MONDAY, NOVEMBER 16, 2026 Tech news, minus the noise.
N News Cast
LATEST
Home  /  Dev
Dev

Linux 7.4-rc2: The AI Patch Flood May Finally Be Peaking

Linus Torvalds released Linux 7.4-rc2 on November 8, and AI-written fixes drove most of the progress. In the eBPF subsystem, fixes now arrive faster than new bug reports. The patch flood may have peaked.

Linux 7.4-rc2: The AI Patch Flood May Finally Be Peaking
Source: OpenClipart-Vectors from Pixabay. Based on Tux design by Larry Ewing ([email protected]) and The GIMP. (CC0) via Wikimedia Commons

Linus Torvalds released Linux 7.4 rc2 on November 8, the second release candidate for the next kernel, and his release note carried a small piece of good news. The flood of AI-generated bug reports and patches that has been drowning kernel maintainers may finally be peaking. In the eBPF subsystem, fixes are now landing faster than new reports arrive.

That is a narrow claim about one subsystem, and Torvalds said so himself. But it is the first sign that the AI era of kernel development might be turning from chaos into routine, and it comes from the most skeptical audience in software.

What Linux 7.4 rc2 actually contains

The 7.4 cycle is big by every measure. The first release candidate, 7.4-rc1, contained roughly 19,671 commits, a record for the kernel, after Torvalds had to hold some code back because he was traveling and the merge window was already the largest in Git history. Linux 7.4 rc2 carries the deferred code, plus bug fixes and cleanups across the tree.

AI is a large part of that volume. The cycle saw 3,008 AI-assisted non-merge commits, which is 2.4 times the figure from the previous cycle. These are changes where AI tools helped find a bug or propose a fix, with humans still required to review and sign off. The surge began after Anthropic launched its Mythos bug-finding model in April 2026, which made it dramatically easier to find flaws in large codebases.

The scale of the flood extends beyond Linux. Microsoft fixed a record 974 flaws in September, its third record month this year after 622 in July and 206 in June. Oracle addressed 1,449 flaws in a single month. The whole industry is finding more bugs than ever, because AI is good at finding them, and the fixing capacity has been struggling to keep up.

Torvalds also merged a backlog of automated Coccinelle cleanup patches that touched thousands of files with simple one-line fixes, and the release carries a FAT32 optimization and AMDGPU work connected to HDMI 2.2. The final stable release is targeted for December 20, 2026.

The eBPF milestone, and why it matters

The encouraging data point comes from Alexei Starovoitov, co-creator and lead developer of eBPF, the technology that lets small sandboxed programs run safely inside the kernel. Starovoitov reported that on the BPF side, fixes are now coming in faster than new bug reports. Torvalds quoted him directly: the fixes are arriving faster than the reports, and that sounds lovely if true.

Torvalds added the honest caveat in the same breath. Most other areas have not reached that point yet, but it is at least a sign that the tide may be starting to turn. Do not hold your breath, but maybe cross your fingers.

The reason this matters is that eBPF is not a backwater. It is the machinery behind modern networking, observability, and security tooling on Linux, and it is exactly the kind of complex, subtle code where AI-generated bug reports used to be mostly noise. If the fixes there are outpacing the reports, it means the tooling got better, the reviewers got faster, or both.

It also matters because kernel maintainers are the hardest audience in the world to impress. They spent the last year complaining, loudly and publicly, about being buried in low-quality AI-generated patches. A public acknowledgment from Torvalds that the tide may be turning is a meaningful signal that AI-generated patches are now landing in critical infrastructure at volume, and surviving review.

What developers should take from this

The practical lesson is about process, not hype. Torvalds has argued that AI-driven changes should be integrated early in development, not saved up and dumped in during the stabilization stretch. The 7.4 cycle is evidence for that: the flood is manageable when the fixes flow continuously, and overwhelming when they arrive in a lump.

For developers watching the AI agent pull request numbers, where agents now start 40 percent of pull requests, the kernel is the stress test. If the most demanding reviewers in open source can absorb AI-assisted contributions without drowning, the rest of the industry can too. The tooling is catching up to the volume, and the review culture is adapting.

The same week brought TypeScript 7.1 beta and NVIDIA open-sourcing Attribution, its GPU debugging tool. The developer tooling landscape is moving fast, but the kernel story is the one with the clearest moral. AI found the bugs. Humans still decide which fixes ship. And for the first time, the humans are winning.

AS
News Cast Editorial Team

We cover AI, gadgets, startups and security — original reporting and explainers, written by humans, for humans. Got a tip? Reach us via the contact page.