May 26, 2026 Linus Torvalds is signaling a tougher stance on incoming code contributions to the Linux kernel. The shift comes after he warned that trivial and poorly timed submissions, some generated or triggered by AI-assisted code reviews, are disrupting the stability of the release process.
The comments came in his weekly update announcing the fifth release candidate (rc5) for Linux kernel version 7.1, where Torvalds noted that this stage of development is supposed to focus on refining stability, not introducing unnecessary changes. Instead, he said the current cycle has been weighed down by what he described as excessive and largely insignificant updates.
“To the surprise of absolutely nobody by now, rc5 is pretty big,” Torvalds wrote, pointing out that the volume of changes is larger than what is typical at this stage. While many of the updates involve minor fixes to drivers, he made it clear that their cumulative impact is a problem.
The Linux kernel follows a structured release cycle. After an initial two-week merge window where new features are introduced, the process moves into a series of release candidates—rc1 through rc7—designed to stabilize the code. By rc5, the expectation is that only critical fixes, particularly regressions, should be addressed. Instead, Torvalds says contributors are continuing to submit non-essential patches.
“These things are ‘fixes’, sure,” he wrote, “but at the same time a lot of them are simply so irrelevant that I think they’d be better off in a linux-next tree and get merged during the merge window.”
He explicitly called out AI-assisted workflows as part of the issue. According to Torvalds, several of the recent submissions were triggered by AI code review tools, which are identifying issues but not necessarily prioritizing whether those fixes are appropriate for the current stage of development.
That distinction matters. Even small, low-risk fixes introduce some level of uncertainty. And when added late in the cycle, they can undermine the stability that release candidates are meant to ensure.
“Trivial fixes may be trivial, and have a pretty low chance of causing problems,” Torvalds wrote, “but ‘low chance’ is still not ‘zero chance.’”
As a result, he said he plans to become “more hardnosed” about rejecting pull requests that do not meet the threshold for urgency. His message to contributors was direct: evaluate whether a change is truly a regression or critical issue before submitting it late in the cycle.
This is not the first time Torvalds has raised concerns about AI’s impact on kernel development. Just a week earlier, he complained that AI-generated bug reports were overwhelming the project’s security mailing list, creating duplication and unnecessary workload for maintainers.
