The protocol aims to address an important market inefficiency: the discrepancy between those with insights (alpha signals) and those with the means of executing upon those insights (consumers). Artificial intelligence (AI), in particular machine learning (ML) can help in modeling market data for illiquid assets and can be leveraged to produce innovative financial instruments.
The protocol consists of several core components that facilitate the decentralized generation, evaluation, and delivery of alpha signals:
- Data Provisioning: Data providers can contribute proprietary datasets, feature engineering code and other data artifacts to the network. These aid model creators in training performant predictors. Data providers share revenue with data scientists.
- Alpha Mining: Alpha Miners (or miners; data scientists, feature engineers, ML researchers, etc.) create and connect machine learning models in alpha-specific subnets that minimize a pre-specified loss function. In doing so, they generate alpha.
- Model Validation: After some amount of time passes, a ground truth emerges against which the released alpha of each miner can be scored by validators. Participants stake tokens and conduct peer review of alpha contributions using sophisticated mechanisms that incentivize honest signals of quality.
- Consensus & Incentive Distribution: Consensus forms between validators on what is the highest quality alpha for a given objective.
- Payment & Delivery: Payment from a consumer instigates a process to deliver the results of the protocol to any on-chain destination. Results are delivered using cross-chain gossip protocols.
Updated 22 days ago