✦ Duo — AI-Consensus Challenge Arena

Duo is an elite, decentralized 1v1 challenge and peer-review arena built on GenLayer. It matches developers, writers, designers, and creatives in high-stakes duels where subjective solutions are evaluated by a decentralized panel of AI validators.

Vercel Web App: https://duo-cyan.vercel.app

Contract (GenLayer Studionet): 0xe86efF20d158671bdc20B61De432B228cEAAbCb4

---

🎯 The Vision

Traditional smart contracts are limited to objective, structured inputs (such as numbers, addresses, and boolean states). They cannot read a block of code and judge if it is elegant, or evaluate a product description for creativity. Consequently, online competitive arenas either have no stakes or rely on slow, biased, and centralized human referees.

Duo leverages GenLayer's Intelligent Contracts to execute subjective judgments trustlessly. Multiple validators run Large Language Models and compare their independent evaluations under the Equivalence Principle. This enables an instant, neutral, and secure judging process.

---

🔄 Core Protocol Flow

1. Host a Challenge: A user selects a category (e.g. Coding), defines the prompt parameters, and stakes a set amount of GEN tokens.
2. Opponent Acceptance: Another user matches the staked amount to lock the duel and active matchmaking.
3. Double Submission: Both participants draft and submit their answers independently.
4. AI Verdict & Resolution: A decentralized validator consensus scores both inputs on Quality, Correctness, and Ingenuity. The higher score secures the entire prize pool, automatically transferred by the contract.

graph TD
    %% Nodes and Actions
    A[Challenger] -->|1. open_challenge + stake| B(DuoArena Contract)
    C[Opponent] -->|2. accept_challenge + matching stake| B
    
    A -->|3. submit_solution| B
    C -->|3. submit_solution| B
    
    D[Any Participant] -->|4. evaluate_challenge| B
    
    subgraph GenLayer Decentralized Network
        B -->|Trigger Consensus| E{gl.vm.run_nondet_unsafe}
        E -->|Execute LLM| F[Leader Node]
        E -->|Execute LLM| G[Validator 1]
        E -->|Execute LLM| H[Validator N]
        
        F -->|Leader Output| I{Equivalence Check}
        G -->|Validator Output| I
        H -->|Validator Output| I
    end
    
    I -->|Consensus: Winner matches & scores within +/- 2| J[Payout Winner & Resolve Duel]
    I -->|Consensus Failure: Scores diverged| K[Cancel Tx / Preserves Stakes]

Loading

---

📜 Contract API Reference

The DuoArena contract exposes the following write and view methods:

Write Methods (State Modifying)

• open_challenge(category: str, prompt: str) -> i32 (payable): Opens a new challenge in the arena. The caller must attach a positive GEN stake.
• accept_challenge(challenge_id: str) (payable): Matches the challenger's stake and locks the challenge to the active matching phase.
• submit_solution(challenge_id: str, solution: str): Allows participants to submit their draft response (allowed during matching or submission phases).
• evaluate_challenge(challenge_id: str): Runs the decentralized AI consensus evaluation. It distributes the combined staked pools to the winner and updates the status to resolved.
• cancel_challenge(challenge_id: str): Allows the challenger to cancel an open challenge before any opponent accepts, returning the full stake.

View Methods (Read Only)

• get_challenge(challenge_id: str) -> str: Returns the full state of the challenge serialized as a JSON string.
• get_challenge_count() -> i32: Returns the total number of challenges initialized in the arena.

---

⚖️ AI Consensus & Equivalence Principle

Duo enforces a rigorous evaluation framework utilizing the Equivalence Principle to guarantee fair and deterministic outcomes on subjective inputs:

1. Deterministic Prompts: The contract instructs the LLM to output a strict JSON block structure. Markdown syntax formatting is programmatically stripped and normalized to prevent simple syntax differences from breaking consensus.
2. Equivalence Bounds Check: In validator_fn(leader_result), instead of requiring a byte-for-byte match on the entire LLM response, the validators verify that:
   • The winner is identical.
   • The scores assigned to the Challenger and Opponent are within a ±2 margin of tolerance.
3. No-Nondet-Leak safety: The contract fixes linter reachability violations by ensuring the gl.nondet calls are kept strictly within leader_fn, preventing validators from invoking non-deterministic commands directly in their validation paths.

---

🛠️ Local Development & Setup Guide

Contract Deployment

1. Install the GenLayer CLI:

   npm install -g genlayer

2. Configure the CLI for the Studio Network:

   genlayer network set studionet

3. Deploy the contract:

   genlayer deploy --contract contracts/duo_arena.py

Frontend Setup

1. Navigate to the frontend directory and install dependencies:

   cd frontend
   npm install

2. Start the development server:

   npm run dev

3. Open http://localhost:3000 to view the web application.

---

🎨 UI design & Aesthetics

Duo features a professional, high-end dashboard interface tailored for developers and competitive professionals:

• Obsidian Dark Mode: Built using a deep midnight grey background (#030712) and slate gray typography, creating a premium visual identity.
• Glassmorphic Navigation: Elements incorporate modern CSS backdrop blur filters for a unified, clean overlay experience.
• Visual Performance Metrology: The AI consensus verdict panel utilizes interactive progress-meter bars side-by-side to visually display and compare validator scores.
• Category Colors: Custom gradients and icons differentiate skill categories dynamically (Indigo for Coding, Pink for Writing, Rose for Design, Green for Math, Amber for Trivia).