Discriminator audit of the emergent-gravity programme, applied to our own work: the thermodynamic-derivation wing has NO observational discriminator. Non-equilibrium Jacobson yields f(R), whose unscreened form is dead by 2.2×10⁴ in PPN γ (Cassini) and whose screened survivor reproduces ΛCDM for any f_R0 including zero. Logarithmic horizon-entropy corrections are 56–120 orders of magnitude too small at every accessible scale. Every zero-parameter holographic dark-energy member is dead. Ġ/G from G ∝ 1/N_horizon is dead by 293×.

4e+114.

  (d) holographic / entropic dark energy
      predicted deviation: Hubble horizon q0 = +0.50 vs
        LCDM -0.55; event horizon at c=1 gives w0 = -0.891, wa = +0.224
      best bound: USR-2026-0006 Delta AIC = +37.5 (killed); the
        registry's own binned w(z) (USR-2026-0004) misses in every
        bin, naive quadrature 5.1, with wa of the
        WRONG SIGN
      VERDICT: (ii) DEAD for every zero-parameter member; (iii) NO
        DISCRIMINATOR for the parametrized survivors.

  (e) anything else, searched honestly
      a0(z) ~ cH(z): predicts a0(0.85)/a0(0) = 1.61;
        auditable lensed bTFR allows [0.87, 1.15] -> 4.3 sigma_bound:
        DISFAVOURED -- and degenerate with LCDM's own predicted
        apparent rise (x3 by z = 2), and the theory does not even
        fix whether it predicts this branch or a constant
      Gdot/G ~ H0: predicts 6.4e-11/yr vs LLR 2.2e-13/yr -> DEAD by 293x
      GW damping: 1e-20 of a Hubble effect -- no handle
      preferred frame: the program predicts exactly zero
      area quantization: live, ~30-270 Hz for 30 Msun, but it tests
        an INPUT of the derivation, not the derivation
      VERDICT: no NEW alive discriminator beyond (a).

  OVERALL -- and this is uncomfortable, which is why it is stated
  first and plainly:

  THE PROGRAM SPLITS IN TWO, AND ONLY ONE HALF TOUCHES THE SKY.

  * The THERMODYNAMIC-DERIVATION wing -- Jacobson's Clausius
    argument, the area law, the Unruh temperature, the first law, the
    trace-free form, the Verlinde screen algebra: EXACTLY what waves
    1-4 built and verified -- has NO DISCRIMINATOR. It reproduces GR
    by construction, its non-equilibrium extension has a free
    function rather than a prediction, and its quantum corrections
    are 60-120 orders below any bound.  In the registry's vocabulary
    this wing belongs in the same category as USR-2026-0037 and
    USR-2026-0038: FRAMING, NOT EVIDENCE.  Four waves of verified
    machinery amount to a REFORMULATION with no current
    observational handle.  That is the finding.

  * The PHENOMENOLOGICAL wing -- a0 = cH0/6 -- does make one
    parameter-free, falsifiable prediction, and it is ALIVE but
    strained: +2.0 sigma_median high in SPARC (a number smaller than
    the mass-to-light systematic), dead in its unscreened form in
    wide binaries, and its cleanest extension (a0 tracking H(z)) is
    already disfavoured at ~4 sigma by lensed data.  CRUCIALLY: this
    prediction CANNOT BE DERIVED from waves 1-4.  It needs Verlinde's
    extra elastic-medium / dark-energy postulate.  The machinery we
    verified does not imply the one thing that can be tested.

  WHAT WOULD CHANGE THE VERDICT (named, with the number needed):
    1. a0 to 0.02 dex, systematics included.  That means fixing
       Upsilon_disk, which currently swings the answer by 0.205 dex
       -- 2-4x larger than the effect being measured.
    2. Low-g_bar resolved RAR at z > 0.5 with gas-inclusive baryons:
       branch 1 vs branch 2 split by 60-80% at z ~ 1.
    3. Gaia DR4 wide binaries in 3D: the estimator spread of 0.22
       must come down to ~0.05 to test MOND+EFE (1.18) at all.
    4. A derivation of a0 FROM the Clausius/area-law machinery.
       Without it, waves 1-4 and the one testable prediction are
       simply not connected.

==============================================================================
HEADLINES (every number labeled)
==============================================================================
a_registry_reread_claims                : 9 claims read live from registry_seed.json
a_sparc_rerun_a0_median                 : 9.1787e-11 m/s^2 (34 knee galaxies)
a_sparc_sigma_median_dex                : 0.0449 dex
a_offset_cH0/2pi                        : +0.0716 dex = +1.60 sigma_median
a_offset_cH0/6                          : +0.0916 dex = +2.04 sigma_median
a_offset_canonical                      : +0.1164 dex = +2.59 sigma_median
a_offset_FREE(rms)                      : +0.0156 dex = +0.35 sigma_median
a_ML_systematic_swing_accepted_band     : 0.2047 dex = 4.30 sigma_median (Upsilon_d 0.40-0.60)
a_ML_systematic_swing_wide              : 0.2244 dex (Upsilon_d 0.35-0.70)
a_published_a0_systematic               : 0.0792 dex = 1.77 sigma_median
a_H0_tension_span                       : 0.78 sigma_median
a_widebinary_verlinde_literal_boost     : 1.666 (velocity, at g_N = 0.3 a0)
a_widebinary_isolatedMOND_boost         : 1.540
a_widebinary_MOND_EFE_boost             : 1.183
a_widebinary_measured_registry          : 0.948 [0.939-0.958] (USR-2026-0047, killed)
a_widebinary_floor_units_verlinde       : 3.3 (conservative floor 0.219)
a_widebinary_floor_units_MOND_EFE       : 1.1
b_fR_unscreened_PPN_violation           : 2.17e+04 (|gamma-1| = 0.5 vs Cassini 2.3e-5)
b_fR0_screening_ceiling_galaxy          : 2.97e-07
b_R2_coefficient_bound                  : a < 4.167e-10 m^2 (lambda < 50 um)
b_R2_natural_bracket                    : Hubble scale exceeds by 62 orders; Planck below by 60
b_max_nonequilibrium_fraction_cosmo     : 5.73e-61
b_natural_nonequilibrium_fraction       : eps ~ 6 Phi_N = 1.27e-05 (Sun), degenerate with 1PN
c_logcorrection_dFoverF_moon_alpha1     : -1.131e-85
c_logcorrection_dFoverF_lab_alpha1      : -4.708e-60
c_allowed_alpha_tightest                : 2.12e+56 (Eot-Wash lab)
c_alpha_needed_for_MOND_at_10kpc        : 4.25e+114
d_HDE_hubble_horizon_q0                 : +0.500 (LCDM -0.550) -- cannot accelerate
d_HDE_registry_kill                     : killed -- DeltaAIC +37.5
d_HDE_eventhorizon_c1_w0                : -0.8911
d_HDE_eventhorizon_c1_wa                : +0.2237 (registry CPL fit prefers wa = -0.548: WRONG SIGN)
d_HDE_binned_w_naive_quadrature         : 5.1 (3 bins, correlations ignored -- directional only)
e_a0z_branch1_ratio_at_z0.85            : 1.612
e_a0z_lensed_bound                      : [0.87, 1.15] -> 4.3 sigma_bound: DISFAVOURED
e_Gdot_over_G_predicted                 : 6.443e-11 /yr
e_Gdot_over_G_bound_LLR                 : 2.20e-13 /yr -> excluded by 293x
e_GW_damping_rate_100Hz                 : 2.13e-38 /s = 9.4e-21 H0
e_area_quantum_freq_30Msun              : 29.7 Hz (4ln2) .. 269.2 Hz (8pi) -- in the LIGO band, but tests an INPUT
overall_discriminator_verdict           : SPLIT -- the thermodynamic-derivation
    wing (waves 1-4) is category (iii) NO DISCRIMINATOR = framing, not
    evidence, same bin as USR-2026-0037/0038; the phenomenological wing
    (a0 = cH0/6) is category (i) ALIVE but systematics-limited, with its
    literal unscreened form (ii) DEAD in wide binaries.  The two are not
    logically connected: a0 cannot be derived from waves 1-4.

STATUS: audit of known results -- reproduced, not discovered.  No
claim here is a measurement of nature; the only new content is the
systematics ladder in PART A3 and the accounting itself.  Nothing
here solves the cosmological-constant problem.
==============================================================================

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.