And what about enabled state, display state, or more complicated binding that changes the structure of a DOM node - e.g. converting the icon on a button to fa-cirle-o-notch while an asynchronous operation is occurring?
It might be fine for extremely simple needs, but if you're going with a UI which dynamically reflects your model and application state, you should IMO go all the way. When done right (which usually means as functional, in the programming sense, as possible), a binding setup greatly reduces the amount of thinking you need to do to keep the UI in sync.
(I'm probably missing something and the library is more capable than it looks; on a closer examination, I do see it doing some kind of vector binding, but I don't think it can do e.g. heterogeneous elements, where the HTML per element differs depending on the model.)