Glossary
Bittensor
A protocol that incentivizes ML model providers to compete on specialized "subnets" — models score each other, and the highest-quality miners earn TAO rewards.
Bittensor is a Layer-1 designed around an incentivized network of ML model providers. Each subnet specializes in a task (text generation, embeddings, image generation, prediction markets, etc.). Miners on a subnet run models and score each other's outputs; validators aggregate the scores; TAO token emissions flow to the highest-ranked miners.
The model attempts to coordinate decentralized ML training and inference via market mechanics. Critics argue that the scoring quality of subnets is hard to verify and that token incentives may not converge on genuinely useful models — these are open questions the protocol's roadmap is actively addressing.