Agentic Game of Life

A multi-agent evolution simulation demonstrating elizaOS's ability to run autonomous agents without an LLM.

Key Features

• Self-Replicating Agents: 40+ agents with DNA that mutates during reproduction
• Emergent Evolution: Natural selection favors survival strategies over generations
• No LLM Required: Custom model handlers implement agent decision-making
• Real-time Visualization: Watch the ecosystem evolve in your terminal

Implementation

TypeScript entry: game.ts (terminal simulation, no LLM).

How It Works

Agents have DNA (speed, vision, aggression, metabolism) that determines behavior:

1. Perceive: Scan for nearby food and other agents
2. Decide: Choose to move toward food, flee from threats, or hunt prey
3. Act: Move, eat food, fight, or reproduce
4. Evolve: Offspring inherit mutated DNA from parents

Watch emergent behaviors like predator-prey dynamics, population cycles, and evolution of traits!

Quick Start

# Standard mode
bun run examples/game-of-life/game.ts

# Fast mode (10x speed)
bun run examples/game-of-life/game.ts --fast

# With statistics
bun run examples/game-of-life/game.ts --stats

Architecture

┌─────────────────────────────────────────────────────────┐
│                    AgentRuntime                         │
│                 (Anonymous Character)                   │
├─────────────────────────────────────────────────────────┤
│  plugins:                                               │
│  ├── plugin-sql          (persistence)                  │
│  ├── bootstrap-plugin    (basic capabilities)           │
│  └── game-of-life-plugin (agent decision handlers)      │
├─────────────────────────────────────────────────────────┤
│  Each tick, for each agent:                             │
│  perceive() → decide() → act() → evolve()               │
│                                                         │
│  All decisions made algorithmically (no LLM!)           │
└─────────────────────────────────────────────────────────┘

What Makes This Cool

• Digital Life: True self-replicating agents with heredity and mutation
• Zero Cost: No API calls, pure algorithmic simulation
• elizaOS Showcase: Custom model handlers, plugin system, agent runtime
• Emergent Complexity: Simple rules → rich ecosystem dynamics