The kissing number in n dimensions is the maximum number of non-overlapping unit spheres that can simultaneously touch a central unit sphere. In 2D it's 6 (hexagonal packing). In 3D it's 12 (the subject of a famous Newton-Gregory debate, settled only in 1953). In 8D and 24D, exceptional lattices (E₈ and the Leech lattice) give exact answers. Almost everywhere else, the kissing number is unknown — bounded above and below, with the gap sometimes spanning orders of magnitude.
Dimension 19 is one of these gaps. The previous best lower bound, due to Cohn and Li, was 11,692. Ho (arXiv:2603.10425) improves it to 11,948 — a gain of 256 spheres.
The construction is beautiful. Start with a binary code (a structured subset of {0,1}^n). Nest a smaller code inside it. Take the quotient — the larger code modulo the smaller one — and discover that the resulting structure is the Clebsch graph, a highly symmetric graph on 16 vertices. Extract a maximum independent set from the Clebsch graph (5 vertices). Lift this independent set back into the original code structure, and the resulting configuration of points on the 19-dimensional sphere gives 11,948 kissing vectors.
The gain — exactly 256, which is 2⁸ — is not a coincidence. It reflects the size of the smaller code used in the quotient. The improvement comes from finding additional structure inside an already optimal-seeming configuration. The Clebsch graph was hiding inside the code, and the independent set of the graph was hiding additional spheres that the previous construction missed.
The structural lesson: sphere-packing bounds in high dimensions are not approached continuously — they jump when someone discovers hidden algebraic or combinatorial structure inside existing constructions. The 256 isn't approximate. It's exact, and its exactness is the signature of a structural insight rather than a numerical optimization.