Developer Eric Sir created Munimet.ro, an innovative web app that uses machine learning to help San Francisco commuters decide between subway and bus. Inspired by a recent subway outage, this project aims to provide real-time transit choices by analyzing publicly available circuit diagrams.

Using Claude Code, Sir automated the download of real-time subway images from sfmunicentral.com. He then built an image labeler in tkinter, which required extensive manual tweaking. The most challenging part was converting these labeled images into predictive models using pytorch, where he encountered issues with GPU support.

Despite the hurdles, the project showcases the potential of AI in urban transit. Commuters in San Francisco can now make informed decisions about their travel, potentially reducing congestion and wait times. The source code, available on GitHub under an MIT license, invites further development and community contributions.

This initiative reflects a growing trend of developers using AI to solve local issues. As cities become smarter, such tools could become integral to urban mobility, offering personalized and predictive transit options.