This blog has been a little while coming, but I guess today just makes for a good time to ask this funny question, following an announcement by Sindre about closing pull requests to external contributors on all his projects.
Since the launch of ChatGPT, Generative AI has moved from chat, to the code editor (first as code suggestions as you write code, helping you write a simple function) to writing a whole application with a chat interface sitting in your code editor, down to not even seeing the code at all and working in the terminal.
This is beautiful to see. I'm sorry, but I'm not a fan of that terminal part, so I don't use it much because I just can't stand something happening somewhere and I can't see it live haha.
GenAI, I believe, is the biggest disruption/innovation of our existence, even bigger than the invention of the wheel or quantum computing.
But this disruptor is seriously changing, or should I say disrupting, the essence of one of the most important things to have ever happened to technology: open source.
I say the essence of open source.
What is the essence of open source?
The essence of open source is the people.
The people who create the technology. The technology that brings people together. That togetherness births even more innovation, technology and, ultimately, opportunities.
This seems to be under threat now.
Why do I say this?
In order to understand why I say this, it makes sense to understand this tiny thing about my perception of open source.
Open source has always been about the contributor
Open source has always been about the contributor first.
This is the reason for ideologies like good first issues and contributor guidelines; programs like Hacktoberfest, Google Summer of Code, Season of Docs and Outreachy; communities like Open Source Community Africa, LFX and MLH; and a whole bunch of other initiatives built around connecting humans to open source.
The contributor-first mantra of open source has created life-changing opportunities for people, groundbreaking innovations and plenty more good stuff that is evidently contributor-driven.
At some point, this contributor focus became so important that contributor data itself became valuable, leading to the development of open source intelligence products like OpenSauced, which was eventually acquired by the Linux Foundation.
It is on the basis of this mantra that I also got where I am in my career today. Haha, yea, my biggest flex right now is that I'm a core maintainer on a project with 20k+ stars and 2 million+ weekly downloads.
Thanks to open source, I've been able to meet people who have touched me at every level.
People who taught me things, people who reviewed my code, people who gave me opportunities, people whose work inspired me. People I probably would never have met if I wasn't contributing to some random repository on GitHub.
And this is the part of open source that I don't want us to lose.
Something changed
I think something started changing even before AI.
Open source became big.
And when something becomes big, it becomes easier to optimize for the wrong things.
Hacktoberfest is probably one of the easiest examples.
For years, Hacktoberfest was one of those moments where the open source community felt unusually alive. October came around and suddenly everybody was talking about contributing.
People were discovering repositories, maintainers were preparing issues, beginners were making their first contributions, an people were writing blog posts about their first PR.
There was excitement.
But there was also a problem.
When you create an incentive around the number of pull requests, eventually people start optimizing for the number of pull requests.
And maintainers started getting spam. Tiny documentation changes that didn't really matter, unnecessary formatting changes.
Changes that technically improved something but solved absolutely nobody's problem. Pull requests created simply because someone needed one more contribution.
This wasn't a new problem. GitHub itself has historically advised maintainers on dealing with “spammy” Hacktoberfest pull requests.
But AI has taken this problem and put it on steroids.
And Hacktoberfest has effectively acknowledged that.
For Hacktoberfest 2026, pull requests no longer count toward rewards. Their own explanation is pretty straightforward: it has become too easy to submit low-effort spam PRs, and the organizers want to reduce the burden on maintainers.
Think about that for a little bit.
An event whose entire identity was heavily associated with contributing pull requests has had to rethink the pull request itself as an incentive.
That tells us something.
And then AI entered the chat
Before AI, writing a bad pull request still required some effort.
You had to find the repository, understand enough of the project to make a change, modify the code, run it, create the PR and you had to explain what you did.
Now?
You can tell an AI:
“Find a good first issue and fix it.”
And in a few minutes, you can have something that looks remarkably like a legitimate contribution.
This creates a very strange problem.
We now have to distinguish between a contribution made with AI and a contribution made by someone who actually contributed something.
And those aren't necessarily the same thing.
I'm not against using AI to contribute to open source though, totally far from it.
I've used AI extensively myself.
If AI helps you understand a codebase faster, write tests faster, investigate a bug, improve documentation or even implement a feature, great.
The problem isn't AI-assisted contribution, the problem is AI-generated contribution without contributor intent.
There is a difference between:
“I found a problem, understood it, used AI to help me solve it and submitted the fix.”
and:
“AI, go find something I can submit to this repository.”
The first is contribution, the second is content generation wearing a contribution costume, and maintainers can feel the difference.
Maintainers are getting tired too
This is the part I think we don't talk about enough.
We have spent years telling developers:
“Go contribute to open source.”
But who is responsible for dealing with all those contributions?
The maintainer.
The same maintainer who already has a job.
The same maintainer who has to deal with issues, releases, security vulnerabilities, documentation, dependencies, users and everything else that comes with maintaining software.
Now imagine that person waking up to twenty pull requests, ten of them generated by AI, three of which are unnecessary, four of which don't follow the project's conventions and just one actually solves a real problem.
The maintainer still has to inspect all twenty.
AI doesn't just increase the ability to create code, It increases the ability to create work for other people, and I think this is one of the most under-discussed consequences of AI in open source.
So maintainers started automating the maintainer
And honestly?
I can't blame them.
We're developers.
When something becomes repetitive, we automate it.
And this is where things get really interesting.
AI is no longer just helping contributors contribute, it is helping maintainers maintain.
Take Vercel's AI SDK as an example.
@gr2m and the project team have explicitly said that their maintenance process is becoming increasingly automated, with AI handling things like reviewing new issues, triaging reports, reproducing bugs and preparing bug-fix pull requests. Their contribution guidelines now tell contributors that a high-quality issue, minimal reproduction, failing test or precise documentation feedback can sometimes be more valuable than a complete pull request.
And when you look at the repository, you can actually see this machinery operating.
Issues are being automatically classified, reproduced and in most cases turned into bug-fix PRs by automated tooling.
I have it from @gr2m today that their ai-sdk-factory software factory has reduced their bug reports from the thousands to a few hundred in count.
That is fascinating.
But it also creates a philosophical problem.
If the maintainer has an AI system that can:
- Receive an issue
- Understand the issue
- Reproduce the bug
- Identify the cause
- Write the fix
- Create the pull request
- Run the tests
- Merge the change
...what exactly is the external contributor supposed to do?
And that question is bigger than Vercel, because this is where I think the open source experience starts to fundamentally change.
The contributor is becoming optional
This is probably the uncomfortable part of my argument.
For a long time, the open source ecosystem was designed around a simple loop:
User → Contributor → Maintainer → Project
You used something, you found a problem, you contributed a solution.
The maintainer reviewed it, the project got better.
The contributor learned something, the maintainer gained another person who understood the project.
Everybody benefited.
AI is turning that into something more like:
User → Issue → AI → Fix → Project
Notice who's missing!? Yea, the contributor.
And this isn't necessarily because maintainers don't value contributors.
Sometimes it's because the economics and incentives simply make automation more attractive.
Why wait for someone to discover an issue, understand your architecture, implement a fix and submit a PR when your automated system can start working on the issue immediately?
Why spend twenty minutes reviewing a beginner's contribution when an AI agent can produce a technically correct patch in seconds?
Why maintain an open contribution surface when the cost of processing external contributions has become higher than the value those contributions provide?
And suddenly, the contributor isn't the centre of the system anymore.
They're an optional input.
And Sindre's announcement is interesting for exactly this reason
This is why Sindre's announcement caught my attention.
Sindre has been one of the most prolific people in the JavaScript open source ecosystem for years.
His projects have introduced countless developers to open source, Node.js and the wider JavaScript ecosystem.
So seeing a maintainer of that scale say, I'm closing the door to external pull requests, feels significant.
Not because Sindre is wrong, I mean... I actually understand the decision.
Maintaining software at that scale is a completely different problem from contributing to it.
If the cost of reviewing contributions is becoming greater than the value those contributions bring, closing the contribution channel can be a rational maintenance decision.
But that's exactly why it worries me.
Because if some of the projects that helped define the open source contribution experience begin saying:
“We don't need your pull request.”
and other projects say:
“Our AI can handle this.”
and contributors are increasingly submitting AI-generated work that maintainers don't want to review anyway...
then what happens to the contributor-first open source ecosystem?
Maybe open source is becoming product-first
This is where I think the bigger shift is happening.
Open source used to feel like a place where people built software together.
Increasingly, it feels like a place where software is built in public.
Those two things sound similar, but they're actually not.
Building software in public doesn't necessarily mean building software with the public.
A company can publish its source code, accept issues, publish releases and still have very little human participation in the actual development process; perhaps that's perfectly fine.
Maybe this is simply the next evolution of open source.
Maybe the future looks like highly automated projects where a tiny team of maintainers operates software with the help of fleets of AI agents.
Maybe that means better software.
Faster fixes, fewer maintainer headaches, more secure projects, more efficient development.
Mehn... I don't know.
But I do know that something is being lost when the human contribution becomes unnecessary.
The thing I don't want us to automate away
I don't want to romanticize the old days either.
Open source has never been perfect.
Maintainers have always been overwhelmed, contributors always submitted bad PRs, projects have always closed contributions, gatekeeping has always existed.
And not every repository needs to be a learning platform for someone else's career.
Maintainers are allowed to protect their time, but I think we need to be intentional about what we are optimizing for, because the value of open source has never only been the code.
It is the people around the code, and the person making their first PR.
It is the maintainer leaving a comment explaining why something doesn't work, and the contributor discovering an architecture pattern for the first time.
It is the random person from another country who becomes your friend because you both care about the same stupid tiny JavaScript package.
It is the opportunity that comes from being able to point at a public contribution and say:
“I built this.”
That stuff doesn't show up in a GitHub star count, it doesn't show up in download numbers, it doesn't show up in an AI agent's benchmark.
But for a lot of us, that is the actual product of open source.
So, is open source no longer exciting?
Maybe the better question isn't whether open source is no longer exciting.
Maybe it's:
What exactly are we trying to preserve?
Because if we're talking about software production, AI is making open source more exciting than ever.
The ability to build, test, review, debug and ship software is becoming absurdly powerful.
But if we're talking about open source as a community, I'm less certain.
The contributor experience that once made open source feel like a playground for anyone willing to participate is being squeezed from both sides.
On one side, there are contributors using AI to produce contributions faster than maintainers can reasonably review them.
On the other side, there are maintainers using AI to eliminate the need for external contributions in the first place.
And somewhere in the middle is the human contributor.
The person who isn't trying to maximize PR count, the one who isn't trying to automate everything, and the one who just wants to learn, build something useful and become part of something bigger than themselves.
I hope we don't accidentally optimize that person out of open source.
Because if we do, we might end up with the most technologically advanced version of open source we've ever seen...
and the least human one.
And that, to me, would be a pretty boring future.
Top comments (0)