← Back to home

EthereumFighter

Pin AI-Agents against each other to make them compete in DeFi trading competition using FHE.

Problem Statement

(EF) - Ethereum Fighter leverages fully homomorphic encryption (FHE) to ensure complete privacy during battles:Players' positions and balances remain encrypted while on-chainTrading strategies stay hidden from opponents until executionContract interactions are encrypted but verifiable on demandFinal outcomes are provably fair while preserving player privacyThis revolutionary approach allows for true strategy-based competition without information leakage.

Solution

🎮 The game starts with connecting players wallets, based on connected wallets you can play either 1vs1 or join public islands to see other players discovering Taiwan. The map of the game holds several action points, mostly hidden around temples. Users can interact with temples to find competition details defining specific chain, duration and cost.Once 2 players join the same Temple players can proceed with AI-Agent selection, based on actual pubic figures there are few interesting options to choose from. Actual AI agent competition begins once both players confirmed their Agent selection, but it is also plan to allow users connect their own agents as long as they meet the rules interacting with the resolver contract.The resolver contract helps to determine the trading competition winner. This is done by allowing agents to trade against real market data delivered by Chainlink, but with fictional balances. Serving also as a educational purpose. Whoever has more USD equivalent at the end of the competition that contract can count wins the competition.Our contract also supports FHE allowing for unique game-play. If players agree to switch to blind mode (FHE) this hides opponent overview on UI, but also utilizes encrypted data for on-chain submission of the transactions. Since FHE supports calculation on encrypted data without revealing it we can still consume Chainlink price-feed and keep player balances hidden until trading competition with fictional balances is over. This allows unique game play where users can increase their stakes in the game.Each agent also has pre-defined commands it can easily operate with when it comes to quick swaps simulation. The winner also moves up in leader boards allowing to find the best AI Agent out there to take donatinos stored in the Temples.🎮 Immersive UI ExperienceThe game features a nostalgic arcade-style interface with modern crypto elements:Character selection screen with detailed agent profilingReal-time battle visualizations for trade executionVintage fighting game aesthetics with blockchain data overlaysDual-player setup for direct head-to-head competitionSupport for spectator mode to watch high-stakes battlesAgents run through Nillion secret LLM and OpernRouter. The main point is to have fun and try out different things onboarding new users into AI agents, tading in DeFi and interacting with FHE.FHE is used to hide participan trades and allow for unique game-play with bluffing capabilities if one of the players would like to raise game stakes during match and even win being behind in trade outcome. Users can choose which mode they like to compete - blind (FHE) or open traditional trading where winner determined by final USD equivalent of AI agent balances stored on the contract internal books (without needing to use actual liqudiity) It is a great way to get an idea how to control AI agetns and at the same time perform buys/sells on popular token pairs.

Hackathon

ETHGlobal Taipei

2025

Prizes

  • 🏆

    Integrate Fusion+ into your project3rd place

    1inch

  • 🏆

    🏆 ETHGlobal Taipei 2025 Finalist

    ETHGlobal

  • 🏆

    Celo L2 Pool Prize

    Celo

Contributors