I.

Now that it’s so easy to just build things, I find myself in a superimposed state. Over the last year, I’ve built a dozen small-to-medium pieces of software that work perfectly for me. They have exactly the feature set I need and no more; I understand how they work, which makes debugging simple; and updating is as simple as pointing my coding harness at the repo. All of these projects are similar to an already existing open-source project, and in the past I might have either contributed to that project or at least forked it to add my feature.

Contributing to open source can be challenging, but maintaining open source is a tough and thankless job. For any given feature, you have to determine whether that addition is general enough and useful enough to incorporate. That means you’re often telling people “no” or asking them to rework it. Additionally, most open source projects have many more features than I will ever use for my own use case - they have to; it’s the balance between being general enough to be useful and scoped enough to avoid bloat.

Great open source projects are usually the vision of an auteur who sees something in the world that should exist. But the very best open source projects achieve something magical: they compound the efforts of many thousands of people on top of that vision into a piece of software that couldn’t have existed before, and one that can change the way people think about software in the first place. Even a smaller project, which compounds only across dozens or hundreds of people’s efforts, can still result in something net new and beyond the vision of the original creator.

II.

Thus my superimposed state: it is both much easier for me, and a better fit piece of software to just build it myself. Generally the cost is a few hours and a few tokens, and the result is exactly what I need. I even open source most of these tools that seem like they could be in any way useful to someone else - I’m giving back to the community right??

On the other hand, if in the future everyone makes the choice to build, then the compounding effect never happens. For every uv or bun that existed previously, there are now 10,000 versions of each, all of which have a small subset of the feature set, but none of which have the aggregate magic that changes the paradigm. Agentic coding does certainly make it easier to contribute directly to open source projects, but right now OSS maintainers are buried under those contributions, and many of them just scratch a personal itch.

Perhaps this is fine - perhaps the perfect, personal tool is better than the compounded, generic open source project. Perhaps paradigm shifts will happen anyway - a thousand thousand leaves fall and eventually one may twist and fall through. But it’s possible we are witnessing a withering of something that was beautiful and will be no more, even if what replaces it is better.