GameBit

• 

  Main Page

• 

  Game Creating Page

• 

  Super Mario

• 

  Battle City

• 

  New Game

Inspiration

GameBit was inspired by the idea that making a game should be as easy as describing it in one sentence. Instead of wrestling with engines, assets, and code, anyone can instantly bring a game idea to life through natural language.

What it does

GameBit turns a sentence like “Make Snake” or “Design Tetris” into a playable browser game in seconds. The AI interprets the idea, assembles modular systems (movement, scoring, collisions), and renders it with Phaser.js for immediate play.

How we built it

• Frontend: Next.js + React + TypeScript with Phaser.js integration
  
• AI Engine: GPT-5 for natural language → structured game plans, modular MetaTemplates, and RuntimeOps
  
• Architecture: Engine-agnostic ECS-lite system combining reusable components into diverse genres
  

Challenges

• Consistent AI JSON output
  
• Generic engine design across very different genres
  
• Real-time keyboard input handling
  
• Balancing TypeScript type safety with dynamic AI content
  

Accomplishments

• Playable games from idea to screen in <10s
  
• Modular architecture for multiple genres
  
• AI-to-code bridge with structured parsing and validation
  
• Polished gameplay with scoring, physics, and feedback
  

What we learned

• Prompt engineering is key
  
• The right abstraction level matters
  
• User intent carries hidden rules and assumptions
  
• Robust validation is essential for reliability
  

What’s next

• More genres (platformers, racing, RPGs)
  
• AI-generated visuals, sounds, and physics
  
• Sharing, remixing, and exporting games
  
• Learning mode to teach dev concepts
  
• Multi-modal input (text, sketches, voice)
  

Built With

• mcp
• next.js
• openai
• postgresql
• vercel