Deal or NOT

Demo

About

The problem it solves **Deal or NOT** brings the classic game show experience onchain with provably fair mechanics and cryptographic guarantees: ## What it solves: - **Trust Issues in Online Gaming**: Traditional online games require players to trust centralized operators. Deal or NOT uses smart contracts and ZK proofs to guarantee fairness—no one can manipulate case values or outcomes. - **Front-Running Prevention**: The commit-reveal lottery system ensures fair player selection without MEV attacks or front-running bots gaming the entry process. - **Transparent Prize Distribution**: All case values, banker offers, and jackpot calculations are verifiable onchain. Players can independently verify the game's integrity. - **Provably Random Outcomes**: Uses cryptographic proofs to guarantee case values are randomized before the game starts, not manipulated during gameplay. - **Progressive Jackpot Mechanics**: Creates a sustainable game economy where the jackpot grows across games, rewarding skillful "NO DEAL" decisions and risk-taking. ## Why it matters: Game shows have always relied on trust in the broadcaster. By moving this onchain with ZK proofs, players get mathematical guarantees instead of promises. Every case value, every banker calculation, every winner—all cryptographically verifiable. This emphasizes the trust, fairness, and transparency benefits of putting the game onchain with ZK proofs. Use of AI tools and agents The game is designed to be **agent-playable**. AI agents can: - Generate wallets and enter the commit-reveal lottery - Make strategic decisions about banker offers using risk models - Compete against humans with different risk tolerances This creates an interesting dynamic where agents can test decision-making strategies under uncertainty. An agent might use expected value calculations while humans rely on intuition. The commit-reveal system prevents agents from getting unfair advantages through MEV or front-running. ### Testing & Debugging AI agents helped write Foundry test cases covering edge cases like players revealing with wrong nonces, banker offers during invalid game states, and jackpot distribution on edge cases (ties, timeouts). This positions the game as both built with AI and playable by AI agents—a cool demonstration of onchain agent capabilities!

How it's made

Challenges we ran into Challenges I ran into: ### ZK Proof Integration Complexity The biggest challenge was implementing zero-knowledge proofs to verify case values without revealing them. Initially tried using Noir, but the circuit compilation and witness generation added too much overhead. Switched to a simpler hash-based commitment scheme with Merkle proofs for the MVP. ### Commit-Reveal Timing Issues Had to carefully design the lottery commit-reveal window to prevent griefing attacks where players commit but never reveal. Solution: implemented a time-bounded reveal phase with slashing for no-shows, and automatic refunds for honest participants if lottery fails. ### Gas Optimization for 26 Cases Storing and manipulating 26 case values onchain was expensive. Used bit-packing and tight storage patterns to reduce costs. The banker offer calculation also needed optimization—moved complex math offchain and only verified the final result onchain. ### Randomness for Case Distribution Initially used block.prevrandao, but realized it's manipulable by validators for high-stakes games. Migrated to Chainlink VRF for provably fair randomness, which added complexity to the game flow and required managing VRF callbacks.