
We build local-first AI agents, the VirusGPT workstation, autonomous production pipelines, and living game worlds — running entirely on hardware we own.
Software should answer to the person who runs it. We build local-first: inference on your GPU, storage on your disk, orchestration on your LAN. The cloud is an option we configure — never a dependency we accept. Every system we ship can be pointed at, inspected, and owned.
Your studio runs on hardware you can touch. No subscription gate on your own creativity, no model that goes quiet because a server farm changed its mind. When the internet is down, the work still ships — because it never left the room.
We build in the open: sources, pipelines and failures documented as we go. Repos you can read, builds you can run, logs you can audit. Trust is something you verify, not something you are marketed.
The network is the workbench. Agents, GPUs and game engines moving as one crew across the LAN — a small studio that runs like a lab, and a lab that answers to exactly one person.