Core Cycle: From Payment to Rewards
The circulation of $DGAI follows a structured process that links users’ inference demands to node contributions:- Initiation of User Payment When a user submits an inference request via the user client, the process unfolds as follows:
- The request is routed to a DGrid Node, which calculates the task fee using the following formula:
Cost = Compute Unit (CU) × Base Price × (1 + Latency Penalty Coefficient)
- Compute Unit (CU): A standardized metric for measuring inference complexity (correlated with model size, number of input tokens, and execution time).
- Base Price: A fixed price per CU (adjustable via governance mechanisms, with an initial value of 0.05 $DGAI).
- Latency Penalty Coefficient: Ranges from 0 to 0.3 (fine-tuned for latency exceeding the target threshold).
- The user’s $DGAI is escrowed via the Bill Contract—a smart contract deployed on the blockchain—to ensure payment upon task completion.
- Distribution of Node Rewards After inference is completed (verified by Grid Nodes and archived in the AI DA Layer), the escrowed $DGAI is distributed via the Bill Contract.
- DGrid Nodes: Compensation for executing inference tasks (proportional to CU, latency, and uptime).
- Treasury: Used for network maintenance (e.g., AI DA Layer fees, security audits) and developer grants.
- Governance Pool: Incentivizes $DGAI holders to participate in voting (e.g., protocol upgrades, parameter adjustments).
- Staking and Lock-Up Requirements Nodes participating in the network must lock up $DGAI as collateral to ensure accountability.
- Governance and Fee Allocation
- Governance Pool: 5% of task payments (plus 10% of annual treasury funds) are distributed to $DGAI holders who participate in proposal voting (e.g., adjusting base prices, whitelisting new models).
- Network Fees: A 2% processing fee is deducted upfront from all task payments to cover smart contract gas costs and AI DA Layer archiving, ensuring transaction efficiency.
- Penalty Mechanisms and Token Burning Node violations trigger automatic penalties to protect network integrity:
- Confiscated Tokens: 5–20% of a node’s staked $DGAI will be confiscated for malicious behaviors (e.g., submitting false inference results, downtime exceeding 72 hours).
- Burning Mechanism: All confiscated tokens are permanently burned, reducing the total supply of $DGAI over time and enhancing scarcity.
Key Processes: Capital Inflows and Outflows
| Inflows (Sources of $DGAI) | Outflows (Uses of $DGAI) |
|---|---|
| Fees paid by users for inference tasks | Rewards for node operators for inference tasks |
| Weekly rewards for DGrid Nodes | Staking lock-up for node participation |
| Governance rewards for voting participants | Network fees (gas, AI DA Layer archiving) |
| Initial distribution (e.g., public sale, airdrops) | Confiscated tokens (burned) |

