Independent recomputation confirms the Van Den Broeck construction IN FULL, correcting C-0013: region IV reproduces to 1% with our closed form, and — under the profile the paper does state (B = alpha(-(n-1)w^n + n w^(n-1)) + 1 with n = 80, E-0032) — the region II sign split also reproduces to 1%: -1.380e30 vs the printed -1.4e30 and +4.865e30 vs +4.9e30 kg, sign boundary w = 0.9812 vs the printed 0.981. C-0013's 'profile-dependence' finding rested on a false premise (the profile was invisible because the canonicaliser strips inline math and, at the time, the extractor missed macro-wrapped equations) and on a C^1 counterexample outside the paper's stated twice-differentiable class. Residual anomaly, unresolved: VDB's printed POINTWISE peak values appear inconsistent with his own stated profile under recomputation, even though his integrated totals reproduce.
What it rests on
Attacks run against it
I recomputed the entire Van Den Broeck region II sign split independently of compute/vdb_pricing.py: I took T^00 from VDB's PRINTED eq. (density) in corpus/S-0010/body.tex (not the repo's symbolic three-Ricci), used exact analytic derivatives of the stated n=80 polynomial B = alpha(-(n-1)w^n + n w^(n-1)) + 1 with w = (R~+D~-r)/D~ (not finite differences), located the sign boundary by bisection at mpmath dps=40 after a 10000-point sign scan (my first secant-based findroot escaped [0,1] to w=1.386 and had to be replaced), and integrated 4*pi*D~*Int (2-w)^2 B^3 T~ dw with adaptive tanh-sinh quadrature split at the root, cross-checked with an 8000-point midpoint rule and an independent c^2/G conversion. Results: E_II,- = -1.3802e30 kg (claim: -1.380e30, ratio 1.00015; VDB print -1.4e30, ratio 0.986), E_II,+ = +4.86546e30 kg (claim: +4.865e30, ratio 1.00009; VDB +4.9e30, ratio 0.993), sign boundary w0 = 0.98119035 (VDB prints 0.981; the CLAIM says 0.9815, which is slightly wrong — T^00 is already positive, +9.0e-34, at w=0.9815, and the repo's own rho flips at 0.9812 on a 1e-5 grid I ran through their finite-difference pipeline). I verified symbolically by two further independent routes (conformal transformation of the 3d Ricci scalar with phi=ln B, and a from-scratch Christoffel computation) that VDB's printed T^00 formula is exactly R^(3)/16pi for the static conformally flat slice — both differences simplified to 0 — validating the K_ij=0 assumption chain. Region IV: independent Ford-Pfenning closed form gave -6.2487e29 kg, ratio 0.9919 vs VDB's -6.3e29, identical to the repo's closed form, confirming 'to 1%'. The residual-anomaly statement also checks out: under the stated profile the true pointwise negative peak is T~00 = -176 at w = 0.576 (10000-point scan), while VDB prints -4.9e2 at w=0.349, where the true density is -1.58e-14 — a ~16-orders mismatch at his printed location, so the printed pointwise values are indeed inconsistent with the stated profile even though integrals reproduce. A3 checks: E-0032's quote matches body.tex line 277 and equations.ndjson index 12 verbatim; git archaeology confirmed the historical premise — at pre-fix commit 6dcf86a fulltext.md had zero occurrences of 'n=80' ('let us choose .') and an alpha-stripped equation, restored by the PDB-26/PDB-30 fixes at 3ecdc26. A4 checks: f identically 1 in region II (region II at r_s in [1e-15, 2e-15] m lies inside r_s < R = 3e-15 m where f=1), the Eulerian observer e_0 equals the primed-coordinate slice normal, and constant v_s is properly flagged as the regime boundary. No attack changed a truth value; the closest hit was the mis-reported boundary digit (0.9815 vs true 0.98119), whose error direction understates the claim's own agreement with the print. || FINDINGS: Independent E_II,- = -1.3802e30 kg and E_II,+ = +4.86546e30 kg match the claim's -1.380e30/+4.865e30 to 0.02% and VDB's printed -1.4e30/+4.9e30 to 1.4%/0.7%. True sign boundary w0 = 0.98119035 (bisection, dps 40); repo's own rho flips at 0.9812; claim's quoted 0.9815 is wrong in the last digit — at 0.9815 the density is already +9.0e-34. VDB's printed T^00 formula proven symbolically identical to R^(3)/16pi via two independent derivations (conformal route and brute Christoffel; both differences = 0). Region IV: -6.2487e29 kg, ratio 0.9919 vs print. Pointwise-peak anomaly confirmed: true peak T~00 = -176/D~^2 at w=0.576 vs printed -4.9e2/D~^2 at w=0.349, where true density is only -1.58e-14. Historical premise verified in git: 'n=80' absent from fulltext.md at 6dcf86a (pre-fix), present after canon 1.1 (3ecdc26); equations.ndjson index 12 = 'B = \alpha(-(n-1) w^n + n w^{n-1}) + 1, \label{B}' matching E-0032 verbatim. Repo integrator at samples=2000 yields -1.380132e30/+4.864045e30, so the claim quotes its own derivation faithfully.