My Desktop Is My Cloud

April 12, 2026 · 1 min read

I started tinkering with AI locally a few months ago. Running models on a 1080 Ti I bought used. It was slow, janky, and I loved every minute of it. There’s something about pulling off a trick with infrastructure in your own home that no cloud service can replicate.

The more I got into this, the more I started seeing the landscape clearly. The incumbent AI companies are hemorrhaging money while unable to shrink their headcount. The tools are getting better every week and the cost of entry keeps dropping. There are gaps everywhere. The market is wide open for people who can move fast and build practical things. I’m ready to be one of them.

I ended upgrading to an RTX 6000 with 24GB of VRAM, found it on eBay. Now I can run both cards together if I want, but I don’t really do that. It’s simpler and more efficient to just use the RTX 6000. One card, one machine, one server. No fuss.

I set up llama.cpp, llama-swap, Speaches for speech, and OpenCode as a coding agent, all running through an Nginx reverse proxy at home. It’s not a data center. But it’s mine, it works, and it proved to me that I can build on this train as it moves. I’m on it now and I’m not getting off.