Writing

Share the Thing

About two weeks ago I open-sourced an ML utility I made to solve a problem I had and I figured it might help a few other people. It's pretty niche.

In that time, more than 3,000 people have run it. There have been 769 unique GitHub cloners and 2,800 Docker Hub pulls. I just did one LinkedIn post and one Show HN post, and the HN post got zero upvotes.

Not only are people using it, but it's being deployed on serious hardware: there are ~100 pulls of my intermediate Docker images, and the only reason you'd do that is to add support for NVIDIA Blackwell. That's big iron.

Docker Hub statistics for michaelrothrock/mixlab, mixlab-cuda, and mixlab-cuda-base showing 2.8K, 68, and 36 pulls respectively

I'm sharing this not to brag (though yes it does feel good to see usage) but to show how far a little utility can go. I just built this for myself, to solve my problem, and shared it.

The big enabler here was agentic coding. I could get the LLM to build something relatively sophisticated, quickly, to solve my problem. And I'm not the only one with this problem — apparently more than 3,000 of you have it as well.

And that's what I hope you take away from this post: if you're sitting on something useful, share it. It doesn't have to be big. mixlab is small, MIT licensed, and now apparently some stranger is adapting part of it for Blackwell hardware. I still can't quite get over that.

Maybe your tool will solve someone else's problem, too.

mixlab on GitHub →