The hard part about using the browser's layout engine is that it is limiting to the designer. We'd have to snap things into say, floating position, etc. And we'd have to know exactly what the designer was thinking when they're dragging some element around. Or they would have to know about floats, and how to use them.

We think (right now) that the best way around this is some kind of constraint system, like CAD apps have. We're not sure what that looks like yet, though.