Decentralized mesh inference

A model is served by pooling heterogeneous consumer/edge devices into a peer-to-peer swarm, unlike datacenter distributed-inference which splits work across tightly-coupled cluster nodes.

grounded in: Trend notable 'Mesh LLM: Distributed AI comp' and the 'run models on your own mesh and CPUs instead of centralized GPU clouds' theme (updated 2026-07-12).