The Linux Foundation summed up results a survey of 234 open source project maintainers and developers assessing the impact of AI tools on development. 59% of maintainers surveyed noted an increase in the number of changes submitted, but about 25% of pull requests submitted were considered low-quality AI-slop. If we consider only large projects, the increase in the number of changes was noted by 69% of respondents, and the share of AI-slop among pull requests is estimated at 31%.
Among other trends:
- AI has made it possible to speed up the execution of tasks related to writing code,
for 52% of respondents, and discussion of proposals for 19%. But 21% of respondents said that the use of AI has only slowed down the solution of very complex problems. - 66% of respondents spend the saved time on eliminating technical debt, 63% on creating new functionality (63%), and 58% on training.
- 71% of maintainers believe that AI lowers the barrier to entry into development of open source projects, but 53% believe that the use of AI undermines trust with the community.
- 35% of projects clearly outlined the possibility of using AI in the rules, and 32% do not have such rules and do not plan to introduce them.
- 61% of development participants said that before submitting changes they conduct a detailed review and full testing of AI-generated code, but do not always mention that they used AI.
- 46% of maintainers believe that AI increases the need for mentors for new participants, and 32% believe that it reduces it.
- 52% use generative AI tools daily, 23% almost every day, 20% weekly, 3% once a month, 2% less than once a month.
- 20% of respondents do not use AI, 58% use AI for non-development tasks, and 59% for development-related tasks.
/Reports, release notes, official announcements.