What is elastic-streets?

Elastic Streets creates animated videos of city maps where the distance between any two points on screen corresponds to how long it actually takes to drive between them at that hour of the day. During rush hour, when travel times are long, the city stretches and swells. In the middle of the night, when roads are clear and drives are quick, the city compresses back down. The result is a 9-second looping animation showing 24 hours of a city breathing in and out. The project works by taking real driving speed data, either from Uber Movement records (available for Manhattan and Seattle from 2019) or from Mapbox traffic estimates (usable for any city). For each of the 24 hours in the day, it runs a mathematical optimization that repositions every street intersection so that the screen distances match the real drive times as closely as possible. It then smoothly interpolates through the 24 positions to produce the final animation. The pre-computed layouts for Manhattan, Seattle, and Los Angeles are already included in the repository, so reproducing the existing videos requires only two commands and no data downloads. Rendering takes roughly 10 to 30 minutes on a standard machine. An interactive browser-based viewer is also included, letting you step through the 24-hour animation in a web page. For cities beyond the three included, the project provides scripts to pull traffic estimates from Mapbox and re-run the full computation from scratch. The README also notes that the project folder includes a CLAUDE.md file that gives Claude Code enough context to rebuild or extend any of the animations conversationally, without needing to read through the scripts manually.