Grossberg uses the (confusing) terms "visible" and "invisible" to refer to the more commonly accepted terms "modal" and "amodal". His contention that "all boundaries are invisible" therefore means that there is an amodal component to those boundaries, which manifests itself as a modal (or "visible") surface brightness percept when there is a brightness contrast signal across that amodal contour.
I have specifically avoided discussion of the F A C A D E theory. In my view the most interesting and significant aspects of Grossberg's vision model are those which have been adopted here in the MLRF model, i.e. the clear and simple principles of boundary completion and surface brightness filling-in. The F A C A D E theory is very much more complex, and has so many different interacting mechanisms and modules that it is difficult to grasp intuitively exactly what is going on in there. So whether or not Grossberg still believes in the simpler (and more elegant) principles of his original BCS / FCS model, it is that model which I have chosen as a jumping-off point for my own theoretical modeling.