Linus Torvalds cut Linux 7.4-rc1 on Sunday, 1 November 2026, closing the merge window on schedule even after spending much of it on his annual vacation. The headline number is the size: almost 20,000 commits, against a more typical 15,000. The shortlog alone runs 24,000 lines. But the number that matters for anyone running Linux in production is not the size. It is the timing of where those patches came from, and what Linus thinks that timing means.

An rc is a testing signal, not an upgrade

First, the operator framing, because release candidates get misread every cycle: rc1 means feature merging is closed and stabilization has begun. It is not a release you deploy. The sensible test targets are a virtual machine, a spare box, or a laptop you can afford to boot into a known-good kernel from. Keep your working kernel installed, test the rc alongside it, and report regressions early, while there is still time to fix them before stable. If you want the full drill on reading a candidate the way its author intends, our guide on how to read a kernel release candidate like Linus does still holds up.

The expected timeline from here is familiar: weekly release candidates, rc2 through rc7 or so, with a stable 7.4 pencilled in for late December. There is also a widely shared expectation that 7.4 becomes the next long-term support (LTS) kernel. That is expectation, not announcement, and it matters because LTS years pull in extra work from vendors who want their hardware supported for the long haul.

Why Linus calls this a healthier model

The most interesting paragraph in the rc1 announcement is not about hardware at all. Torvalds writes that he is much happier seeing the flood of AI-triggered work arrive during the merge window than showing up mainly as extra fixes during the rc period. His hope, in his words, is that AI fixes are becoming part of normal development rather than a wave landing during stabilization. Hence the "much healthier model" verdict.

Read that carefully, because it is narrower than the headlines suggest. Linus is not saying AI writes the kernel now, and he is not saying AI patches sail through unreviewed. In the same note he describes the release as large and full of fixes for issues found by AI, adding that patches are more and more suggested by AI too, although they often get rewritten by humans. The healthier part is about when the work arrives: in the open merge window, where it gets the normal review and testing cycle, instead of as a pile of late fixes during the stabilization phase, when every change carries more risk.

That distinction is worth keeping in mind the next time someone claims AI has either saved or ruined kernel development. The 7.4 window suggests something more boring and more useful: AI as a bug-finding pipeline feeding the same human review process, earlier in the cycle. It is also consistent with the kernel's earlier pushback against AI-generated noise: as we covered in the kernel's rejection of AI labels, the project cares less about how a bug was found than about whether the fix is correct and reviewable.

Linus himself adds the honest hedge: part of the window's size may simply be the LTS expectation pulling work forward. So the record is not proof of an AI revolution. It is evidence of a workflow change plus a calendar effect, and the rc phase will be the real test. If rc2 through rc7 stay calm, the healthier-model claim holds. If they fill up with late AI-driven fixes anyway, it does not.

What actually lands in 7.4

Strip out the noise first. About a third of the diff is AMD GPU register offsets and bit masks dumped into huge headers, a recurring event Linus tells readers to ignore. Ignore the AMD headers and the picture looks fairly normal: about half drivers, with the rest spread across architecture updates, filesystems, networking, Rust, BPF, memory management, and tooling like perf.

The hardware highlights compiled by Phoronix from the merge window include AMD Ryzen Zen 6 preparations, AMDGPU HDMI 2.1 enabled by default, early Apple M4 support, Google Tensor G5 and Pixel 10 enablement, continued Qualcomm Snapdragon X2 work for PCs, Intel IPU8 camera support for upcoming Nova Lake laptops, virtual machine security tweaks, Raspberry Pi 10-inch display support, and performance work including Radeon iGPU improvements. Treat that list as a merge-window snapshot, not a final changelog: features can still be reverted or reworked during the rc phase, and the stable release notes in December are the authoritative list.

The one deadline hiding inside this news

The same weekend quietly closed another door. Greg Kroah-Hartman shipped Linux 7.2.15 on 1 November as the final 7.2.y release, marking the branch end-of-life. If you are still on 7.2, there will be no more fixes coming for it, including security fixes. The move is to 7.3.y, and our Linux 7.3 stable upgrade playbook walks through that upgrade with the usual precautions: verify the new kernel boots, keep the old one until you are sure, and check DKMS modules like out-of-tree graphics drivers before you commit.

What to do this week

  • Nothing, on production. rc1 is for testing, and 7.3.y remains the current stable line.
  • Test, on expendable hardware. Boot rc1 in a VM or on a spare machine, keep your known-good kernel, and exercise the hardware you actually own: graphics, Wi-Fi, suspend and resume, and any out-of-tree modules.
  • Watch the rc cadence, not the commit count. A quiet rc2 and rc3 would confirm the healthier-model story better than any headline number. A noisy one would tell you the AI fix flood merely moved.
  • Plan the December window. If 7.4 lands late December and takes the LTS mantle, distro kernels will follow in early 2027. That is the upgrade to schedule, not this rc.

The kernel just had its biggest intake ever, and its maintainer's reaction was cautious approval rather than alarm. The work arrived early, through the normal process, and now it gets the normal scrutiny. That is the story of 7.4-rc1: not a revolution, but a pipeline growing up.

Sources

  • Linus Torvalds, "Linux 7.4-rc1", Linux Kernel Mailing List, 1 November 2026: https://lkml.iu.edu/hypermail/linux/kernel/2611.0/00796.html
  • Greg Kroah-Hartman, "Linux 7.2.15 (final 7.2.y release)", as reported by Linux News, 1 November 2026: https://www.linuxnews.net/articles/final-release-for-the-7-2-stable-series
  • Phoronix, "Linux 7.4-rc1 Released With Nice Performance Improvements, AMD Ryzen Zen 6 Preparations & AMDGPU HDMI 2.1 Default", 1 November 2026
  • The Register, "Linus Torvalds says Linux kernel is out of 'stabilization phase' and using 'much healthier model' for AI contributions", 1 November 2026