“The debate isn’t whether AI can write code. The debate is whether AI can become a responsible member of an open source community.”
The New Divide in Open Source
The open source community has survived many revolutions.
Git replaced centralized version control.
GitHub changed collaboration.
Continuous Integration automated testing.
Static analysis found bugs before humans did.
Generative AI is different.
For the first time, developers are asking a question that isn’t just technical.
Should AI-generated code even be accepted into Free and Open Source Software projects?
Depending on which project you contribute to, the answer ranges from “Absolutely” to “Not under any circumstances.”
Two Very Different Views
Supporters see AI as another developer tool.
Just like compilers, debuggers and IDEs increased productivity, AI can remove repetitive work and help newcomers understand large codebases.
Critics see something entirely different.
They argue that AI makes it almost free to produce thousands of pull requests while making maintainers spend hours reviewing work that contributors themselves may not fully understand.
The real question becomes:
Has writing code become cheaper than reviewing it?
What Linus Torvalds Thinks
Linus Torvalds has consistently argued that AI should be treated as a tool rather than an ideology.
If AI helps developers produce better software, use it.
If it creates noise, improve the workflow rather than banning the technology.
The Linux kernel still judges patches on technical merit, not on whether an AI assistant was involved.
“AI is a tool, just like other tools we use.” — Linus Torvalds
Why Some Communities Are Saying No
Projects such as Zig, QEMU, and NetBSD have adopted much stricter policies.
Their concerns aren’t about whether AI can write working code.
They’re about review cost, legal uncertainty, and preserving the learning culture of open source.
Daniel Stenberg’s Warning
Daniel Stenberg, creator of curl, has seen firsthand how AI-generated security reports can overwhelm maintainers.
“The better the crap, the longer time and the more energy we have to spend.”
The issue isn’t simply bad reports—it’s that AI can generate convincing bad reports that still require careful investigation.
Simon Willison’s Middle Ground
Simon Willison offers a balanced perspective.
Use AI if it helps.
But never submit code you don’t understand.
His rule is simple:
Never commit code you cannot explain.
The Real Scarcity Isn’t Code
AI can generate code at incredible speed.
But maintainer attention, review time, and trust remain limited.
Open source isn’t running out of code.
It’s running out of experienced reviewers.
My Take
The important distinction isn’t between developers who use AI and those who don’t.
It’s between developers who remain accountable and those who outsource responsibility.
AI can write code.
Only humans can own it.
And in open source, ownership has always meant responsibility.
Further Reading
- Linux Kernel Mailing List
- Simon Willison’s Blog
- Daniel Stenberg’s Blog
- Zig Language News
- QEMU Contribution Guidelines
- Apache Software Foundation AI Guidance
- Open Source Initiative (OSI)