Linux Foundation Report: AI Makes Coding 52% Faster, but a Quarter of PRs Are Low-Quality Contributions

On October 20, the Linux Foundation released its latest research report, "Navigating Open Source Software Development in the Generative AI Era," examining the impact of generative AI on open source software development, code contributions, and community maintenance.

The report presents a set of data with a striking contrast: respondents estimated that after using generative AI, the average time spent on code-writing tasks fell by 52%; at the same time, open source project maintainers estimated that about 25% of the Pull Requests (PRs) they receive are low-quality contributions, referred to as "AI slop." In large open source projects, this proportion rises further to 31%. 59% of maintainers said that since the rise of generative AI, the number of code contributions their projects receive has increased; in large projects, this figure reaches 69%.

1 AI writes code faster, but complex tasks do not necessarily save time

This Linux Foundation study was published by LF Research, sponsored by Red Hat, and supported by Futurewei, OpenSSF, and LF AI & Data.

In its official news release, the Linux Foundation said that generative AI's improvement of open source development efficiency already covers multiple stages, far beyond code generation. Across the six categories of open source development activities covered in the study, respondents reported varying degrees of time savings. Among them, code writing saw the largest average reduction in time, reaching 52%; even communicative tasks such as proposal discussions achieved time savings of about 19%.

However, this efficiency gain does not apply to all scenarios. For highly complex tasks, 21% of respondents said that using generative AI actually required more time. This means that the efficiency advantages AI shows in generating code and handling routine development tasks may not directly translate into overall efficiency gains for complex software engineering tasks.

The time saved has also not all been converted into more code output. 66% of respondents said they invested the efficiency gains in addressing technical debt, 63% in developing new features, and 58% in learning and skills improvement.

2 One in four PRs is considered "AI slop," increasing pressure on maintainers

Compared with code-writing efficiency, what is more noteworthy in the report is the pressure the open source maintenance stage is under.

59% of maintainers reported an increase in the number of code contributions, but these new contributions cannot all be directly converted into project value. According to maintainers' estimates, about one quarter of the PRs received are low-quality content. In large projects, this proportion reaches 31%.

"AI slop" here generally refers to AI-assisted contributions that are of poor quality, lack sufficient verification, or require maintainers to invest a great deal of effort just to determine their value. It should be emphasized that this proportion comes from estimates by survey respondents, not from objective statistics obtained by researchers after item-by-item inspection of PRs on platforms such as GitHub, nor does it mean that all low-quality PRs can be confirmed to be AI-generated.

But the maintenance pressure reflected in this set of data deserves attention.

In the past, developers had to spend considerable time understanding the project structure, writing code, and running tests before they could submit a PR. With the spread of AI coding tools, the threshold for generating code and submitting changes has dropped markedly. For contributors, proposing a code change may take only a very short time; for maintainers, determining whether the code is correct, whether it fits the project architecture, and whether it introduces security issues still requires effort.

Especially in large open source projects, maintainers also need to deal with issues such as code compatibility, long-term maintenance costs, contributor communication, and project governance.

The Linux Foundation noted in the report that maintainer burnout has become one of the most prominent concerns for large open source projects.

This change has also introduced a new dimension for measuring the productivity of AI coding: if the cost of code generation falls sharply, but the cost of code verification, review, and maintenance does not fall in tandem, then the efficiency gains may be redistributed among different roles.

3 71% believe AI has lowered the barrier to open source, but community trust has not risen in tandem

At the same time, however, 53% of maintainers disagree with the statement that "generative AI has improved community trust."

Collaboration in open source communities depends on many factors beyond code. Maintainers not only need to know whether a piece of code can run, but also need to judge whether the contributor understands the changes, can explain design choices, and has the ability to continue maintaining the code if problems arise later.

The report also found that 46% of maintainers believe generative AI has increased the need to mentor new contributors, while 32% believe this need has decreased.

This also shows that although AI has lowered the technical barrier for new developers to enter open source communities, newcomers still need to understand project norms, collaboration processes, and software engineering practices. Some of the work originally used to guide code writing may be shifting toward code verification, issue explanation, and contribution quality management.

4 Only 35% of projects have a formal AI contribution policy

The Linux Foundation survey results show that only 35% of projects have a formal policy explicitly permitting the use of generative AI in code contributions. Another 32% of projects have neither a relevant policy nor plans to develop one. On the contributor side, 61% of respondents said they conduct detailed review and complete functional testing before submitting AI-generated code. But disclosure of AI usage still lacks consistent standards.

This creates a practical problem: when maintainers receive a code contribution, they may not clearly know which parts were generated by AI, what verification the contributor performed, and whether the code changes were fully tested. For open source projects that rely on volunteer maintainers, this information asymmetry may further increase review costs.

The Linux Foundation recommends that open source projects establish clear policies on AI-assisted contributions, set requirements for code testing, documentation, and disclosure of AI usage, and improve contribution attribution and verification mechanisms. A relevant person in charge at OpenSSF also emphasized in an official statement that AI can speed up code writing, but contributors still need to understand and test what they submit, and maintainers should not bear all security review responsibility alone.

5 Who bears the verification cost of AI coding's efficiency dividend?

Over the past two years, discussion of AI coding has largely centered on code generation speed, task completion rates, and developer productivity. From code completion to coding agents capable of autonomously modifying repositories, running tests, and submitting PRs, automation capabilities are steadily increasing.

This Linux Foundation report offers another perspective: once the speed of code production increases, the bottleneck in the software engineering process may gradually shift toward review, verification, and maintenance.

This change is especially evident for open source projects. Code contributions to open source projects can come from any developer around the world, but maintainer permissions are usually concentrated in the hands of a relatively limited group of core maintainers. AI tools can enable more people to quickly generate code, but they cannot guarantee that maintainers' time and energy will grow at the same rate.

The problem is that as AI-generated code becomes the norm, open source communities need to establish verification and governance mechanisms that match it. In the future, when measuring the value of AI coding, in addition to how much coding time developers save and how many tasks agents complete, attention should also be paid to the proportion of code ultimately accepted, the time required for review, the cost of defect fixes, and the long-term maintenance burden.

Link to the Linux Foundation report:

This article comes from the WeChat public account "InfoQ" (ID: infoqchina), author: Cai Fangfang, published by 36Kr with authorization.