UR

The subnet follows a distributed architecture where validators continuously create protein folding challenges and distribute them to miners via the Global Job Pool (GJP). Miners receive job specifications (PDB IDs, system configurations) and perform molecular dynamics simulations using OpenMM, uploading trajectory results to S3. Validators periodically query all miners for job status, retrieve and evaluate results against energy baselines, run credibility validation (state-checkpoint checks), compute rewards using a centralized reward registry, and update miner scores via EMA.

An organic API allows external users to submit custom folding jobs. Background tasks handle job creation, job updates, reward distribution, synchronization with the Bittensor metagraph, and health monitoring.

Miners work on protein folding challenges placed in the Global Job Pool (GJP) and upload results to S3 for validation. Each miner uses a unique random seed to ensure diverse solution exploration. If miners submit identical results, their rewards are zeroed to enforce innovation.

Validators evaluate solutions by computing energy metrics and running evaluation pipelines that include credibility checks (state-checkpoint verification) and reward calculation using a registry-based system. Scores are updated using exponential moving average (EMA) after jobs complete.