Simon Willison prompted Claude Fable 5.2 Medium to generate a three.js animation of Yonkers standing up as a mech while stomping around Manhattan, Queens, and Brooklyn. The model correctly interpreted a mobile keyboard typo where “átomos” meant “stomps” and used OpenStreetMap data to render the scene. The resulting video shows the borough rising from the map with a mechanical motion.
This example demonstrates how current text-to-3D tools handle specific geographic constraints and natural language errors without requiring manual correction. It also highlights the ability of these models to combine external mapping data with generative animation instructions in a single prompt. The output serves as a practical test case for developers evaluating spatial reasoning in generative AI workflows.
- The prompt contained a deliberate spelling error for “stomps”
- OpenStreetMap data drove the geographic placement
- Claude Fable 5.2 Medium executed the animation
