Most AI building still happens in a chat box. You type a request, it spits back a single result, you type again, and you scroll up and down a thread trying to remember which version you actually liked. Replit just shipped something different for that: Canvas, a visual workspace where the AI builds alongside you instead of inside a chat window.
Their pitch is the honest part. "The best design work doesn't happen in a chat box. You need space to explore ideas, create variants, and iterate." So instead of one answer at a time, you get a canvas you can spread out on — lay several versions of a landing page side by side, push on the one that's closest, and keep going.
It's built for making websites, apps, and marketing assets, and it folds AI image, video, and audio generation into the same workspace. So the mockup, the hero image, and a short clip to go with it can all come out of one place, rather than three separate tools you stitch together afterward.
If part of your job is mocking up landing pages or marketing visuals, this is worth a look as a faster way to get to a first draft. The promise is seeing a handful of rough directions in the time it usually takes to brief one, then pushing on the version that has legs.
I'd treat it as early. Replit only just launched Canvas, so I haven't put it through real client work yet, and "explore variants" tends to read better in a launch post than it performs on deadline. Still, the canvas idea is the right one for design work. Run your own next mockup through it and you'll learn more in one sitting than any demo will tell you.