The core problem Rivalz addresses is that today’s powerful AI systems operate in silos, disconnected from the vast ecosystem of real-world resources, diverse data streams, physical infrastructure, and on-chain economies. This fragmentation prevents AI from reaching its full potential, as it lacks a secure, verifiable, and standardized way to access and coordinate these assets, while blockchains lack the native ability to interact with this off-chain intelligence and utility.
Solution Overview
Rivalz solves this by building a World Abstraction Layer, a decentralized infrastructure designed to connect AI with the world’s resources. Our solution for this hackathon focuses on the connectivity layer of this system: the Agentic Data Coordination Service (ADCS). ADCS is a network of AI oracles that serves as a developer-facing tool, allowing smart contracts to securely request and consume outputs from any off-chain AI model or data source. It provides the core Web3 infrastructure for building complex AI workflows directly on-chain.
Project Description
Our project is the deployment and integration of Rivalz’s ADCS infrastructure on the Hyperion network, providing a foundational tool for developers to receive verifiable AI insights directly on-chain.
Core Functionality: ADCS allows developers to bring any AI workflow on-chain. Through our system, a smart contract can call an “Adapter”, a configurable template that defines a specific AI task. This Adapter orchestrates off-chain AI and Data Providers (e.g., LLMs, predictive models, data feeds) to execute the task and delivers a verifiable result back to the smart contract, enabling it to act on that intelligence.
User Interaction & Benefits: For developers, ADCS is a powerful infrastructure service that drastically simplifies building with on-chain AI. Instead of creating bespoke oracle and AI pipelines, they can use our platform to easily compose and call upon complex AI workflows. End-users benefit from the resulting dApps, which will be smarter, more automated, and capable of a new range of intelligent on-chain actions.
What excites us: We are excited to provide the Hyperion ecosystem with a fundamental piece of Web3 infrastructure that empowers builders. Fusing ADCS with Hyperion’s high-performance capabilities will create the perfect environment for developers to pioneer the next generation of sophisticated on-chain AI applications.
Community Engagement Features
To gamify the platform and onboard developers and users to ADCS on Hyperion, we will launch the “Rivalz AI Builder Campaign”:
Testable Features/Tasks & Points System:
Web3 Infrastructure Track Task: Guide developers to use the ADCS UI to create a new custom AI service (Adapter). (Task: Create 1 Adapter | Reward: 50 Points)
Onchain AI & Data Track Task: Challenge developers to deploy a simple smart contract on the testnet that calls an ADCS Adapter, demonstrating an on-chain AI workflow. (Task: Deploy 1 Consumer dApp | Reward: 1000 Points)
Community User Task: Allow users to interact with a demo dApp (e.g., an AI Wallet Summarizer) that makes a request to ADCS. (Task: Use demo dApp | Reward: 250 Points)
Gamification and Onboarding:
This points-based system creates a leaderboard and incentivizes participation at all levels. It guides developers through using our core infrastructure and building on-chain AI workflows, while allowing general users to experience the power of the dApps created, effectively onboarding the entire community.
Getting Involved
Community members can get involved with our project in two key ways:
Developers can directly contribute by utilizing the ADCS infrastructure to build their own innovative dApps and on-chain AI workflows on Hyperion.
General users will engage with ADCS indirectly every time they use the growing number of dApps and services within the ecosystem that are built on our AI oracle platform.
How does ADCS ensure the verifiability of off-chain AI outputs when injected on-chain? Is there a cryptographic proof, commit-reveal system, or consensus layer validating responses?
Hi @han, glad to know that Rivalz looks interesting to you! Regarding your questions:
1. Are the dApps using this system truly AI-powered, or are they just passing data through from off-chain sources?
They can be both. ADCS is a flexible system that supports simple data pass-through from providers (like fetching a token price), but its core power lies in enabling truly AI-powered dApps. Developers can create “Adapters” that chain together data providers and AI models (like LLMs), allowing smart contracts to request and act on sophisticated AI analysis, sentiment scores, or generative content, not just raw data.
2. Can anyone define and deploy their own AI tasks on-chain, or does it require deep technical knowledge?
Defining an AI task is designed to be simple. A user can define a new task by creating an “Adapter” through the ADCS user interface without deep technical knowledge of AI infrastructure. This involves selecting from available AI/data providers and writing a simple text prompt. However, to use that AI task, a developer is still needed to write and deploy the smart contract that calls the ADCS oracle. Example smart contracts using ADCS: ADCS-core/contracts/src/mock at main · Rivalz-ai/ADCS-core · GitHub. These are relatively simple but soon we will create a couple more interesting demos using combinations of multiple data and AI providers.
3. Will there be a way to discover dApps built on this infrastructure—like a showcase or directory?
Yes, there will be a section on the platform highlighting these dApps with the next v2 update with a better UI and better onboarding experience for both data/AI providers and dApp builders.
After the off-chain components process a request, there is a dedicated “Upload Proof On-chain” step. This proof, which validates the entire process from request to result, is uploaded to the Rivalz Rollup.
Let me know if you have more questions, happy to answer.
Thanks for the detailed answers - really appreciate the clarity!
The Adapter system sounds like a solid balance between flexibility and accessibility. Making AI capabilities modular for smart contracts is no small feat. Quick follow-up: in cases where multiple AI models are chained together, how do you handle latency or conflicting outputs, especially for time-sensitive dApps?
Just one more question : Beyond the AI Wallet Summarizer demo, what real-world use-cases do you think ADCS is uniquely positioned to enable that weren’t possible before?
We handle latency and conflicting AI outputs through the flexible design of our Adapters. There are a few ways to set it up:
Here’s some examples:
Pipeline Style: You can chain models together, like an assembly line. First, you feed data to AI model ‘A’ for an initial analysis, then pass its result to model ‘B’ for the next step, like refining the idea or making a final decision.
“Second Opinion” Style: To handle conflicting results, you can run data through model ‘A’ and model ‘B’ at the same time. You can even have another AI analyze both outputs to pick the best one or create a single, combined answer.
Depending on how many adapters and providers you want to use you can have a system that implements both or even more complex logic.
When it comes to speed, the AI models themselves are the biggest factor (e.g., Gemini Pro vs. Gemini Flash). Since that’s up to the provider, our approach is to give developers choices and use the model(s) that better serves their needs.
These are some of the more general use cases we have in mind. We will be working closely with dApp developers to generate more ideas and establish what features, data and AI providers are in high demand so we can plug them within our product.
Automated DEX & LP Management
Enable DEXes to use AI oracles for dynamic fee adjustments, optimized liquidity provision, and mitigating impermanent loss based on predictive analytics.
Dynamic Lending & Risk Assessment
Empower lending platforms with AI oracles for real-time risk assessment of collateral or borrowers, allowing for dynamically adjusted LTVs and interest rates.
AI-Powered Yield Optimization
Create autonomous DeFi vaults that use ADCS to analyze market conditions and rebalance portfolios or shift yield farming strategies to maximize returns.
On-Chain Generative AI
Facilitate dApps where smart contracts can request the creation of AI-generated content, for example, minting dynamic NFTs whose art or metadata is generated by an AI model based on on-chain events.
Smart Contract-Based AI Agents
Enable the creation of autonomous on-chain agents. These smart contracts can use ADCS to gather information, analyze it with AI, and execute complex tasks or transactions without direct human intervention.
AI-Assisted Reputation Scoring
A service that provides a simple on-chain reputation or trust score for a wallet address. ADCS would analyze the wallet’s transaction history and on-chain behavior with an AI model to generate this score, which can be used for things like Sybil resistance or building trust in p2p interactions.
