A Linux Foundation survey finds that developers estimate large time savings from generative AI, while open-source maintainers face more low-quality proposed changes. Respondents put the average reduction in code-authoring time at 52 percent and the poor-quality share of incoming changes at about a quarter. Formal project rules on AI use remain uneven, and complex tasks sometimes take longer.
Research by the Linux Foundation, the organization supporting open-source software projects, examines how generative AI changes development work. Survey respondents estimated an average 52 percent reduction in code-authoring time. The results vary with the task: for highly complex work, 21 percent said AI made the job take longer. These are participants’ assessments rather than independently timed experiments covering every project.[1], [2]
Maintainers estimated that about a quarter of incoming pull requests, the proposed changes submitted to a project, were low-quality AI output. That estimate reached 31 percent in large projects. Maintainers described lower barriers to contributing, alongside disagreement about improvements in community trust. The separate tasks of preparing changes and reviewing them leave maintainers handling the growing flow of submissions.[1]
Only 35 percent of projects had a formal policy permitting generative AI use. Another 32 percent had no policy and no plans for one. Linux Foundation research executive Hilary Carter called for clear expectations, contributor responsibility for testing and disclosure, and better tools for maintainers. Carter argued that broader participation needs support that avoids overloading the people who review and maintain projects.[1]
