ProofFund
Agentic grants, verified by markets.
Demo
About
The problem it solves Web3 grants are often subjective, slow, and lack accountability for developers after funds are released. Our platform solves this by turning public goods funding into a data-driven, market-based process. When a developer applies for a grant, an AI agent evaluates their past on-chain history using Sourcify to generate a trust score. Based on this score and the proposal, the AI generates strict on-chain KPIs and opens decision markets tied directly to specific funding chunks. Crucially, the developer does not receive all the money at once. Instead, traders bet with real capital on whether the developer will hit their milestones, and these market bets dictate when the money is released. If the market bets "YES" (approves the project), the first chunk of funding is unlocked immediately to start development. After a 14-day development window, smart contracts automatically verify if the on-chain KPIs were met. If the developer delivers, the market resolves, the final chunk of funding is released to the developer, and the winning traders claim their payouts. This ensures funders only pay for proven, measurable results, governed entirely by market consensus.
How it's made
Challenges we ran into System Scale vs. Hackathon Timeline: The project architecture required building four distinct systems from scratch: a BigQuery ML pipeline, a Python AI agent, Layer-2 smart contracts, and a React frontend. Completing the end-to-end integration of all four within 48 hours exceeded the time limit. Middleware Integration: The Solidity contracts (Decision Market, Treasury, KPIVerifier) are fully written and deployed. However, building the Python middleware (web3.py) for the AI agent to securely manage keys, sign transactions, and handle testnet block times required more development hours than we had left. Demo Execution: To prove the concept works, we prioritized the core logic. We kept the Sourcify ML pipeline running live, but temporarily shifted the smart contract execution math (pricing and pool state changes) into the frontend. This allowed us to demonstrate the exact market mechanics without relying on incomplete middleware.