Run a Node

Run a Node — Become a Distributed AI Compute Contributor

Operating a Zus Node transforms your device into a verifiable participant within a decentralized federated AI infrastructure. Whether deployed through a browser extension, Telegram-integrated client, or VPS environment, each node contributes compute, validation, and network stability to the protocol.

Nodes are not passive endpoints — they are active agents in the distributed training lifecycle:

• Execute local model training cycles

• Generate gradient updates

• Encrypt and submit updates to the global aggregation layer

• Participate in peer validation and network health verification

• Maintain uptime commitments under protocol-defined thresholds

All node activities are cryptographically verifiable and anchored to the Solana network for transparency, performance, and auditability.

Reward Mechanism Design (On-Chain Verified)

The incentive layer is engineered to align compute contribution, model quality, and network reliability with tokenized rewards.

Mechanism	Technical Description	Reward Source
Proof-of-Uptime (PoU)	Verifiable node availability tracked via on-chain heartbeat submissions and network monitoring oracles.	Base points emitted per validated online minute.
Proof-of-Training (PoT)	Gradient updates evaluated by the global federated aggregator. Quality scoring based on convergence contribution and variance metrics.	Performance-weighted bonus allocation.
Proof-of-Connectivity (PoC)	Peer-to-peer handshake validation and latency consistency checks across the node mesh.	Stability multiplier applied to base rewards.

Reward Computation Logic

Each node receives a composite reward score calculated through:

Total Score = (PoU Base × Tier Multiplier) 
            + (PoT Quality Score × Training Weight) 
            + (PoC Stability Bonus)

All reward calculations are:

• Anchored via smart contract execution

• Transparent and auditable on-chain

• Non-custodial and automatically distributed

• Resistant to manipulation through cryptographic verification

Node Tier Structure

Node classification determines reward multipliers and workload allocation priority.

🧩 Browser Node

• Lightweight execution via Chrome Extension

• Optimized for retail contributors

• Performs micro-training cycles

• Lower compute threshold, broader decentralization coverage

💻 VPS Node

• Dedicated server environment

• Higher training batch allocation

• Improved gradient throughput

• Enhanced PoT earning potential

⚙️ Partner Node

• Institutional-grade deployment

• Enterprise compute participation

• Priority model aggregation routing

• Maximum multiplier tier

Security, Slashing & Integrity Enforcement

To preserve protocol integrity:

• Idle nodes below minimum uptime thresholds are automatically downgraded.

• Malicious submissions (invalid gradients, spoofed activity) trigger automated slashing mechanisms.

• Repeated violations result in reputation score decay and potential suspension.

Smart contract–based audits ensure:

• Deterministic reward issuance

• Zero manual intervention

• Transparent enforcement

Economic Alignment

The Node & Reward architecture is designed to:

• Incentivize real compute over speculative activity

• Prevent Sybil attacks via performance-weighted scoring

• Encourage long-term uptime commitments

• Align token emissions with measurable AI training contribution

This framework ensures that contributors are rewarded not for merely being online — but for meaningful, verifiable AI computation and network reliability.