GeoJSON Map Viewer
Simon Willison’s Weblog Simon Willison
Simon Willison built a GeoJSON map viewer after needing to show local boundary files and export them as PNGs. It started as a simple ask and turned into a tool that can place multiple boundaries on one map.
Based on reporting by Simon Willison’s Weblog, Simon Willison — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Simon Willison needed a way to show GeoJSON files on a map and export the result as a PNG while helping Natalie gather maps of local political boundaries. The boundaries in question included the Granada Community Services District and the Midcoast Community Council, so this wasn’t a toy problem. It was a very practical one, and it needed a working result fast.
He asked GPT-5.6-Sol for tool suggestions, and it went further than expected: it proactively built one. After a few iterations using Claude Code for web and Fable 5.1, the result became the finished GeoJSON Map Viewer. That’s the kind of workflow that feels a little like software being assembled in real time, because in this case that’s basically what happened.
The GeoJSON itself came from another place entirely. Willison says that if you ask ChatGPT Work for boundaries for almost anything, it will grind through Government data sources, extract and combine files, and produce what you need. He used that process to get a polygon for the exact boundary of the El Granada GCSD, and another GeoJSON file for the Midcoast Community Council near Half Moon Bay, CA.
The end result is a link that displays both boundaries together in the new map viewing tool. Simple enough on the surface, but the interesting part is how many pieces had to line up: a mapping need, generated boundary data, and an AI-built tool that could make the whole thing visible in one place.
My take — AI-written commentary, not fact-checked reporting
This is the sort of AI use that actually makes sense: narrow, concrete, and tied to a real task instead of some grand demo about productivity. The bigger story is that the dull stuff — finding boundaries, stitching files together, making a map exportable — is where these systems keep proving their worth. Everyone loves the flashy model release; the boring utility is where the useful future keeps hiding.
Read more about this at: Simon Willison’s Weblog