>I used to try a few OSS projects before building anything myself. Recently, with LLM coding assistants/agents, I’m spinning up small, single-purpose tools tailored to my workflow. They’re not “better” than mature OSS (in fact they are often brittle and narrowly scoped) but they solve my exact problem faster than learning how to install, configure and use a new tool.

Ive been coding a great deal of things. I'm approaching things that before LLM coding I wouldnt have gotten into.

>And it got me thinking, if more developers do this what happens to OSS?

In my case I just spent a few weeks building a new OSS project, yet unreleased publicly, to scratch multiple itches. It does things that essentially i have no competitor. My main starting problem was that I wanted to make a change to another open source project but it's in golang and i dont know go.

What happens to OSS when we get essentially unlimited new options of things in every language? The new golden era is among us.

>Do we risk fragmentation into many private one-offs instead of shared libraries?

Do we risk having LLM coders fixing every problem and expanding the shared big projects as well?

>For maintainers: are you seeing fewer users, less feedback, or different kinds of contributions?

Looking at insights on github, my big project, which ive neglected for years has died lol.

> They’re not “better” than mature OSS (in fact they are often brittle and narrowly scoped) but they solve my exact problem faster than learning how to install, configure and use a new tool.

Maybe this is an artifact of just having been around for a long time, but it's extremely rare that I need to use a new tool to solve a new problem. 99.5% of the time, the existing ones are just fine. In the rare cases they aren't, I'll write up a quick script or program to do the job. I don't use genAI for that because I can do it faster and better myself.

In any case, my answer to the main question is no, LLMs have had no impact on my use of OSS.