How to Responsibly and Effectively Contribute to Open Source Using AI

With the influx of AI tooling, it’s never been easier to contribute to open source communities. These tools are capable of gathering context quickly, “understanding” repositories faster than ever before. They provide instant summaries about repositories that, previously, would have meant reading lines and lines of code. They can fix bugs in programming languages you don’t know, and ultimately allow more contributors to get involved, which (almost) every open source project wants.

Experienced contributors benefit as well. Being able to give AI tooling the context an experienced contributor already knows makes them even more effective. They can be given more tedious tasks that aren’t yet (or can’t be) automated—for example, I’ve tasked AI to follow the OpenTelemetry operator’s release steps to prepare a release PR).

But just because AI tooling allows these opportunities does not mean that it rises to the occasion every time. It is incredibly easy to create “slop” contributions that just end up slowing down an open source community. An influx of low-effort AI contributions steals time away from maintainers and ultimately leads nowhere.So the real question is: how can we use AI to effectively contribute to open source communities? As a maintainer of some OpenTelemetry and Honeycomb repositories, here are my thoughts.

The reality of contributing to open source

Whether you’re using AI tooling to contribute or not, every contributor to an open source community is vying for maintainer attention. These are the folks who ultimately are responsible for the thing you’re changing, and without them, your contribution is not going to get accepted. But they are humans with lots of different responsibilities, hobbies, activities, and interests and they are likely not committing 100% of their time to the communities you’re contributing to. Time is precious, and the faster a maintainer can engage with your contribution, the more likely it is to be accepted.

The amount of time a maintainer spends on your contribution is based on complexity and priority, but also on trust. The more trusted you are as a contributor, the safer a maintainer will feel accepting your contribution. Trust is gained in a lot of ways, but being a consistent contributor who responds to feedback is the quickest.

How is AI hindering you?

If successfully contributing to an open source community is all about making the most of the attention a maintainer can give you (or needs to give you based on how much they trust you), how is AI tooling working against you?

Large pull requests

First, they can hinder you by creating pull requests (PRs) that are too large. With or without AI tooling, the larger the PR, the harder it is for the maintainers to review.. But with AI tooling, it has become easier to make large sweeping changes in one fell swoop. Large PRs require a lot of maintainer attention, a resource that we’ve already established is scarce. By submitting large PRs, you actively reduce the chances of your change being merged in a timely manner.

Instead, when submitting changes using AI, make your PRs small. Even if the AI can make all the changes in one massive PR, have it break things up. This is good advice when not using AI tooling as well, but it’s an extremely important instruction to include to your prompts when setting AI loose on an open source repository.

Large comments

Second, when you use AI to respond to feedback, they can hinder you with giant comments. AI likes to be verbose by default. When a maintainer asks a question, if you task AI to respond to it and it responds with five paragraphs, that’s a lot of information for the maintainer to take in.

Each comment in a PR or issue is debt—context that a future reader needs in order to understand the active discussion. If your AI is joining in on the conversation with verbose comments, that adds a lot of debt. This debt takes more precious maintainer attention, making it less and less likely that your change gets merged in a timely manner.

Instead, prompt your AI to be succinct. Task it to use a specific persona and/or tone. Your response should gain you trust by being direct, on topic, and, as much as possible, human. It may change some day, but for now, maintainers are still looking to engage with other humans: we want to hear what you think. Maintainers still trust humans more than machines, and you’re more likely to gain trust if your responses aren’t generic AI responses. That doesn’t mean you can’t use AI to respond, only that when you do, use it intentionally. Do use it to help articulate your thoughts. Don’t use it to think for you.

Prevents the maintainers from seeing you

Using AI tooling to make large PRs or comments has the additional side effect of eroding trust with the maintainers. If maintainers suspect your PR or comments are submitted by AI without your own verification, they’ll likely decide the contribution needs additional review, thus lengthening the time to merge.

Blatant AI contributions may also result in maintainers being wary of the engagement they’ll receive from the submitter. All open source communities suffer when a person submits a PR and then abandons it. With AI tooling, it’s a lot easier for someone to quickly submit a change and then never look at it again, so you have to be careful of the bias a maintainer may have when you submit AI contributions. If they see the AI contribution and have the bias that you won’t reengage on this change, they may ignore you.

Vibe coding is fun—reviewing vibe-coded code is not. It gets more fun when the maintainer gets to engage with the submitter. By staying engaged with your contribution, you build trust and increase the changes of your change being accepted.

How can AI help you?

What are the best ways to use AI to contribute to an open source community? Probably one of the best ways to use AI is to help overcome a language barrier. If you don’t know the language of the maintainers, AI can help. But, as we cautioned before, you have to use it in the right way. Make sure to instruct your AI to be succinct and to stay on topic. Have the AI explain that you’re using this tool to help with a language barrier. You want the maintainers to see you, the human submitter, behind your AI usage.

In addition to language barriers, AI can help with how you write. If you’re not confident in your ability to express yourself clearly though text, AI can help take your stream of consciousness and output something articulate. Again, review the response to make sure it represents what you were struggling to put into words. Try to make it not sound like AI.

AI is also extremely helpful for understanding a repository. Before AI tooling, contributing to a new repository could mean a lot of reading. That could mean a lot of time commitment or, if you’re not a programmer, it could be an impossible task. AI is incredibly good at summarizing. Try tasking it to explain the different folders in a repo or what a complex function does and it will probably get it right. Use this feature to increase your understanding of the community to which you are contributing. The more you understand, the better contributions you’ll be able to submit with or without AI tooling.

AI is a tool, not a contributor

Ultimately, AI is another tool at your disposal. Like most tools, it can help you or hinder you depending on how you use it. Open source communities are just that: communities. They flourish on meaningful, repeat engagement. Make sure when contributing with AI tooling that you respect that these are communities of people with limited attention. Use AI tooling to help you become a part of the community, not burden it.