Open Source Project Turns Every City on Earth into a Graph Neural Network

aigleeson · x · 2026-08-20

Background: Urban planners and epidemiologists have long struggled to model cities as living systems. Traditional GIS tools treat them as static layers, failing to simulate complex interactions (e.g., how closing a subway line affects pedestrian flow six blocks away).
Solution: This project models cities as graphs (nodes and edges), enabling complex queries about the dynamic relationships between streets, buildings, transit routes, and human flow.
Availability: The author has released the code on GitHub for free, transforming cities worldwide into Graph Neural Network models.

More from Apps