My personal AI benchmark: "Generate an SVG of a frog with a Habsburg jaw."
Summary
An in-depth look at a month-long AI benchmark where 14 models were prompted with a single instruction to generate an SVG of a frog with a Habsburg jaw. The article catalogs 42 runs and 42 SVG outputs, highlighting tendencies such as editorialization and, in some cases, royal framing, as well as differences in determinism across models. It offers insights into prompt design and model behavior that are useful for AI tool users and content creators.