The crowd
is the GPU.

BRAIN AUTO estimates every execution target, picks one for your mode and privacy setting, runs it, and attaches a receipt. Today chat runs on a configured model provider; the browser network runs verified parallel compute. Each answer says exactly which.