A few months ago, I wrote about how you can still just open source things. I’ve since been thinking of what happens next for open source, as AI models continue to get even better at coding.
A fair concern is AI will make open source irrelevant-ish. Why share code when anyone can generate it? Why contribute when a model can write it faster?
Right now (and this may very well change!) I think we’re going to see more open source, not less. Not necessarily better open source yet in the short-run (although some will be better)! But certainly more.
The other day I wanted to manage PostHog without using its UI to add/remove feature flags, set cohorts, that kind of thing.
So I put together piglet, which is a CLI for some of PostHog’s management API. It took ten minutes? Less? Of course you can build your own version too, just as quickly. I added a skill plugin: “create a flag called voice-ai with 50% rollout.” That took two minutes with prior art.
The process to make piglet was straight-forward. Prompt/plan like this:
- here are the API docs for posthog
- use click (the open source Python CLI library)
- give me json, table, and raw output modes
- follow prior art from the cased-kit CLI
- write tests
- do real-world testing using the API key in my env
Half a dozen small nudges in the coding session, and a couple of style preferences, and that was it.
The code is fine. It’s trivial, of course (it’s an API wrapper), but still good. So that’s ten minutes for something I’d actually use, release, and put on GitHub with a README + installation instructions. It would have taken me a few hours before; and honestly I probably wouldn’t have even bothered. Seems like a good thing?
We’ve all heard the cost of writing software is collapsing. For certain kinds of software (like an API wrapper), that’s obviously true. So is it worth open sourcing something like this? Sure! Because why not?
The friction that used to stop people from open sourcing (polish, documentation, just running the damn push command) is lessened when the investment is trivial. AI is good at helping with that stuff, too.
And here’s the really good thing: open sourcing forces the right kind of investment. Out of a feeling of public open source responsibility, I’ll actually maintain it, and that is good for everyone, myself included.
We keep hearing about abundant software. More apps, of course, but also small, specific, sharp tools. Things that used to live as hacky one-off scripts, or never get written at all.
The commercial open source playbook (open core, developer relations, enterprise features), that’s still around. But alongside it, we’re going to see a lot more people just… make things and put them out there. You can still just open source things, and a lot more people will. Some will get usage, most won’t. And that’s fine.
The bigger question now is how to judge what’s out there. But that’s a question for another day and another post. Hint: I think it’s about that same promise (and necessary burden) of maintenance.
