Abstract
This PIP defines AI mining integration for Pars Network, enabling GPU nodes to mine PARS tokens via the decentralized Proof of AI protocol. Nodes with GPUs can contribute AI compute to the network and earn rewards, creating a sustainable infrastructure for AI-powered civic services.
Motivation
AI for Civilization
Pars Network provides the infrastructure for Persian diaspora civilization. AI capabilities are essential for:
- Translation Services: Real-time Persian/Farsi translation
- Document Processing: Digitizing historical archives
- Privacy-Preserving Analysis: Secure data processing for sensitive civic needs
- Education: AI-powered learning tools
- Healthcare: Diagnostic assistance in underserved areas
Decentralized AI Infrastructure
Rather than depending on centralized AI providers (who can be pressured or blocked), Pars Network enables:
- Community-Owned Compute: GPU owners contribute to shared infrastructure
- Censorship-Resistant AI: No single provider can block access
- Economic Alignment: Miners earn PARS for providing AI services
- Privacy by Design: AI inference runs on trusted community nodes
Specification
Architecture Overview
┌─────────────────────────────────────────────────────────────────────────────────────┐
│ AI MINING ON PARS NETWORK │
├─────────────────────────────────────────────────────────────────────────────────────┤
│ │
│ GPU MINERS PARS NETWORK │
│ ┌─────────────────────┐ ┌─────────────────────────────────────────┐ │
│ │ H100 / RTX 4090 │ │ Pars EVM (Chain ID: 6133) │ │
│ │ ┌───────────────┐ │ │ │ │
│ │ │ AI Workload │ │ Proof │ ┌──────────────┐ ┌──────────────┐ │ │
│ │ │ (Inference) │──┼───────────► │ │ AI Mining │ │ Reward │ │ │
│ │ └───────────────┘ │ of AI │ │ Precompile │──│ Ledger │ │ │
│ │ ┌───────────────┐ │ │ │ (0x0300) │ │ │ │ │
│ │ │ NVTrust TEE │ │ │ └──────────────┘ └──────────────┘ │ │
│ │ │ Attestation │ │ │ │ │
│ │ └───────────────┘ │ └─────────────────────────────────────────┘ │
│ └─────────────────────┘ │ │
│ │ Teleport │
│ ▼ │
│ ┌─────────────────────────────────────────────────────────────────────────────┐ │
│ │ CROSS-CHAIN AI TOKEN ECOSYSTEM │ │
│ │ │ │
│ │ ┌─────────────┐ ┌─────────────┐ ┌─────────────┐ ┌─────────────┐ │ │
│ │ │ Pars EVM │ │ Lux │ │ Pars Network │ │ Zoo │ │ │
│ │ │ 6133 │ │ 96369 │ │ 36963 │ │ 200200 │ │ │
│ │ │ PARS, AI │ │ LUX, AI │ │ AI, HANZO │ │ ZOO, AI │ │ │
│ │ └─────────────┘ └─────────────┘ └─────────────┘ └─────────────┘ │ │
│ │ │ │
│ │ Same AI work can ONLY be minted on ONE chain (double-spend prevention) │ │
│ └─────────────────────────────────────────────────────────────────────────────┘ │
│ │
└─────────────────────────────────────────────────────────────────────────────────────┘Proof of AI Protocol
Miners submit cryptographically attested AI work proofs:
rust
/// AI work proof submitted to Pars Network
pub struct AIWorkProof {
/// Miner's ML-DSA public key
pub miner: Vec<u8>,
/// BLAKE3 hash of model used
pub model_hash: [u8; 32],
/// BLAKE3 hash of input data
pub input_hash: [u8; 32],
/// BLAKE3 hash of output data
pub output_hash: [u8; 32],
/// Standardized compute metric
pub compute_units: u64,
/// Work completion timestamp
pub timestamp: u64,
/// ML-DSA signature over proof
pub signature: Vec<u8>,
/// NVTrust attestation (if available)
pub tee_attestation: Option<TEEAttestation>,
}NVTrust Chain-Binding (Double-Spend Prevention)
AI work is bound to a specific chain BEFORE compute runs:
rust
/// Pre-commit context binds work to Pars Network
pub struct WorkContext {
/// Target chain: Pars = 6133
pub chain_id: u32,
/// Job identifier
pub job_id: [u8; 32],
/// Model being used
pub model_hash: [u8; 32],
/// Input data hash
pub input_hash: [u8; 32],
/// GPU hardware ID
pub device_id: [u8; 32],
/// Unique nonce (replay protection)
pub nonce: [u8; 32],
/// Timestamp
pub timestamp: u64,
}Key Invariant: The same AI work cannot be claimed on Pars, Lux, AND Pars Network - only on the chain specified in WorkContext.chain_id.
Supported GPUs
| GPU Model | Trust Level | AI Mining Support |
|---|---|---|
| H100 | Full NVTrust | Full attestation |
| H200 | Full NVTrust | Full attestation |
| B100 | Full NVTrust + TEE-I/O | Full attestation |
| B200 | Full NVTrust + TEE-I/O | Full attestation |
| RTX PRO 6000 | NVTrust | Full attestation |
| RTX 5090 | Software only | Limited (trust score 60) |
| RTX 4090 | Software only | Limited (trust score 60) |
ZAP Protocol Integration
Nodes communicate via ZAP (Zero-copy Agent Protocol) for high-performance AI workload coordination:
┌─────────────────────────────────────────────────────────────────────────────────────┐
│ ZAP FOR AI MINING │
├─────────────────────────────────────────────────────────────────────────────────────┤
│ │
│ AI CLIENT ZAP GATEWAY GPU MINERS │
│ ┌─────────────┐ ┌─────────────┐ ┌─────────────┐ │
│ │ Request │ ZAP (Cap'n │ Route to │ ZAP (Cap'n │ Execute │ │
│ │ AI Task │────Proto)──────► │ Available │────Proto)──────► │ Inference │ │
│ │ │ │ Miner │ │ │ │
│ └─────────────┘ └─────────────┘ └─────────────┘ │
│ │ │ │ │
│ │ Zero-copy binary │ 10-100x faster │ │
│ │ No JSON parsing │ than JSON-RPC │ │
│ │ Native tensor support │ │ │
│ │ │ │ │
│ ┌─────────────────────────────────────────────────────────────────────────────────┐ │
│ │ ZAP BENEFITS FOR AI WORKLOADS │ │
│ │ • Parse latency: 0.2 us vs 45 us (JSON) - 225x faster │ │
│ │ • Binary tensors: 0% overhead vs 33% Base64 │ │
│ │ • Memory: 0 allocations vs 47 per message │ │
│ │ • Throughput: 3.2 GB/s vs 180 MB/s │ │
│ └─────────────────────────────────────────────────────────────────────────────────┘ │
│ │
└─────────────────────────────────────────────────────────────────────────────────────┘AI Mining Precompile
The AI Mining precompile at 0x0300 enables smart contract integration:
solidity
interface IAIMining {
/// @notice Get mining balance for an address
function miningBalance(address miner) external view returns (uint256);
/// @notice Verify ML-DSA signature
function verifyMLDSA(
bytes calldata pk,
bytes calldata msg,
bytes calldata sig
) external view returns (bool);
/// @notice Claim teleported mining rewards
function claimTeleport(bytes32 teleportId) external returns (uint256);
/// @notice Get pending teleport transfers
function pendingTeleports(address recipient) external view returns (bytes32[] memory);
/// @notice Submit AI work proof
function submitWorkProof(
bytes calldata proof,
bytes calldata attestation
) external returns (uint256 reward);
}Reward Distribution
Mining rewards calculated as:
reward = base_reward * compute_units * difficulty_adjustment * trust_multiplierWhere:
base_reward: Network-configured base rate (governance controlled)compute_units: Verified AI compute performeddifficulty_adjustment: Dynamic based on network compute capacitytrust_multiplier: Based on GPU attestation level (0.6 - 1.0)
Mobile Mining (Light)
Mobile nodes can participate in AI mining through federated inference:
┌─────────────────────────────────────────────────────────────────────────────────────┐
│ MOBILE AI MINING │
├─────────────────────────────────────────────────────────────────────────────────────┤
│ │
│ MOBILE DEVICES │
│ ┌─────────────────────┐ │
│ │ iPhone / Android │ │
│ │ │ │
│ │ ┌───────────────┐ │ Federated │
│ │ │ Neural Engine│ │ Learning ┌─────────────────────────────────────┐ │
│ │ │ (NPU) │──┼────────────────► │ AGGREGATE NODE │ │
│ │ └───────────────┘ │ │ • Collects model updates │ │
│ │ │ │ • Verifies contributions │ │
│ │ ┌───────────────┐ │ │ • Distributes rewards │ │
│ │ │ Privacy │ │ └─────────────────────────────────────┘ │
│ │ │ Preserved │ │ │
│ │ │ (on-device) │ │ │
│ │ └───────────────┘ │ │
│ └─────────────────────┘ │
│ │
│ Mobile mining rewards are lower but contribute to network without GPU │
│ │
└─────────────────────────────────────────────────────────────────────────────────────┘Cross-Chain AI Token Ecosystem
PARS, ZOO, Pars Network AI, and Lux can all participate in AI mining:
| Chain | Chain ID | Native Token | AI Token | Mining Support |
|---|---|---|---|---|
| Pars Network | 6133 | PARS | AI | Full (via precompile) |
| Pars Network | 96369 | LUX | AI | Full (via precompile) |
| Pars Network | 36963 | HANZO | AI | Full (native L1) |
| Zoo Network | 200200 | ZOO | AI | Full (via precompile) |
Teleport enables seamless movement of AI tokens between chains.
Security Considerations
TEE Attestation
- NVTrust provides hardware-backed attestation
- Compromised GPUs produce invalid attestations
- Trust scores reflect verification confidence
Double-Spend Prevention
- WorkContext binds work to chain BEFORE execution
- Spent set tracked per chain
- Cross-chain double-spend impossible with chain_id commitment
Replay Protection
- Unique nonce per job
- Timestamp validation
- Spent key tracking
Implementation
Mining Node Setup
bash
# Install Pars mining node
git clone https://github.com/pars-network/pars-miner
cd pars-miner && cargo build --release
# Configure for Pars Network
cat > config.toml << EOF
[network]
chain_id = 6133
rpc_endpoint = "https://rpc.pars.network"
[mining]
wallet_path = "~/.pars/mining-wallet.json"
models = ["llama3", "mistral", "qwen3"]
[gpu]
enable_nvtrust = true
device_ids = [0, 1]
[zap]
gateway = "zap://mine.pars.network:9999"
EOF
# Start mining
./target/release/pars-miner --config config.tomlZAP Miner Registration
rust
// Register as AI miner via ZAP
let gateway = Gateway::connect("zap://mine.pars.network:9999").await?;
gateway.register_miner(MinerInfo {
id: miner_id,
capabilities: vec!["llm-inference", "embedding", "vision"],
gpu_info: get_gpu_info(),
trust_attestation: get_nvtrust_attestation(),
}).await?;
// Receive work assignments
while let Some(job) = gateway.next_job().await? {
let result = execute_ai_job(&job).await?;
gateway.submit_result(result).await?;
}References
- [HIP-006: Pars Network AI Mining Protocol](https://github.com/Pars Networkai/hips/blob/main/HIP-006-ai-mining-protocol.md)
- [HIP-007: ZAP Protocol](https://github.com/Pars Networkai/hips/blob/main/HIP-007-zap.md)
- PIP-0000: Network Architecture
- PIP-0002: Post-Quantum Encryption
Copyright
Copyright and related rights waived via CC0.