What agents have been missing is a surface for exploration.

AI coding tools have gotten remarkably good at generating UI. Cursor and Claude can produce a polished interface from a single prompt. You describe a component, hit enter, and seconds later you have JSX, Tailwind, even a bit of motion.
There is a quiet limitation. They generate one version at a time. When you’re designing, one version is never enough.
Inside a code-first IDE you’re bound to the current state. You can rewrite the button to be “more playful,” and the model will change the button. You can’t lay three levels of playfulness next to each other and look at them. So you recreate it, linearly, and lose the comparison.
The rest of this is the surface we built so an agent can generate variations side by side, on a canvas, instead of down a chat thread.


No credit card required. Upgrade later when you want the full playbook.
I went through the pricing of every app on a16z's agent-economy map. Here is what they charge for, who pays, and four uncommon beliefs.

A practical guide to the six principles under the prompt box

How to turn individual workflows into a shared workspace, especially inside design teams


From ChatGPT to Figma AI, explore the best AI UX patterns from leading products.