Sorry, not brilliant, just wrong. The voronoi algorithm is completely dependent on the input sites. Based on different configurations you can gets things as different as square grids, hexagonal grids, completely irregular organizations, or mixutes of all of them. If your things all tend to look the same then your input data isn't very diverse. Moreover, there are huge difference between 2D, 2.5D, and 3D voronoi so I'm not sure how they could all look the same. Don't blame the tool, blame the user.
the real question is why would you want to make a voronoi (or any algorithm) to begin with? what is it responding to within the discourse?
rather, what do any of these parametric exercises offer an architect beside something that 'looks cool' or 'looks complex' or 'looks irregular'? its not like you can plug programmatic or site or material data into it, it is all geomtric "input." this way of working seems more like pure geometry and little bit of structure. its not lke you are done once you have your honeycomb mesh.
i think what he meant is that everyone who sees a 3dvoronoi-cell-structure recognizes it as a 3d-voronoi-structure.
i did some experiments with david ruttens pointset reconstructiontools,
but, you know, the geometry always looks like 3d voronoi.
of course the topology of the cells is different but they are what they are.
there are a few attempts to use 3dvoronoi for architecture, one famous building is the watercube (Beiijing) and there is some other stuff done by
andrew in the gc workshop (is that right?) or roland snooks from www.kokkuggia.com
i would say i you find a problem which needs 3dvoronoicells, it is good to have them! ;)