Extremal code × active mesh × photon falsifiers × rank-48 tensor

Passes 3973–3980. Exact finite calculations are separated from proposed photonic experiments and the still-pending external Monster execution.

global maximum A₄

Exact clique proof on 945 admissible weight-four words.

296

active mixers

Plus 105 signed swaps; 69 layers; exact zero residual.

48 / 904

relations / constants

Literal five-fiber coherent multiplication tensor.

192

extremal stabilizer

Element orders 1,2,3,4,6,12 and geometry 45+6+6.

Global code extremality

The 945 candidates form a degree-624 graph and split into symmetry orbits 135+810. Exact orbit-reduced branch-and-bound proves the maximum clique size is 57.

A₄ ≤ 57, attained by [36,17,4]
W(z)=1+57z⁴+852z⁸+7332z¹²+57294z¹⁶+…+z³⁶

The intersection-two graph is SRG(45,16,8,4) plus two octahedral K₂,₂,₂ components.

Active-device compiler

compiler	active mixers	signed swaps	total	layers
common order	302	96	398	69
independent input/output order	296	105	401	69

The second design lowers genuine mixing hardware by six only when swaps can be absorbed into routing. The exact fixed-order cut-rank lower bound is 253 total adjacent factors.

Six competing photon models

Direct-sum mode capacity with invariant causal front.
Transverse-structure axial group delay.
A falsifiable mode-count-dependent hidden-node alternative.
An engineered W33 graph-kinetic delay spectrometer.
Independent hyperentangled tensor factors.
Self-similarity as coupling/crosstalk control rather than variable c.

The experimental design independently sweeps mode count, transverse momentum, W33 Laplacian sector, and factorization of equal total dimension. The engineered W33 sectors are:

spec(12I−A) = {0,10,16}; declared delay target 0:10:16

Literal coherent tensor

fibers 1|27|36|40|160
48 orbitals
904 nonzero pᵢⱼᵏ out of 110,592
SHA-256 a17a375552c102d281e3ba79b602b67aa54a426fd68b1878293763ea65395ff9

Monster boundary

The abstract 51,840-pair queue and strict harness are present. Portable serialized Monster words and an executed class-fusion artifact are not.

Not claimed: uniqueness of the maximum code, globally optimal mesh size, physically free swaps, hidden photon nodes, variable vacuum light speed, achieved capacity, measured delay, Monster embedding, fabrication, laboratory performance, or remote PDF success.