Bittensor coordinates specialized AI competitions on-chain while the heavy model work runs off-chain, with rewards routed through TAO and subnet alpha tokens.
Subnets run off-chain tasks and scoring, while on-chain Yuma Consensus and root or dTAO signals decide who gets paid in TAO.