Linus Torvalds: AI Reviewers Are Now Writing More Linux Fixes Than Any Human Team

Linus Torvalds doesn't exaggerate. When he describes something as unusual, he means it in a precise, technical sense — not as a rhetorical flourish. So when he wrote that Linux 7.2-rc7 is "one of the biggest rc6s we've had in years," that's a data point, not hyperbole.

On August 9, 2026, Torvalds released Linux 7.2-rc7 with more than 400 fixes signed off by over 230 contributors. For context: a release candidate at rc7 — one week before stable — typically contains a handful of last-minute fixes. 400+ is not typical. It is, as Torvalds put it, "the new normal."

The reason, he explained plainly: AI review tools.

How This Actually Works

This is not AI writing kernel code. That distinction matters. What Torvalds is describing is AI acting as a code reviewer — automated tools that scan the kernel source continuously and surface bugs, edge cases, and potential vulnerabilities that human reviewers didn't catch, or caught and deprioritized.

TechRadar's coverage notes that the workflow remains human-controlled: AI tools flag potential issues, human maintainers triage those reports, and kernel policy still requires a human to review and cryptographically sign off on every patch before it lands. AI finds; humans decide; humans write.

The result is a backlog effect. These tools are surfacing bugs that existed for years — some of them dormant issues that nobody had time to track down during a normal review cycle. When those reports now arrive in volume, maintainers are processing them in volume. Hence the unusually large rc7.

Torvalds' Reaction

"I can't say that I'm exactly thrilled about the size of this all," he wrote in the release announcement. But he acknowledged the reality: "It is what it is: the new normal with a lot of fixes, many of them due to review by various AI tools."

That's a characteristically Torvalds sentence. Not an endorsement, not a complaint — just an observation about what is true. The tools are doing something useful. The side effect is that "useful" means more work, more scrutiny, and larger release candidates from now on.

Betanews covered the AI review dynamic in Linux 7.2 in more depth, noting that AI-assisted code review is now broadly integrated into the kernel's development pipeline — not as a formal policy decision, but as an emergent reality driven by individual contributors and subsystem maintainers adopting the tools independently.

The Broader Implication

Linux 7.2 stable is expected around August 16, 2026. If the pattern holds — and Torvalds' language suggests he expects it to — future kernel releases will continue to ship heavier than they used to, with more fixes per cycle, surfaced by a class of reviewers that doesn't sleep, doesn't have a day job, and doesn't need to deprioritize low-severity bugs to focus on the important stuff.

That's actually the best version of how AI integrates into a mature open-source project: not by replacing the humans who understand the system deeply, but by giving those humans a more complete picture of where the problems are. The Linux kernel has 30+ years of accumulated code. No human team was ever going to review all of it. AI tools apparently can — and they're finding things.

The question now is whether other major open-source projects follow the same path, and whether the maintainer communities handling the resulting surge of reports have the capacity to keep up. For the Linux kernel, Torvalds' answer seems to be: we don't have a choice, and that's fine. The work is real. Better to do it than not.

Linux 7.2 will likely be the most thoroughly reviewed major kernel release in the project's history. That's not a bad thing to be.