AI Cuts Coding Time by 52%, but Open-Source Maintainers Pay the Price

Generative AI is changing how open-source software is produced, but its impact is not limited to faster coding. In its report, Navigating Open Source Software Development in the Generative AI Era, the Linux Foundation found that respondents estimated a 52% average reduction in time spent on coding tasks. At the same time, maintainers are dealing with a larger flow of contributions that still require careful evaluation.

Key findings

• 59% of maintainers said their projects had received more code contributions, rising to 69% among large projects.
• Maintainers estimated that about 25% of pull requests were low-quality contributions, with the figure reaching 31% in large projects.
• 71% agreed that generative AI lowers the barrier to open-source participation, while 53% disagreed that it improves community trust.
• Only 35% of projects had a formal policy covering AI-assisted contributions; 32% had neither a policy nor plans to create one.
• For complex tasks, 21% of respondents said generative AI actually increased the time required.

The term “AI slop” should be read carefully. The figures are estimates from surveyed maintainers, not an audit of every pull request on GitHub or another platform. They also do not prove that every low-quality contribution was generated by AI. The broader issue is that AI reduces the cost of proposing a change, while checking correctness, architectural fit, security implications, and maintenance requirements still demands human attention.

Where the time savings go

The saved time is not simply being converted into more code. Some 66% of respondents said they used the gains to address technical debt, 63% to build new features, and 58% for learning and skills development. This suggests that AI can improve how developers allocate their time, but individual productivity gains do not automatically produce equivalent gains across the whole collaboration process.

AI also lowers the technical barrier for newcomers without eliminating the need to understand project conventions, explain design decisions, or maintain changes after submission. Forty-six percent of maintainers said the need to guide new contributors had increased. Community work may therefore be shifting from teaching basic implementation toward validating code, explaining failures, and managing contribution quality.

Governance must catch up

As AI-assisted code becomes normal, disclosure, testing, and a contributor’s ability to explain a change may directly affect review costs and trust. The Linux Foundation recommends that projects define clear policies covering AI use, testing, documentation, and disclosure. Related OpenSSF guidance emphasizes that contributors should understand and test what they submit; maintainers should not be left to carry the entire security-review burden.

The report does not show that AI inevitably harms open-source quality. It shows that productivity needs a broader definition. In addition to coding speed and agent task completion, projects may need to track acceptance rates, review time, defect-remediation costs, and long-term maintenance load. Code generation can be accelerated, but turning generated code into dependable software still requires verification.

Source: InfoQ 中文