A claim with its caveats still attached
The claim at the center of the coverage of Linux 7.4-rc2 is narrower than a turning point. Read the source post and it arrives attributed, conditional and local.
Torvalds released the second release candidate of Linux 7.4 on November 8. It is, he wrote, "a big -rc2 to go with the largest ever -rc1," with a diffstat dominated by automated coccinelle updates he had held back from rc1. Mask that noise out and the shape is ordinary: roughly a third drivers, a third networking and bpf, a third everything else. Then the sentence the coverage is built on: "much of this was driven by various AI fixes. I think I'll stop repeating that, and we can just take it for granted."
What follows keeps its attribution. "This time around Alexei claims that on the bpf side at least the fixes are now coming in faster than the new reports, and that sounds lovely if true." Alexei is Alexei Starovoitov, co-creator and lead developer of eBPF, the subsystem that runs sandboxed programs inside the kernel. "Claims" and "if true" are Torvalds' words, in the primary source, and they are the difference between a measurement and a report from a named subsystem developer.
One subsystem is not the kernel
Torvalds states the limit himself in the next breath: "I think most other areas haven't reached that point yet, but it's at least a sign that we _may_ be starting to see the tide starting to turn. Don't hold your breath, but maybe cross your fingers we'll be getting there in the not entirely distant future."
That is the whole evidentiary position. No count of bpf reports. No count of bpf fixes. No statement of how many reports were valid, fixed, or duplicates, and no window over which the two rates were compared. If the relationship held and persisted, it would matter: a sustained excess of resolved, comparable reports over new valid reports could reduce an existing backlog. The post supplies neither that accounting nor evidence that the relationship will persist, and it says nothing about drivers, filesystems or the rest of the kernel.
Why the flood built up
The Register, which reported the rc2 post on November 9, ties the surge in flaw-finding to AI tools becoming widely usable this year, pointing to Anthropic's Mythos model, launched in April, in its framing. The volumes it reports elsewhere are large: Microsoft addressed a record 974 flaws in September, its third record month of the year, and Oracle addressed 1,449 flaws in a single month. Those are The Register's reported counts for other vendors' patch loads, not kernel numbers, and patch counts alone do not establish what caused them. They describe the environment kernel maintainers have been working in while AI-assisted reporting has grown.
What to watch in rc3
The next release candidate, expected in about a week on the kernel's usual cadence, may not resolve the uncertainty either. What would broaden the picture is specific: counts of new valid reports, of fixes resolving those reports, and of the outstanding backlog, over a consistent window, repeated for other subsystems. Repeated qualitative claims would widen the anecdote; they would not settle it.
Until then: Torvalds relayed a qualified claim about bpf and suggested that most other areas had not reached the same point.
Sources
- [1] Linus Torvalds, "Linux 7.4-rc2", Linux Kernel Mailing List archive, November 8, 2026 — all quotations, read in full.Read source
- [2] The Register, "Linus Torvalds sees signs AI-prompted patch flood has peaked", November 9, 2026 — the Mythos framing and the Microsoft and Oracle patch counts.Read source
- [3] XenoSpectrum, "Linux 7.4 Sees AI-Tagged Commits Jump About 2.4x", November 2026 — the 3,008 versus 1,257 tag tally and the tag's abolition during the window.Read source