We started off building a canvas-based codegen tool for a small set of 50 early users, many of which were designers, who pushed us to build better tools for them.
There wasn’t really a creative tool that let designers easily iterate on their UIs. After seeing what could be done with the newest iterations of image models, we doubled down on solving this problem. I’ve talked with over 100 designers and it’s ever so apparent that a change is needed for their workflows.
I did have previous ideas that we discarded. A year ago from today, I was interested in building agents for backend infrastructure. We switched from that to a canvas, and from there it evolved into the AI-native design tool it is today.
Two questions I keep coming back to: how much further can we go, and what can we achieve in the next 6 months? There will be a point where it feels like we’ve discovered fire for the first time. When will the Cursor moment for design be achieved? It’s clear that we’re getting close. But I think the answer won’t be as obvious as it was for those in the more verifiable domains. It will look different for us. And over the past year we’ve built a strong intuition for how we can achieve success among the design community.