Research Playground

Where we experiment, learn, and share the future of AI-powered development

Vibe Coding

The following project was created using a methodology called 'Vibe Coding' popularized by Andrej Karpathy. Vibe coding is where you pair-program with AI, except you let the AI do 99-100% of the actual coding. If you have an error, you simply copy and paste it to your chat interface. When this is paired with tools that allow agents to automate the creation of files and spinning-up of servers, this is extremely powerful, and very fast. We created the beta for Green Metropolis in about 4 hours using this methodology. A typical software beta for a full-featured application can take weeks or months.

Latest Project: Green Metropolis

Creating a Sustainability Simulation Game Using Vibe Coding with Claude 3.7

Green Metropolis is a fully-functional sustainability simulation game that demonstrates how AI-assisted development can democratize software creation. This game was built through collaborative coding with Claude 3.7, showcasing how developers of all experience levels can create complex, meaningful applications. NONE of the code was written by a human. It was all generated by AI. None of the code was scrutinized by a human. Our Developer simply copied and pasted the code into their coding environment and created files as requested. AI did the rest, even generating prompts for us to use to generate assets with other AI systems.

By publishing this project, we're opening new pathways for developers to learn production-ready techniques beyond "tutorial hell" while addressing important global challenges through technology.

Methodology - it all starts with a simple prompt

  1. We used Grok in order to generate 10 ideas for a game based on trending ideas. 'generate 10 ideas for a web-based game based on trending ideas' was the prompt that started it all.
  2. We then used Claude 3.7 to generate the code for the game. We simply copied and pasted the idea to Claude and asked it to begin creating the game. It generated the code for the game, even going as far as organizing it into files we could download and move to our coding environment.
  3. When it came to generating assets (images) we asked Claude to let us know the name of the image (so we could name our files to match the code) and asked for a prompt that was suitable for AI to generate the image. We then copy/pasted this prompt to Midjourney in order to generate assets.

Key Features

  • Complex system of metric-based tiles such as 'Luxury Apartments' each with its own bonuses such as income or power generation.
  • Real-time resource management and consequence visualization
  • Adjacency bonuses and penalties. Putting a factory next to an apartment building creates more pollution. A park next to a factory decreases it.
  • Modular, extendable code. This fits well into one of AI's greatest limitations - the context window. We can start up new conversations to deal with minor bugs using modular code once the application grows too large to fit the whole thing into a single conversation.

Technologies Used

  • Vibe coding methodology for rapid, sustainable development
  • Claude 3.7 for collaborative code generation and problem-solving
Try the Demo

Our Research Philosophy

At InfinitIQ Tech, our playground is where research meets practical application. We believe that the future of software development lies in the collaborative space between human creativity and AI capabilities.

Every project we undertake in this space aims to:

Upcoming Research

We're constantly exploring new ways to democratize software development through AI collaboration. Stay tuned for upcoming projects that will showcase:

Join Our Research Community

We believe in open collaboration and shared learning. Our research playground is open to developers of all experience levels who want to contribute to the future of AI-assisted development.

Explore our GitHub Repositories to see our current projects and consider contributing your own ideas and code.