At Hyperbolic, we’re focused on making decentralized AI and other advanced services scalable and secure. Traditional verification tools—like opML or zkML—often force developers to choose between efficiency and security. Our Proof of Sampling (PoSP) protocol, developed with researchers at UC Berkeley and Columbia University, solves this tradeoff by delivering high security, minimal overhead, and predictable incentives for honest behavior.
EigenLayer, a restaking platform that extends Ethereum’s security guarantees to Actively Validated Services (AVS), now integrates PoSP. With this integration, AVS built on EigenLayer gain access to a robust, cost-effective verification mechanism that aligns all participants around honest computation. The result is secure, scalable, and verifiable AI services delivered at a fraction of the cost.
Why Proof of Sampling (PoSP)?
Proof of Sampling (PoSP) ensures that verifying computations—such as AI inference or other data-intensive workloads—remains both secure and efficient. It’s a sampling-based protocol that reduces the need to verify every single result. Instead, it selects certain results at random, verifies them, and imposes clear penalties for dishonest behavior.
This approach mirrors the logic of random ticket inspections in a public transport system. Just as random checks and fines encourage all passengers to buy tickets, PoSP’s random audits and penalties incentivize all network participants to remain honest. Under PoSP, it’s economically irrational to cheat, and honesty becomes the best strategy.
Key PoSP Advantages
Efficient Verification: PoSP adds less than 1% computational overhead, allowing nodes to continue operating at near-native speeds.
Robust Security: Participants know random checks can occur at any time, making fraud too risky to be profitable.
Game-Theoretic Incentives: PoSP is designed so that honest behavior is always the rational choice, achieved through a pure strategy Nash Equilibrium.
Scalable for AI Services: PoSP can handle large-scale decentralized AI workloads, ensuring that high-performance compute and AI inference remain verifiable and trustworthy.
Integration with EigenLayer
EigenLayer allows developers to create AVS that leverage Ethereum’s existing validator set and security model. By integrating PoSP, EigenLayer provides AVS with:
Randomized Validation: Outputs are randomly audited, guaranteeing unbiased verification without overloading the network.
Reduced Costs: Minimal overhead and efficient verification reduce the cost of running secure, large-scale AVS.
Verified AI Services: AI inference outputs can be continuously and reliably validated, bringing real trust to decentralized AI systems.
For developers, this means they can confidently build AI-driven services—such as AI model inference or data-heavy computations—knowing every output is at risk of random verification. This deters misbehavior and maintains a secure, predictable environment for decentralized workloads.
Concrete Results
Security at Scale: Even as AVS usage grows, PoSP ensures fraud-proof verification.
Lower Costs: Developers save up to 75% compared to centralized solutions.
Performance: PoSP delivers verifiable results without slowing down AI pipelines.
Competition with Centralized Solutions: AVS secured by PoSP now rival centralized alternatives, while retaining decentralized benefits.
Future Applications
PoSP can enhance a wide range of AVS scenarios built on EigenLayer:
AI Inference Validation: Trust that each inference output is correct.
AI Training and Data Verification: Maintain the integrity of training data and outputs.
Data Availability & Rollups: Securely verify that L2 computations remain accurate.
Cross-Chain Operations: Extend PoSP’s fraud resistance to interoperability solutions.
By integrating PoSP into EigenLayer’s AVS restaking, Hyperbolic helps developers build secure, scalable, and cost-efficient decentralized AI services. PoSP sets a new standard for verification—one where honesty pays, computation scales, and verification overhead is kept low.
About Hyperbolic
Hyperbolic is democratizing AI by delivering a complete open ecosystem of AI infrastructure, services, and models. Through coordinating a decentralized network of global GPUs and leveraging proprietary verification technology, developers and researchers have access to reliable, scalable, and affordable compute as well as the latest open-source models.
Founded by award-winning Math and AI researchers from UC Berkeley and the University of Washington, Hyperbolic is committed to creating a future where AI technology is universally accessible, verified, and collectively governed.
Website | X | Discord | LinkedIn | YouTube | GitHub | Documentation
