In the Timaeus, Plato assigns each classical element a regular solid: fire is the tetrahedron, air the octahedron, water the icosahedron, earth the cube. These particles fill the universe completely — no void between them. The universe itself is a sphere, the most perfect shape, rotating uniformly.
The geometry is impossible (arXiv:2504.02374). If you fill all of space with polyhedra — any combination of tetrahedra, octahedra, icosahedra, and cubes — the boundary cannot be a sphere. Polyhedra have flat faces, straight edges, sharp vertices. Pack them however you like: the outermost surface will always have bumps and hollows where the polyhedral faces meet the boundary. A smooth sphere requires smoothly curved boundary elements, and no finite packing of polyhedra provides this.
Worse: such a universe cannot rotate. If the boundary has protrusions, rotation would push those protrusions into the void beyond the universe — but there is no void beyond the universe in Plato's cosmology. The universe is everything. There is nowhere for a bump to go. So either the boundary is smooth (impossible given the packing) or the universe doesn't rotate (contradicting the Timaeus explicitly).
This is not a minor inconsistency. It is a collision between two of Plato's deepest commitments: that matter is composed of geometric primitives, and that the cosmos is a perfect sphere. Either the elements are polyhedra and the cosmos is bumpy, or the cosmos is smooth and the elements are not polyhedra. You cannot have both.
What makes this interesting is that it went unnoticed for 2,400 years. Commentators from Aristotle onward criticized Plato's cosmology on physical grounds — why should fire be a tetrahedron? — but not on geometric grounds. The incompatibility between space-filling polyhedra and spherical boundaries is elementary by modern standards. It required no mathematics unavailable to the Greeks. It was hiding in plain sight, unnoticed because the question about the boundary's geometry was never asked.