Linux 7.2-rc4: AI-Discovered Bug Fixes Become the "New Normal"

The fourth release candidate of the Linux 7.2 kernel, “Linux 7.2-rc4,” has been made available. According to Michael Larabel of Phoronix, who reported on it on July 19, Linus Torvalds described this rc4 release as the “new normal,” indicating that the detection of bugs by AI/LLMs is becoming a standard practice in kernel development.

Scale and Characteristics of rc4

In his announcement of 7.2-rc4, Torvalds noted, “This week has felt like people were on summer vacation, but looking at the stats, I was obviously wrong. Everything is perfectly normal.” He continued, “Of course, ‘normal’ now refers to the ‘new normal,’ where there is a substantial amount of fixes overall. This is not a small rc4.”

While most of the fixes consist of minor changes of one to several lines, which is typical for an rc4 release, the volume of changes is by no means small. Notably, in the can/bcm subsystem, fixes identified by syzbot and sashiko stand out, along with some additional self-tests included in this release.

Torvalds further explained, “There are a few larger fixes, but nothing unusual.” While there is some noise related to AMD GPUs, the majority of the activity involves code relocations and additional comments, thus differing from the large-scale header drops seen in previous years.

Establishing AI/LLM in Bug Detection

This rc4 release includes numerous fixes for bugs discovered by AI and large language models (LLMs). During the Linux 7.2 development cycle, AI/LLMs have become a routine part of the workflow, autonomously detecting kernel code issues and proposing repair patches.

Torvalds’ description of this as the “new normal” reflects the acknowledgment that AI-supported kernel development is no longer an extraordinary occurrence but a standard process. Previously, human kernel developers would spend significant time identifying bugs through code reviews and testing. Now, AI can pinpoint issues in a much shorter timeframe.

New Features in 7.2 and Upcoming Schedule

The stable release of Linux 7.2 is expected in August. The earlier release candidates have already introduced notable features such as AMDGPU HDMI 2.1 FRL support and Cache Aware Scheduling in 7.2-rc1, along with enhanced BPF JIT hardening in 7.2-rc2.

Additionally, enhancements to Cache Aware Scheduling have been shown to boost MySQL performance by up to 360%, marking a significant improvement for enterprise users.

7.2-rc4 consolidates bug fixes on top of these new features, signaling steady progress in the stabilization phase of development.

Transformative Impact of AI on Kernel Development

The influence of AI/LLM-based code analysis on kernel development extends beyond simply improving the efficiency of bug fixes. AI has demonstrated its ability to uncover edge cases and subtle bugs tied to specific architectures that human developers might overlook, potentially enhancing the overall quality of the kernel.

At the same time, a new workflow has emerged wherein humans now verify patches generated by AI. Torvalds’ specific mention of “nothing unusual” in this release suggests an awareness of the potential risks of unexpected side effects stemming from AI-generated patches—a cautionary note against complacency as AI becomes more integrated into the development process.

Editorial Opinion

The growing adoption of AI/LLM for bug detection in Linux 7.2-rc4 showcases its contribution to enhancing productivity in kernel development. Over the next three to six months, similar applications of AI in other large-scale open-source projects are likely to be explored. In security-critical code areas, AI could accelerate the identification of vulnerabilities that are challenging for human reviewers to detect.

In the long term, the role of kernel maintainers may undergo significant changes. Instead of focusing primarily on writing code, their responsibilities may shift toward evaluating and integrating AI-generated patches. Over the next 1–3 years, this could significantly alter the division of labor within open-source development. For subsystems where a shortage of maintainers is a bottleneck, AI-driven automation could address these challenges and accelerate development.

A key question remains: how many AI-detected bugs should be merged without human review? While this decision currently rests with Torvalds, as the scale of AI involvement grows, clear standards for evaluating AI-generated patches will become essential. The kernel community must decide when and how to define quality benchmarks for AI-contributed patches, a decision that will shape the development process over the next year.

References

• “Linux 7.2-rc4 Released: Looking Like The “New Normal"", by Michael Larabel — Phoronix, 2026-07-19T21:16:16.000Z (ARR)
• Source URL: https://www.phoronix.com/news/Linux-7.2-rc4-Released