Google Antigravity & Gemini 3 Pro: Tech Showcase

This project is a technical demonstration designed to test the coding capabilities of the new Gemini 3 Pro model within the Google Antigravity environment.

The goal was to push AI-assisted development to the limit: asking the AI to recreate the logic, textures, and behavior of the classic Windows 3D Maze screensaver using modern web standards (WebGL & Three.js).

Why this demo matters: Rather than just a nostalgic visual, this application serves as a benchmark for autonomous code generation. The maze generation algorithm, the "bot" navigation logic (left-hand rule), and the real-time rendering pipeline were architected and refined using Gemini 3 Pro's advanced reasoning.

Demo Features

• AI-Driven Logic: Procedural maze generation on every refresh.
• Performance: Optimized WebGL rendering running smoothly in the browser.
• Accuracy: Faithful reproduction of the original OpenGL "glitches" and camera movements.

Controls

The demo is self-running to demonstrate the autonomous navigation logic:

• Click: Toggle ambient sound.
• Double Click: Fullscreen mode.

About the Developer

I am a Creative Technologist exploring the intersection of retro-computing and modern AI. This project is part of my research into the practical applications of Google Antigravity for web development.

Inspect the AI-generated code on GitHub: github.com/edumarrom/agy-3d-maze-screensaver