The future of code involves waiting for Copilots (Even though they get faster all the time, we'll just ask more and more of them). We took a stab at visualizing the design decisions that our Copilot makes while it's coding the design you've prompted it for. Instead of just spinners or the raw stream of text like chat UIs, we overlay this visualization where the design will appear on the canvas. Each visualization reflects a design variant that Copilot works on, and is computed in real-time based on the token streams from the LLM.