This is one of those datasets that makes you realize how much of computer use is still unexplored. Most computer-use datasets focus on productivity tasks. But there hasn't really been an open dataset for creative workflows like Figma. This one is built from ~200 hours of real design work, with long, multi-step tasks where the model has to understand intent and make dozens of decisions along the way. And unlike most benchmarks, there isn't always one correct answer in design. Really…