53,"slug":54,"count":55,"latest_published_at":56},"Deals","deals",524,"2026-10-17T07:00:00.000Z",{"name":58,"slug":59,"count":60,"latest_published_at":61},"Hardware","hardware",265,"2026-10-18T13:24:23.000Z",{"name":63,"slug":64,"count":65,"latest_published_at":66},"Consumer Tech","consumer-tech",236,"2026-10-16T23:36:26.000Z",{"name":68,"slug":69,"count":70,"latest_published_at":71},"Science","science",203,"2026-10-17T12:40:00.000Z",{"name":73,"slug":74,"count":75,"latest_published_at":76},"Startups","startups",134,"2026-10-18T12:02:57.000Z",{"name":78,"slug":79,"count":80,"latest_published_at":81},"Software","software",129,"2026-10-15T22:21:06.000Z",{"name":83,"slug":24,"count":84,"latest_published_at":85},"Dev Tools",118,"2026-10-16T15:52:39.000Z",{"name":87,"slug":88,"count":89,"latest_published_at":90},"General","general",82,"2026-10-16T16:06:38.000Z",{"name":92,"slug":93,"count":94,"latest_published_at":95},"Gaming","gaming",77,"2026-10-17T16:59:39.000Z",{"name":97,"slug":98,"count":99,"latest_published_at":100},"Reviews","reviews",37,"2026-10-14T10:00:00.000Z",{"name":102,"slug":103,"count":104,"latest_published_at":105},"How-To","how-to",8,"2026-10-05T09:00:00.000Z"]