Modular Compute Protocol

MCP powers decentralized AI training at scale:

  • Task Assignment: Micro-tasks are dynamically matched to VisionNodes

  • Reputation System: Contributor scores based on task success, uptime, latency

  • Task Verification: Quorum-based or zkML-based proof of result integrity. If quorum fails, zk proof is required for validation.

  • Resource Optimization: Scheduler avoids node overload and ensures fairness

  • Slashing Logic: Faulty or malicious contributors are penalized via token slashing and reputation decay

PreviousVisionXNextValue Accrual

Last updated