Pythia

Polymarket processed $9B in 2024 election volume. One wallet moved BTC odds 12% with a single $30M position. Another user created 50+ wallets to farm airdrops and manipulate market odds. The platform had no way to distinguish 1 whale from 1,000 real humans.

Pythia fixes this. It enforces true sybil resistance: every bet requires a World ID ZK proof verified on-chain. 1 human = 1 capped bet per market. No whale manipulation. No multi-wallet gaming. The nullifier hash is stored on-chain and can never be reused for that market.

The resolution pipeline is fully automated via Chainlink CRE:

1. CRE Runtime reads active markets and pending resolutions from the Pythia contract on World Chain Sepolia
2. CRE Secrets protects API keys (OpenRouter, event data sources) inside TEE
3. Confidential HTTP fetches real-time event data from 3 sources:
   • CoinGecko for crypto price markets
   • ESPN for sports markets
   • Open-Meteo for weather markets (location-aware, extracts lat/lon from market question)
4. AI Analysis via OpenRouter (Gemini 2.0 Flash) analyzes fetched data against the market question, returns outcome (YES/NO/INVALID) with confidence score
5. On-chain resolution only fires if AI confidence >= 70%. Merkle attestation (bets root + payouts root) submitted after resolution is confirmed

Markets that never reach 70% confidence have a 30-day emergency resolution fallback, preventing locked funds.

Mobile-first UX: Tinder-style swipe right for YES, left for NO. Designed for World Mini Apps distribution to 15M+ World ID holders.

Smart Contract (Solidity + Foundry):

• Pythia.sol (531 lines) handles market creation (6 categories: CRYPTO, SPORTS, POLITICS, WEATHER, ENTERTAINMENT, OTHER), World ID verification, betting with per-market nullifier tracking, AI-powered resolution, proportional payout distribution, and emergency fallback
• 39/39 Foundry tests covering market lifecycle, sybil resistance, odds calculation, resolution, claims, Merkle attestation, and admin functions
• Deployed on World Chain Sepolia: 0x6158fa6bA28a664660B3beb4F8992694dbAD4fAC

CRE Workflow (TypeScript, 830 lines):

• 6-step pipeline: Read markets -> Fetch prices via Data Feeds -> World ID verification layer -> Fetch event data via Confidential HTTP -> AI outcome analysis via OpenRouter -> On-chain resolution + Merkle attestation
• Real keccak256 Merkle tree computation from actual on-chain bet records (not mocked)
• Location-aware weather data: parses city names from market questions and fetches geo-specific forecasts
• ConsensusAggregationByFields for DON consensus on resolution outcomes

Frontend (Next.js + wagmi):

• Swipe-based trading UI for mobile World Mini App
• Live market feeds from on-chain contract
• AI-powered price predictions via Jupiter integration
• Real-time odds display and market statistics