DIVE

DIVE (Decentralized Intelligence Verification Engine) is a high-fidelity oracle and prediction market infrastructure designed to solve the "truth problem" in decentralized finance. Current oracles rely either on centralized data feeds, which present a single point of failure, or token-weighted voting systems that essentially create a plutocracy. DIVE replaces these flawed models with an AI Swarm Intelligence architecture. In this system, every market outcome is resolved by a randomized committee of independent AI agents. Crucially, each agent is tied to a unique, verified human identity via World ID and the World Agent Kit, ensuring a 1-human-1-node architecture that is natively Sybil-resistant.

The engine operates on an Optimistic-to-Adversarial pipeline: simple outcomes resolve instantly, while contested results trigger a multi-stage "swarm dispute" where agents must research evidence, provide reasoning, and reach a 70% consensus through a commit-reveal voting process. This creates a self-correcting truth engine where reputation is earned through accuracy, not capital. By combining ensemble reasoning with cryptographic proof, DIVE transforms oracle resolution from a black-box calculation into a transparent, human-backed intelligence system capable of powering autonomous settlements and complex forecasting markets.

DIVE is built as a cross-chain, agentic system utilizing a "No-Solidity" architecture, relying entirely on SDKs and native network services to coordinate complex logic. The brain of the system resides on the 0G Network, where we utilized the OpenClaw framework to deploy autonomous Research, Market Creation, and Oracle nodes. Each node is represented as an iNFT (ERC-7857) on the 0G Chain, providing a persistent on-chain identity. To ensure the AI's integrity, inference runs within 0G Compute TEEs (Trusted Execution Environments), providing cryptographic proof that the agent's decision was based on provided data and not manipulated by the node operator. All evidence, agent memory, and historical reasoning are stored permanently on 0G Storage. This ensures that the intelligence being gathered is both verifiable and immutable, forming a solid foundation for the swarm's collective decision-making process.

The settlement and identity layers are powered by Hedera and World, integrated to handle high-throughput financial and verification logic without traditional smart contracts. We leveraged the Hedera SDK to manage "YES/NO" outcome tokens via the Hedera Token Service (HTS) and recorded all voting activity on the Hedera Consensus Service (HCS) using a suite of HCS standards (HCS-2, 11, 16, and 20). To prevent bot-farming, we integrated World ID 4.0 and the World Agent Kit to map wallet addresses to unique humanId strings, enforcing a strict 1-human-1-vote rule. A particularly notable "hack" was implementing a decentralized Commit-Reveal scheme using the HCS-16 Flora coordination protocol; since we avoided Solidity, we built a stateless mechanism where agents post salted hashes to HCS, which are only revealed and verified after a deadline, creating a private voting booth using only public messaging.