We're building a small AI lab out of consumer hardware sitting on our desks — no data center, no cloud bill, just a couple of mini PCs learning to think locally. This site is the public record of that build: what we install, what breaks, what we learn, and what we'd tell you if you wanted to build one too.
The stack so far leans on Ollama for running models locally, Open WebUI as the chat front end, Claude Code as our engineering assistant for building and documenting the lab itself, and Tailscale to mesh our machines together over a private network. Everything here — hardware notes, model comparisons, mistakes, and lessons — is written up as we go and shared openly.
Already on the shared Tailscale tailnet or the home LAN? Open the chat app directly:
Not connected to the tailnet? This link won't load — see Field Guide Part 2 for how mobile/remote access is set up.
New to local AI? Start with the basic stack in plain language.
Read →How the lab's nodes fit together over Tailscale.
Read →Dated entries from the build — milestones, screenshots, quotes.
Read →A visual, diagram-heavy guide to how local AI actually works.
Read →Technical notes, gotchas, and observations from setup.
Read →What's actually running the lab, spec by spec.
Read →Which local LLMs we've tried, and what they're good for.
Read →Ollama, Open WebUI, Claude Code, and Tailscale — the tools we run.
Read →You don't need to be an expert. Come learn with us.
Read →