Newton's constant from a jammed packing the area-law coefficient is computed and measured, and it measures the Planck length in monad diameters

Spacetime here is a jammed packing of identical spheres one Planck diameter across - monads - entangled with their neighbours and updated once per Planck time; isotropy forces the sphere, the disorder and the spin-1/2 Heisenberg interaction. Newton's constant is not fundamental but the vacuum's entanglement density, G = 1/(4 eta_A), and counting entanglement across a surface returns the black-hole area law with a computable coefficient. The route on to the field equations is Jacobson's, inherited not re-derived. That makes the central identification - monad diameter is Planck length - testable, not a calibration. At every box size the full Hilbert space can reach it comes out short - by more than this paper long thought. The entropy across a cut is not proportional to the bonds it crosses: it is S = a c + b ln(n_A n_B/N) with b -> 3/2, the logarithm of an ordered ground state with three Goldstone modes, and at every reachable size the logarithm is the larger term. The quantity the anchor needs is a, and measured on a design calibrated against a state whose answer is known exactly it is a = 0.0474+/-0.0013. There are two targets for it and the comparison has to name which it is against, because the anchor identity is a phi_J z = T(3)/C^2 and the two differ by exactly C^2 = 3. Against the target at the derived C = sqrt3, namely pi/(18 phi_J z) = 0.04545, the measurement is 1.50 sigma above on its statistical error - and the systematic is unfavourable to the model, which is stated and quantified below. Against the target that d = l_P would require, 0.13635, the same measurement is short by a factor of 2.5 to 2.9 - and that is the figure to quote only if C = 1, which this paper does not hold and the measurement rejects at 30 sigma. ==> So "short by a factor of three" is the content of the result and not a discrepancy in it: the factor is C^2, and C is derived rather than fitted. And the order in which the two arrived is stated here rather than left to be reconstructed: d = l_P was the expectation this paper was written with, the measurement refuted it, and the reading that puts the factor at C^2 came afterwards - so C = sqrt3 is post hoc with respect to this measurement, and we label it so. ==> What makes it more than a rescue is that it is not free: C is derived twice without using a - from causal covering of the cell's diagonal and, independently, from the magnon speed - it predicts C = sqrt2 in two dimensions, where the run returns 1.488+/-0.609 and is 0.12 sigma from it, and the test that would break it is put on record below before it is run. ==> A post-hoc reading with an a-priori test attached is a hypothesis, which is what this paper claims it to be. That shortfall also puts d/l_P at 0.59 to 0.64, and the black-hole coefficient at k_A = 9.8 to 11.5 under (P1) against the 4 that c = d/tau would give - or at k_A = 3.84-4.01 once (P1) is dropped - the same arithmetic read forward, since the factor of 2.9 is how d/l_P comes out at 0.59 instead of 1. The systematic runs the wrong way and we say so: on the same error the band's lower edge sits 1.50 sigma above 0.04545 and its upper edge 7.7 sigma, so the derived target sits just below the measured band rather than inside it - and the convergence in N is untested either way. That shortfall is not a missing coefficient but a measurement. The propagation speed is what the dynamics return, c = C d/tau, and writing c = d/tau silently sets C = 1; keeping the update rule instead gives d/l_P = 1/C, so the factor above is C = 1.70. Three quantities are then one: d/l_P, the shortfall and k_A are all functions of that single number, k_A as C^-2, so that the area law is recovered at C = sqrt3 rather than missed - while the tick stays at t_P by (P3). And a separate geometric argument identifies the number itself as sqrt3: light crosses the cell's diagonal rather than its diameter. Over 138 packings at N = 12 to 24 the size trend is +0.00082+/-0.00040 per monad, 2.0 sigma - no larger than the scatter. Neither the level nor its convergence is established: the boxes reach only 2.1 to 2.7 diameters, and the packings have not reached bulk. Which way it will go is registered here, before the run, so that the decisive calculation cannot afterwards be read in whichever direction suits the model: the slope above is treated as flat, and the expectation put on record is that a does not converge to 0.13635 but stays near 0.047 - if a rises to 0.136 at a few hundred monads the C reading is wrong, and if it stays near 0.047 the factor C^2 = 3 is permanent and derived. Either way the test is one number and the side being bet on is on the record. So what this paper settles is which quantity the anchor constrains, not its thermodynamic-limit value; reading it as a refutation would be as unearned as reading the number it replaces was as a confirmation. One substance carries the dark sector, split by a theorem not by hand: dark energy is the medium counted globally, rho_L = D(n-n_0), vanishing exactly where the packing sits at contact. A separation pinned at contact would give w=-1 with dot G = 0, and that is the wrong sign. "A slight overshoot" fixes the state, n>n_0, and not the rate: an overshoot relaxes, so n falls, and the model's own dot G/G = 2H(1 + w)/y then returns 1 + w_0>0 (quintessence, DESI's side of -1) and dot G>0, the two running oppositely - dark energy decaying logarithmically while the constant rises - and both signs are the ones ranging, the planetary ephemeris and DESI already carry. The magnitude is not predicted; what is, is the correlation sign(dot G) = sign(1 + w_0), and the price is H_0: the model's value is bounded above by LambdaCDM's own 67.40 and buys motion towards DESI's w_0 by falling. The ending follows and is not a tear: rho_L falls to exactly zero at contact, where the monads touch without overlapping and the bonds - which are the entanglement of overlap - cease with them, so the end of this universe is the erasure of its geometry rather than a rip, a recollapse or a thermal equilibrium. The date is branch-dependent: ln a_c = 7.085 on one branch and 72.82 on the one adopted, unobservable on either. One thing is owed and we mark it rather than settle it: whether the excess density is an overlap or a denser ordering, and on the first reading where the equilibrium sits - the spin sector alone puts it off contact by far more than the cosmology can use. Dark matter is the same medium in local circulation, its shear variance read as mass, tracking no baryons and diluting as a^-3. It is computed, not fitted - two fewer free parameters than any particle account - and direct detection ends it. What the sector owes is a density with its own spatial profile: writing rho_dark ~ rho^p, no single exponent is admissible - p_dm = 0 misses the halo contrast, p_dm = 2 grows the cosmic dark-to-baryon ratio as clustering proceeds, and p_dm = 1 would make that ratio universal when it varies by three and a half decades between the solar neighbourhood and a dwarf spheroidal (where an earlier reading of p_dm = 1 as the target is retracted). A mobile carrier was the candidate way to supply it, and whether that carrier forms is now measured, with the classical answer no: in a box of eight diameters a localised twist comes apart in ten ticks, indistinguishably from a gapless magnon, with under three per cent of its chirality left. If the object exists it is intrinsically quantum, and the question returns behind the exponential wall. Averaging closes by an identity: sigma_2 of a Jacobian is a null Lagrangian, so inhomogeneity does not back-react. Space is finite and closed, its curvature predicted into a two-sided window LambdaCDM leaves free. Neither black holes nor the Big Bang has a singularity, and where the fabric tears, what it carried is destroyed: energy and information are not globally conserved. The area and first laws return the Hawking temperature exactly; and evaporation is not the tear - unmaking a monad at a horizon is forbidden by invariant (5), so no gamma signature follows from it. What replaces it is graph destruction driven by rho_L: the mass lives in the geometry, the repair re-jams into a new graph, and the rate is tau = 2.675 H_0^-1 = 39.5 Gyr with no free parameter - exponential decay rather than M^-2, and nothing emitted. There is no mediating graviton; what radiates is a pattern in the entanglement, and the wave equation that pattern obeys returns general relativity's quadrupole formula with its exact coefficient, unaided. The one load-bearing assumption is supersymmetry below 10^9 GeV, unobserved; the Standard Model is adopted rather than competed with, and matter is a separate theory built on this one - deferred not missing. A companion Letter, fifteen pages, states the central measurement and the argument against it, for a reader who wants the result before the record.