Today we're shipping spatial context. This lets you draw directly on your screen + pass it as context to the model. AI can easily analyze screenshots. But, it has no idea where exactly you want it to look. Now you can just show it. It's very intuitive. Demo: