A developer used Claude Code and Opus 4.5 to build an interactive map of San Francisco parking regulations. The tool surfaces street cleaning, meter, and permit data to help drivers avoid tickets. It was born from a personal commute headache and the frustration that such a straightforward public data project was never built.

The author describes this as a new era of bespoke software, where LLMs act as high-level compilers. He delegated frontend work to Claude, focusing on backend and requirements. This approach turned a daydream into a viable weekend project, demonstrating how AI can act as a force multiplier for developers with limited time.

The app ingests GeoJSON from data.sfgov.org and sfmta.com, but raw files were too large for mobile. A custom Go tool pre-processes the data hourly. Initially built with Leaflet, the author later rewrote it in pure JS, optimizing coordinate systems for performance and simplifying the codebase.