Chapter 3 Decentralized Mining Network and Proof of ZK Work
PreviousChapter 2 Introduction to the Zypher’s Safe AI NetworkNextChapter 4 Token Utility & Allocation Design
Last updated
Was this helpful?
Last updated
Was this helpful?
For large-scale AI-on-chain applications, an efficient and scalable decentralized ZK proof generation network is indispensable:
Centralized Solutions’ Bottleneck: If you rely on a centralized cloud service to generate the massive quantity of proofs, you may face expensive fees and limited computing resources; plus, if that central node fails, is attacked, or gets censored, the entire system halts.
Decentralized Value: By distributing ZK proof tasks among mining nodes across the globe, the network can utilize idle or cheaper computing power, reinforced by cryptography and on-chain incentives to guarantee accuracy. This lowers costs significantly and, during usage spikes, the network can auto-scale to handle demand.
Upgrading the Trust Model: In a decentralized architecture, users no longer rely solely on one centralized provider. Instead, security is collectively maintained by distributed nodes worldwide, greatly increasing system resilience and anti-censorship.
PoZK is a consensus mechanism that integrates PoS (Proof of Stake) and PoW (Proof of Work) ideas, specifically for measuring ZK proof generation workload and distributing rewards fairly between miners and users. Combined with the Cobb-Douglas incentive model, PoZK accounts for a miner’s computing power, token staking, and user participation metrics (e.g., task frequency, staked tokens), providing a balanced approach to efficiency and fairness. By employing zero-knowledge proof technology, PoZK transforms AI Agent execution into verifiable computational tasks, assigning a uniform “Work” metric so that mining nodes in a decentralized network can compete more efficiently and fairly. Thus, users can not only get verifiable AI inference results but also rely on the PoZK consensus framework to fortify network security through robust incentives.
Standardizing ZK Workload: The computational cost of generating ZK proofs is quantified into “Work,” making it comparable across tasks.
Decentralized Incentives: Network nodes (miners) are rewarded for undertaking and completing these ZK computations, thus safeguarding the trustworthiness of AI Agents.
3.2.1 Miner’s ZK Proof Workload Calculation
Core formula:
Circuit Size (gates): The number of logic gates in a ZK circuit, directly reflecting the task’s complexity. Larger gate counts mean more time and resources required.
α (Alpha): A weighting factor for different ZK solutions (such as Plonk, Groth16, Risc0, etc.), balancing differences in circuit structures and computing demands (e.g.,Plonk：α=1，Groth16：α=0.2，Risc0 zkVM：α=4).
The reward allocation formulas presented in this section are derived from the Cobb–Douglas production function, a foundational model in economics that represents the relationship between multiple input factors (such as capital and labor) and total output. Originally formulated to describe industrial productivity, the Cobb–Douglas function has since been widely adopted across various domains for its ability to model diminishing returns and weighted contributions of multiple inputs.
Zypher Network adapts and extends this model to fit the context of decentralized zk-proof mining and usage, introducing a multi-factor reward allocation mechanism that balances computational effort (Work), application-level contributions, and staking activities on both application and individual levels. This adaptation ensures fair, incentive-aligned distribution of network rewards to both miners and users, based on measurable participation and resource commitments.
In simplified notation:
This structure ensures:
Work-based fairness: Miners are rewarded proportionally to their ZK contribution.
App-level equity: Applications with higher computational complexity or staking depth receive a greater share of the reward pool.
Individual incentive alignment: Users and miners are incentivized to stake and participate actively.
Parameter Definitions
Variable
Description
n
Total reward pool allocated for the current epoch
work
Total ZK proof computation performed by all miners during the epoch
usage
Total task volume generated by all users during the epoch
app_cs
Circuit size (complexity) of the specific application
total_cs
Total circuit size across all applications in the epoch
app_staking
Total tokens staked across all users within the application
total_app_staking
Total tokens staked across all applications in the network
self_work
ZK proof work performed by the individual miner
app_work
Aggregate proof work contributed by all miners in the application
self_num
Number of ZK tasks submitted by an individual user
app_num
Total number of user-submitted tasks for the application
self_staking
Tokens staked by the individual miner or user
total_staking
Total tokens staked across the entire network
a, b, c, d
Adjustable weighting parameters set by governance (DAO) to tune reward distribution logic
3.3.3 Incentive Distribution & Reward Claiming
Rewards are calculated and distributed over a fixed interval (called an Epoch, typically every 24 hours). The general steps include:
Workload & Contribution Statistics: Record how many ZK proofs miners have generated during the Epoch, and how many tasks or staked tokens users have contributed.
Reward Pool Calculation: After referencing the set formulas and weights, figure out how many tokens go to miners and to users at the end of the Epoch.
Automatic Rewards Distribution: Once the Epoch ends, the system automatically settles and transfers rewards to miner and user accounts.
Feedback Loop & Growth: This model encourages miners to continue providing computing resources and users to generate more tasks for ZK proofs, fostering a virtuous cycle of security and network engagement.
3.3.4 Sustained Development of the Incentive Ecosystem
Miner Incentives: Miners can stake more tokens to qualify for more ZK tasks, thereby earning higher mining returns. They can also invest in hardware upgrades to complete larger or more complex circuits first.
User Incentives: Users can also stake more tokens, gaining higher service quality or additional task feedback rewards, thus guiding them to actively use AI Agents and continue circulating tokens within the ecosystem.
A decentralized ZK proof network operates on a “Task Posting–Miner Bidding–Verification & Incentive” cycle. Concretely:
Posting ZK Tasks
When an AI Agent user undertakes an operation requiring ZK proofs (e.g., a financial transaction, in-game battle), they submit the tasks along with circuit data, reward pool details, etc., to an on-chain contract.
The tasks could include circuit instructions, data to be proven, or encrypted materials.
Miner Bidding
Mining nodes in the PoZK network constantly monitor for available tasks. Based on their hardware specs, staked tokens, network latency, etc., they decide whether to “take the job.”
Once a miner confirms acceptance, it starts generating a ZK proof. If a miner times out or fails, other miners can step in, forming a competitive mechanism.
Submission & Verification
The miner completes the proof within the allotted timeframe and submits the proof file on-chain for contract validation.
If the proof is deemed valid, the miner is rewarded with tokens proportional to their “Work” and staking parameters.
Users Also Receive Incentives
In some strategies, if users (the task initiators) hold and stake tokens, they may receive a portion of mining rewards or a fee rebate, thus incentivizing them to continue using AI Agents.
This is commonly implemented by including a user dimension (“Play” or usage) in the Cobb-Douglas function.
The Decentralized ZK Task Marketplace not only lets miners leverage their unique computing advantages but also encourages greater user participation, driving overall network prosperity and diversity.
Advantages of PoZK:
Transparent Computation & Verification: All nodes can see circuit gate counts and the α weighting, comparing differences in workload and greatly minimizing collusion and hidden operations. With “universal verification,” any anomalies or arbitration requests can be quickly scrutinized.
Security with Zero-Knowledge Proof: Unlike simple hashing or stake-based checks, ZK proofs hinge on rigorous circuit correctness and cryptography. As long as at least one honest node exists in the network, fraudulent proofs are difficult to disseminate widely.
Highly Distributed Nodes & Scalability: PoZK encourages global participation with varying hardware specs, preventing computing or staking from being overly concentrated in a few entities. The network can scale adaptively, significantly boosting resilience to external attacks and single points of failure.
Different ZK proof schemes have widely varying demands for CPU, memory, and parallelism. To accommodate diverse real-world deployments, Zypher Network offers multiple ZK proofing choices (e.g., Plonk, Groth16, Risc0 zkVM):
Minimum Configuration:
CPU: 4 cores
Memory: 32 GB
Bandwidth: 1 Gbps Suited for small-scale or experimental setups, offering feasibility for lightweight tasks or test networks.
Professional Nodes:
CPU: 96+ cores
Memory: 768+ GB
High-speed network connections (10 Gbps or higher) Ideal for large-scale AI Agent inference, frequent DeFi trading, real-time PvP gaming, etc., dramatically reducing proof generation and verification latency.
