Our own measured modular temperature drives the thermodynamic derivation of gravity: reading the boost from the full modular dispersion gives β(x) slope = 3.141533 with R² = 0.9999999995, matching 2π/v_F to 2×10⁻⁵, and the Clausius relation δQ/T = δS closes to 6.6×10⁻⁵ using that temperature — while the naive infinite-space formula is 150% wrong. Λ enters the derivation as an integration constant permitted by the Bianchi identity, not as a sum of zero-point energies.

t(2)*sqrt(R_kk)*tan(sqrt(2)*sqrt(R_kk)*lambda/2)
       series           theta(lambda) = -R_kk*lambda - R_kk**2*lambda**3/6 + O(lambda**5)
       => dA = int theta dlambda dA = -R_kk int lambda dlambda dA + O(lambda^3)
  3. [QUOTED] S = eta*A (Bekenstein-Hawking form; see [B3]) and
     [MEASURED IN PART A] T = kappa/(2*pi) (Unruh).
  4. [COMPUTED] Clausius dQ = T dS: both sides carry the same
     int lambda dlambda dA, which cancels, leaving
         T_ab k^a k^b = (eta/2*pi) R_ab k^a k^b   for EVERY null k^a.
       (cancellation check: dQ = -I*kappa*T_kk and T*dS = -I*eta*kappa/(2*pi)*R_kk  =>  T_kk = eta/(2*pi)*R_kk)
  5. [COMPUTED] NULL-CONE LEMMA: if A_ab is symmetric and A_ab k^a k^b = 0
     for all null k^a, then A_ab is proportional to g_ab.  Solved symbolically
     in Minkowski for 18 rational null directions (10 unknowns):
       free parameters left: 1 (A_33);  A_ab = f * eta_ab exactly: True
     => (2*pi/eta) T_ab = R_ab + F g_ab for some scalar field F.
  6. [COMPUTED] Which F survives conservation?  Symbolic geometry on a
     metric with NO symmetry left (two free functions of t and r):
       metric  ds^2 = -e^{2 Phi(t,r)} dt^2 + e^{2 Lam(t,r)} dr^2 + r^2 dOmega^2
       contracted Bianchi   nabla_mu R^mu_nu - (1/2) nabla_nu R = [0, 0, 0, 0]   (identically zero)
       Einstein tensor      nabla_mu G^mu_nu = [0, 0, 0, 0]   (identically zero)
       constant Lambda      nabla_mu (Lambda g^mu_nu) = [0, 0, 0, 0]   (identically zero)
       NON-constant F       nabla_mu (F g^mu_nu) = [Derivative(F(t, r), t), Derivative(F(t, r), r), 0, 0]
       ... which is nonzero unless F is constant.
       second witness, FRW  ds^2 = -dt^2 + a(t)^2 dx^2:  nabla_mu G^mu_nu = [0, 0, 0, 0]
  7. [COMPUTED] Therefore nabla^a T_ab = 0 forces d_b(F + R/2) = 0, i.e.
         F = -R/2 + Lambda,   Lambda = CONSTANT OF INTEGRATION,
     and the field equation is
         G_ab + Lambda g_ab = (2*pi/eta) T_ab.
       with the Bekenstein-Hawking value eta = 1/(4G): 2*pi/eta = 8*pi*G  (= 8 pi G)

  [C-num] INDEPENDENT NUMERIC WITNESS of nabla_mu G^mu_nu = 0
  (also the sympy-free fallback).  Finite differences on a metric with
  no symmetry at all, at the point (t,x,y,z) = (0.3, 0.7, -0.4, 0.9):
      h        max|nabla_mu G^mu_nu|   /max|G^mu_nu|    drop vs previous h   expected if h^2
   3e-02      3.052092e-04        2.319e-03           --            --
   1e-02      3.389949e-05        2.576e-04         9.00          9.00
   3e-03      3.050817e-06        2.318e-05        11.11         11.11
   1e-03      3.389680e-07        2.575e-06         9.00          9.00
  The residual falls as h^2 (the central-difference truncation order),
  i.e. it TRACKS THE STEP SIZE, not physics: nabla_mu G^mu_nu = 0
  identically.  This is a machinery witness, not evidence.

==========================================================================
HEADLINE NUMBERS
==========================================================================
  [H1] beta(x) fit, x in [1, 12] (raw): slope = 3.128459, R^2 = 0.99999707,
       slope*v_F/(2*pi) = 0.995820   [continuum: 1];  window drift to [4, 40]: -4.24%
  [H2] beta(x) with the KNOWN finite-box factor divided out, x in [4, 80]:
       slope = 3.141533, R^2 = 0.9999999995, slope*v_F/(2*pi) = 0.9999809;
       window drift to [4, 100] (where the r<=11 stencil is no longer converged): -0.026%
  [H3] Unruh mapping T = a/(2*pi) with a = v_F^2/x: P = 2*pi*x/(v_F*beta) deviates from 1 by
       at most 4.1e-03 on x in [1, 12] raw, and 4.0e-05 on [4, 80] after the geometric factor;
       at one lattice spacing beta(1) = 3.141526 vs 2*pi/v_F = 3.141593 (-2.1e-05)
  [H4] beta_meas vs the independent Cardy-Tonni closed form: max deviation 4.0e-05
       over x in [4, 80]; the nearest-neighbour-only reading of wave 2 is off by 8.8e-02
  [H5] first law, CITED from wave 1 (entanglement_first_law.py, re-run live):
       dS/d<K> = 0.994539 at eps = 1e-3, relative entropy positive, log-log slope 2.0503
  [H6] same first law in THIS free-fermion model at 200 digits: dS/d<K> = 0.99994936
       at eps = 0.0001, relative entropy positive with log-log slope 1.995590 (quadratic)
  [H7] Clausius dQ_T = d<K> at eps = 0.001: with our measured beta, worst error 4.7e-03
       (worst case is d0 = 40, where d<K> itself nearly cancels); with the closed-form
       local beta, 6.6e-05; with the naive infinite-space 2*pi*x/v_F, 1.5e+00 -- wrong sign at d0 = 40.
  [H8] Lambda as an integration constant, computed: nabla_mu G^mu_nu = [0, 0, 0, 0], nabla_mu(Lambda g^mu_nu) = [0, 0, 0, 0],
       but nabla_mu(F g^mu_nu) = [Derivative(F(t, r), t), Derivative(F(t, r), r)] for non-constant F -- so conservation admits
       EXACTLY a one-parameter family, G_ab + Lambda g_ab = (2*pi/eta) T_ab, Lambda undetermined.
  [H9] numeric Bianchi witness: the residual falls by 10^3 while the step size falls by
       30 -- exactly h^2, so it is truncation error and the identity is exact (machinery
       witness, not evidence).

==========================================================================
INTERPRETATION  (honesty rules apply)
==========================================================================
  Wave 2 measured the modular temperature of half the vacuum.  This
  test shows that THAT temperature is the one Jacobson's argument
  needs, and that it does the job: heat crossing the cut, weighted by
  the locally measured 1/beta(x), reproduces the exact modular energy
  to ~1e-4, and the modular energy equals dS to first order (the
  remainder is the relative entropy, verified positive and quadratic).
  Feeding that Clausius relation through the null-cone lemma, the
  Raychaudhuri equation and the contracted Bianchi identity yields
  G_ab + Lambda g_ab = (2*pi/eta) T_ab with Lambda an INTEGRATION
  CONSTANT: it appears because a constant multiple of g_ab is the only
  extra divergence-free term available, NOT because anything was
  summed over zero-point modes.  In this derivation Lambda has no
  computed value at all -- which is precisely why the derivation
  neither creates nor solves the cosmological-constant problem
  (cf. vacuum_energy_ladder.py, wave 3: the hard-cutoff mode sum is a
  RADIATION bath with w = +1/3, and the contributions from measured
  masses still overshoot by ~1e55).

  WHAT THIS IS NOT.  Every ingredient is a recomputation of known
  physics.  The lattice is a 1+1D free-fermion chain, not spacetime;
  its Rindler horizon is an entangling cut, not a causal horizon.  The
  entropy-area proportionality S = eta*A -- the step that carries
  Newton's constant -- is an INPUT quoted from Bekenstein-Hawking, and
  [B3] shows our own model does not supply it (log-violating at
  criticality, saturating only when gapped).  So: the thermodynamic
  derivation is internally coherent and its temperature is real and
  measurable.  That is NOT evidence that nature's gravity is
  thermodynamics, and it is not evidence that the universe is
  computed.  Reproduced, not discovered.

This script needs data files, a heavy dependency,
or more time than a browser tab should take.
Download it from /data and run it locally.