How AI Replaced Traditional Coding to Build a Full NYC Map
How AI Replaced Traditional Coding to Build a Full NYC Map
A DeepMind engineer has created a full, interactive map of New York City without writing traditional code. The project presents NYC as an isometric pixel world, similar to classic city-building games like SimCity 2000, allowing users to explore streets, blocks, and buildings directly in a web browser.
Instead of manual drawing or conventional programming, the map was built using a data-driven and AI-assisted process. Detailed 3D city data from sources like CityGML and Google Maps was first rendered from an isometric viewpoint to form the city’s structure. These renders were then converted into pixel-style visuals using generative AI models, including Nano Banana Pro and a customized Qwen Image-Edit model.
To maintain accuracy, building positions and metadata were carefully tracked so each generated scene aligned correctly across the entire city. The final result sits between art and engineering — neither hand-drawn nor traditionally coded.
The project highlights how AI agents can now replace complex manual workflows, opening new possibilities for large-scale visual world creation.
