Bittensor is presented as a peer-to-peer market for machine intelligence, where models are ranked and rewarded based on their informational value to other models. The system uses a digital ledger to track these rankings and incentivize participants. A key challenge is preventing collusion, which the paper addresses through a trust-based incentive scheme that rewards well-connected and trusted peers. The network employs techniques like conditional computation and knowledge distillation to optimize bandwidth and enable offline model usage. The goal is to create a decentralized and collaborative environment where diverse intelligence systems can thrive and be monetized. To combat collusion, a loss function is used to force honest nodes to connect. Ultimately, the framework aims to reward the production and dissemination of valuable knowledge within the network.