AI cuts coding time, but it's also filling open source's inbox with crap

A recent survey from the Linux Foundation reveals that while generative AI is cutting coding time for open source developers, it is also inundating maintainers with a significant amount of low-quality AI-generated pull requests. Nearly a quarter of incoming code changes are deemed low-quality AI contributions, causing a surge in issues related to AI-generated submissions, false positives, and duplicate reports.

Linus Torvalds, a well-known advocate of AI tools, acknowledges their potential but remains critical of low-quality submissions. The foundation asserts that projects require clearer policies and advanced tools to manage the influx of AI-generated contributions. Currently, just over half of the maintainers believe that GenAI has improved community trust, while 71 percent see it lowering barriers to entry.

However, the benefits of AI are not evenly distributed. For complex tasks, 21 percent of developers report that AI has extended coding time. Moreover, only 35 percent of projects have formal policies allowing the use of GenAI, while 32 percent lack any policy altogether, creating ambiguity within the community.

Hilary Carter, SVP of Research at the Linux Foundation, emphasizes that open source projects must adapt to the growing number of AI-generated contributors. This includes setting clear expectations for GenAI use, imposing responsibility on contributors for testing and disclosure, and providing maintainers with better tools to handle the influx of pull requests.

Large projects face a more significant challenge, with an estimated 31 percent of pull requests classified as AI slop. The workload goes beyond coding changes, as seen with the curl project, which announced moving low and medium severity flaws to public view due to an overwhelmed security team. Project maintainers are feeling stretched, overloaded, and close to their limits.

The Linux Foundation's Carter stresses that addressing these fundamentals is crucial for harnessing AI's potential to expand participation without overburdening maintainers or increasing costs.