Nodes on waitlist
—
Frontier AI is trained in a handful of buildings owned by a handful of companies. Meanwhile the largest pool of idle compute on earth sits under desks and in living rooms. LOOM is a protocol for weaving those threads into one fabric — volunteer and paid contribution, open research priorities, governance by the people who supply the cycles.
You're on the loom. We'll write when the worker client ships.
Your email stays private — only anonymous hardware stats are shared.
Folding@home proved goodwill can mobilize exaflops. LOOM starts there, then layers non-speculative compute credits redeemable for inference on the models you helped train. You weave, you wear.
Training runs are proposed publicly and scheduled by contributor vote weighted by delivered — not pledged — compute. The agenda belongs to the people supplying the cycles.
No trusted center. Redundant computation and proof-of-work-done verify results from anonymous nodes; governance weight decays without contribution, so power can't pool and sit.