Please Don't Start An Open Source Project, Join One

The effect of AI on PRs has gotten a lot of press. Slop PRs, untested PRs, just the profusion of PRs. Slop bugs, slop security reports, all sorts of slop problems. Definitely a problem.

But I would also propose that something else happening in the open source space that is hurting the open source community: Everyone thinks they can start their own project now. The stream of web frameworks, proxy servers, and all of the other code projects that we frankly had too many of even before AI has accelerated even more. Everyone seems to think they can run their own project now.

Everyone is, in some sense, even correct about that. They can now do all the technical work of running their own separate project.

It was already difficult to get anyone else to use your project in the past, but now there’s even more projects to choose from and the first couple of brave users who may have tried your project are instead trying to run their own. The feeder system for open source is getting flooded out.

Since as I said, everyone is basically correct that they can now run a full project when they couldn’t before, the fix will need to be cultural. So, I exhort you, if you are thinking of starting a project in a language and problem space that is already overrun with alternatives…

Please think harder about joining an existing project.

I wrote a few months ago that in an AI era, the real value of a codebase is in how much contact with the real world it has had. Two projects that have one user, or in many cases, clearly zero users, is vastly less useful than one project that has had even two users. That second use, that second check against reality, the second cross-check that, yeah, the system actually works for something other than the exact way the original author was using it, that second user is the most important user a project will ever have. With a second user, the project instantly stands out above the pack of projects with 0-or-1 users. Not enough on its own to succeed, but perhaps enough that success becomes possible, rather than a complete pipe dream.

AI hasn’t created this problem, but it made it much worse.

If you have a clear thesis about why your project is new and unique, solves a structural problem in some other project that can only be solved with a new project, or can just in general articulate why your goals can’t be achieved by joining with someone else, then by all means, start something up. There are many projects that can clear this bar. There are many, many more that can’t.

Which sounds better, to be “the most important user the project has ever had” or to rule over the wasteland of your zero-user project?

To those thinking of starting a project, I ask you to consider harder looking for one to join.

To those not thinking of starting a project but who might speak to people who will, I ask you to consider whether this is a good cultural change that should be advocated for and to start recommending to people to join existing projects rather than starting their own.