Dollama

A self-hosted AI cluster protocol that turns the computers I already own into a personal compute network -- a lightweight laptop can call on a desktop GPU, tasks run in parallel across machines, and I can draw on free network capacity when I need more. It's a chance to put idle compute to use, and to gain some independence, resilience, and sovereignty over my own AI access instead of renting all of it back from someone else's cloud.

What it proves: cost/accuracy/latency tradeoffs on real hardware — distributed inference I designed, benchmarked, and operate myself.

Visit dollama.net →