In the 1600s, Prince Rupert wagered that a cube could pass through a hole cut in an identical cube. He was right. If you prop a cube on its corner at the right angle, you can bore a square tunnel through it large enough for the original cube to pass through. The key is that a cube's diagonal cross-section is wider than its face.
Mathematicians found that this self-passage property — now called Rupert's property — held for every convex polyhedron they checked. Tetrahedra, octahedra, dodecahedra, icosahedra, prisms, all of them. In 2017, the observation was formalized into a conjecture: every convex polyhedron is Rupert. Every 3D shape with flat faces and no indentations can fit through a tunnel bored through an identical copy of itself.
Steininger and Yurkevich (arXiv 2508.18475) proved the conjecture wrong. They constructed a convex polyhedron — 90 vertices, 240 edges, 152 faces — that cannot pass through a hole in itself, in any orientation. They named it the noperthedron. The proof required checking approximately 18 million possible orientations computationally, ruling out every possible angle of passage.
The counterexample needed 90 vertices. Simpler convex shapes — anything with fewer vertices that anyone tested — turned out to be Rupert. The property fails only when the surface has enough facets to close off every possible diagonal cross-section that might provide a wider tunnel. The noperthedron is essentially convex enough to be smooth in every direction, leaving no orientation where the internal diagonal exceeds the external profile.
What interests me is the structure of the surprise. The conjecture had overwhelming empirical support — every natural example of a convex polyhedron is Rupert. The Platonic solids are Rupert. The Archimedean solids are Rupert. The conjecture failed not because anyone found a natural counterexample, but because someone deliberately constructed one at a level of complexity that nature and mathematical tradition hadn't explored. The counterexample required a shape with 90 vertices specifically designed so that no orientation yields a cross-section wide enough for self-passage.
This is a specific instance of a general pattern in mathematics: conjectures that hold for every simple case and fail only at a level of complexity that makes the failure invisible to casual exploration. The problem isn't that mathematicians weren't looking hard enough. It's that the space of convex polyhedra is vast enough that the region where the property fails can hide for centuries — accessible only by deliberate construction, not by surveying familiar shapes. The shape that won't fit was always there. It just wasn't in anyone's collection.