The 3D grid result is well known, perhaps it would be fair to call it trivial (you can even throw away all but 2 layers of one of your dimensions). As you say, the Menger Sponge is a subset of the 3D grid, so the students had to find a construction which dodges the holes. To me, the result isn't surprising, but it is pleasing. But the really cool part of the article in my mind is the open problem at the end: can you embed every knot in the Sierpiński tetrahedron?