Generative AI is changing how open source software gets written, and the picture has two sides. The Linux Foundation estimates that AI can cut open source coding time in half. On the other side, maintainers estimate that about 25% of incoming pull requests are low-quality output, what the industry calls "AI slop."
The study, published on October 20, 2026, collects answers from people working on open source projects. The productivity gain applies across all kinds of development tasks, for contributors and maintainers alike. Some 59% of maintainers report growing AI-based contributions, and 71% agree that GenAI lowers the barriers to entry for newcomers.
Image: Wikipedia
The downside falls on the maintainers. A low-quality pull request still has to be read, evaluated and rejected or corrected, and review time grows faster than the number of useful contributions. The same imbalance has surfaced on the Linux kernel, where Torvalds' supersized updates have become the new normal and security triage struggles to keep up. GitHub shows the same strain, where monthly commits doubled in four months driven by coding agents.
The Linux Foundation expects the share of low-quality contributions to keep rising as more newcomers and users have AI prepare patches for them. The question is already dividing projects: Debian is weighing whether to ban or regulate LLM use in its contributions, while NetBSD banned it back in 2024.
Torvalds and AI: the kernel case
Torvalds himself has shifted his stance. After several revisions of his position on AI-assisted contributions, he posted his first LLM-assisted patch, aimed at an old cache contention problem that appears when many CPUs read the same file. It touches shared libraries, databases and web servers, an issue several developers had tried to fix over the years without a fully satisfying solution.
No comments yet. Be the first!