We demonstrate via simulated annealing that an unstructured network of N qubits, subject only to a degree-3 regularity constraint and spectral energy minimisation, sponta- neously partitions into ⌊N/8⌋copies of the Q3 hypercube graph — the unique 3-regular, vertex-transitive graph on 8 vertices supporting a distance-4 error-correcting code realis- able as the face-adjacency graph of a regular octahedron in three dimensions. Over 100 independent trials from random initial conditions with N = 24, perfect Q3 crystallisation occurs in 94% of runs. We prove that Q3 is the unique optimal target by ruling out all competing graphs on independent geometric and coding-theoretic grounds: the Petersen graph fails both the convex polyhedral embedding test and the 4-cycle (distance-4 parity check) requirement. Frustrated configurations (N ̸≡0 (mod 8)) produce high-energy partial clusters that cannot close their parity-check circuits, providing a discrete model of quantum vacuum fluctuations. When inter-cluster bonding is permitted, the isolated octahedra spon- taneously form bridge connections, assembling into a connected lattice network. The code and all simulation data are publicly available for independent reproduction. v2.0 (2026-06-12): an in-PDF dated status/erratum note has been added reflecting the June 2026 canon audit (DRIFT/ANCHOR ledger); see the paper's status note for the specific corrections, supersessions, or upgrades. 2026-06-20 legacy erratum: This version adds a canon erratum note to the legacy paper. Crystallisation remains important but toy/embedded and K33 artefact status changed The body is preserved as a historical derivation trail; the erratum note identifies the current ANCHOR/DRIFT status and superseded claims. 2026-06-21 canon refresh: This version incorporates the 2026-06-21 ANCHOR/DRIFT/PTMS canon refresh and rebuilt local PDF.
David Elliman (Sun,) studied this question.