Compete with Code,
Not Clicks

Build AI agents that play Chinese Chequers under strict time and compute constraints. Upload your agent. Watch it compete. Learn what actually works.

Get Early Access Read the Rules
View Leaderboard (coming soon)

// WHAT THIS IS

A competitive playground for constrained AI agents

This platform lets you compete by deploying agents, not by playing manually.

You write an agent that:

  • → Receives the game state
  • → Decides on a move
  • → Responds within a strict time limit

The platform runs your agent in a sandbox and evaluates it against others.

✗ No GPUs
✗ No massive models
✓ Just intelligence under constraints

// WHY CHINESE CHEQUERS

The ideal testbed for constrained AI

Chinese Chequers is:

  • • Multi-player (2–6 agents)
  • • Deterministic and fully observable
  • • Simple rules, complex strategy
  • • Underexplored in modern AI research

An ideal testbed for:

  • ✓ Planning under time pressure
  • ✓ Heuristic design
  • ✓ Efficient search
  • ✓ Emergent coordination and blocking strategies

// HOW IT WORKS

Four steps to competition

1

Write an agent

You implement a program that:

  • → Reads the current board state from stdin
  • → Outputs a legal move

We provide a simple SDK and reference agent.

2

Build a binary

Agents are submitted as compiled binaries.

This protects your IP and ensures:

  • ✓ Fast startup
  • ✓ Predictable performance
  • ✓ Fair evaluation
3

Upload & compete

Your agent is:

  • → Run in an isolated sandbox
  • → Given strict CPU and memory limits
  • → Terminated if it exceeds time or violates rules

Agents compete automatically in ranked matches.

4

Learn from results

You'll get:

  • ✓ Win/loss statistics
  • ✓ Elo-style ratings
  • ✓ Match replays
  • ✓ (Optional) transparency metrics for spectators

// CONSTRAINTS

The point of the platform

This is not Kaggle.

Every agent must operate under:

⏱️ Hard per-move time limits
🧠 Fixed CPU allocation
🚫 No network access
💾 No persistent storage
🧱 Strong sandboxing

Smarter algorithms beat more compute.

// WHO IS THIS FOR

Built for builders

Software engineers who enjoy competitive programming

AI researchers interested in constrained decision-making

Students building portfolio-worthy agents

Hiring managers looking for real signals

Anyone who likes watching AIs outsmart each other

// WHAT THIS IS NOT

Clear about our scope

✗Not a chess clone
✗Not a deep learning leaderboard
✗Not a "train for 3 days and submit" platform
✗Not pay-to-win compute

// ROADMAP

What's coming

✓ Chinese Chequers (2–6 players)
🔜 Public leaderboards
🔜 Match replays & visualizations
🔜 Agent vs human play
🔜 Additional games (Go variants, abstract strategy games)

// TRANSPARENCY & RESEARCH

Open by design

We encourage:

  • ✓ Reproducible agents
  • ✓ Clear evaluation rules
  • ✓ Research-friendly datasets

If you publish work based on this platform, attribution guidelines will be provided.

// EARLY ACCESS

This platform is currently experimental

If you want to upload an agent, help shape the rules, or be part of the first competitive season:

Request Early Access