bitsofintelligence101-lab/games

This repository is an experiment comparing how two different AI models build the same game when given identical, detailed instructions in a single attempt. Rather than being a game itself, it is a collection of independent implementations meant to be compared side by side, so anyone curious about AI coding ability can see how two different models solve the exact same problem. Each game idea was written up as a detailed specification first, then handed unchanged to two AI models, referred to in this repo as Fable and K3. Every resulting project lives in its own separate folder and shares nothing with the others, which lets someone directly compare architecture choices, file organization, code quality, and bugs between the two models' attempts. There are two game concepts covered: Cats versus Dogs, a browser based real time strategy game in the style of Command and Conquer with resource harvesting and base building, and Rocket Arena, a local multiplayer style game inspired by Rocket League with cars, a ball, and boost mechanics. The folder names follow a pattern combining the game and the model, so cc_fable and cc_k3 are two versions of the same strategy game, while rl_fable and rl_k3 are two versions of the same sports game. The Fable built strategy game runs as a single HTML file with no setup needed, while the K3 version uses a more modular file structure and needs a simple local web server to run because of how it loads its code. Both versions of the sports game are Node.js projects built with the Three.js 3D graphics library and the Rapier3D physics engine, started with a couple of standard install and run commands, or alternatively through an included Dockerfile. Each individual project folder has its own README with specific setup instructions, controls, and notes on how that particular build was constructed, so this top level README mainly explains the comparison structure rather than any single game's mechanics.