Senior Linux kernel maintainer cuts back reviews amid flood of AI-generated code
Lorenzo Stoakes, of ARM, a kernel maintainer since 2014, told the community he will stop reviewing code in three critical areas because he can no longer keep up with the volume of AI-generated patches.

- Reading
- 5 min
Lorenzo Stoakes, of ARM, a kernel maintainer since 2014, told the community he will stop reviewing code in three critical areas because he can no longer keep up with the volume of AI-generated patches.
The warning from one of the kernel's pillars
Lorenzo Stoakes, an ARM engineer and Linux kernel maintainer since 2014, told the community he will give up most of the reviewer duties he had held for years. According to CNews, citing a report from TechPowerUp, Stoakes will stop reviewing code in the areas of memory management (memory management), the core (core), and the project's Rust subsystem.
He is not leaving the kernel: he remains the formal maintainer of the parts for which he was already directly responsible. What changes is the volume of voluntary review work he did outside those areas, a task that consumed growing amounts of time as the number of AI-generated contributions surged in 2026.
The number behind the decision
Stoakes' justification has a concrete data point: according to CNews, by the end of 2026 about 25% of the pull requests submitted to the Linux kernel were already generated by artificial intelligence, rather than by a human developer writing the patch line by line. That means that, out of every four proposed code changes to the kernel, one originated from an AI model.
According to the report, Stoakes argued that without the necessary amount of time he can no longer guarantee a quality review in areas that are not his direct responsibility as a maintainer. Instead of simply dropping the flow of patches he received, he chose to redistribute that load among four other people in the community, avoiding concentrating the problem on a single reviewer. He also publicly apologized to the rest of the group, saying he needs to ease his inbox given current contribution trends.
This partial step back carries weight because Stoakes is no ordinary contributor: he is the author of the book The Linux Memory Manager, a reference on how the kernel manages memory, and has contributed to Linux since 2014, for almost 13 years, having been actively involved in developing the memory management subsystem.
Torvalds had already been complaining about the same problem throughout 2026
Stoakes' case is not isolated. Linux creator Linus Torvalds spent much of 2026 publicly commenting on the effects of generative AI on the kernel's workflow:
- In March 2026, he said he was surprised by the jump in the volume of patches proposed for Linux 7.0 RC6, attributing the increase to the growing popularity of AI tools among programmers.
- In May 2026, he said the mailing list dedicated to kernel security became practically unmanageable because of a huge flow of AI-generated vulnerability reports: the main problem was not the low quality of the findings, but the massive duplication of reports about the same flaws.
- Also in May, he publicly criticized AI-generated bug reports for disrupting the kernel's development process.
Interestingly, Torvalds' stance on AI is not one of outright rejection. In January 2026 he admitted he doesn't know how to program in Python and said that, when he needs code in the language, he now generates it with AI instead of looking for ready-made snippets on Google, a habit he already had before. In November 2025, he had praised vibe coding (generating code via AI instead of writing it) as a legitimate practice, but added an important caveat: critical components should not be left in the hands of someone who only does vibe coding.
The contrast is telling: the project's creator sees value in AI as a personal tool, but acknowledges that the volume of automated contributions is overwhelming the kernel's collective review and security processes.
What this changes for those who maintain or contribute to open source code
The Stoakes episode is a concrete symptom of a problem that any open source project maintainer, Brazilian ones included, is likely already feeling: PR review stopped being a bottleneck of human volume alone and became a bottleneck of AI volume as well. Models generate patches, bug reports, and even entire series of fixes in minutes; reviewing each one carefully still takes the same amount of time it always did, because human judgment work doesn't scale along with it.
For those who maintain repositories on GitHub or GitLab in Brazil, a few points from this case serve as practical reference:
- Split the load, don't pile it on one person. Stoakes avoided dumping the entire flow onto a single successor; he spread it across four reviewers. Smaller projects can replicate this logic by creating narrower areas of responsibility instead of a single maintainer trying to cover everything.
- Explicit review scope. He defined, in writing, which areas he would stop reviewing. This avoids the ambiguity of PRs that sit waiting for review indefinitely with no one taking ownership.
- Duplication is the silent enemy. The problem Torvalds reported on the security list wasn't low quality, it was duplicated volume. Automatic duplicate-triage tools, already common in large trackers, tend to become mandatory, not optional, in projects that receive external AI-driven contributions.
What remains open
Neither CNews nor TechPowerUp, the report's original source, details whether the kernel will adopt any formal policy for identifying or limiting AI-generated PRs, such as requiring explicit disclosure of AI use upon submission, a practice that already exists in some smaller projects. There is also, as of this publication, no official position from the Linux Foundation on the matter, nor an announced plan to structurally address the growth of these 25% of automated contributions. Stoakes' case reinforces that the discussion is no longer hypothetical: real maintainers are reducing the scope of their work because of it, and the community still has no ready collective answer.
Translated from the Brazilian Portuguese original · Read the original
PyPI redesigns project pages to expose security signals for each package
The official Python package repository completed, in November, the overhaul announced in July: a dedicated security tab shows provenance, attestations, and warnings of possible compromise for each release.





