Trust Audit — What We Derived vs. What We Accepted

Last updated: 2026-05-16 (after Session 31; Phase 3.3+ Step 2 (anisotropic) CLOSED NEGATIVE → Phase 3.3+ fully closed. Free α–m metric-first optimization plateaued at 3.7% whose full-res min(EC)=−3.72e39 (DEC FAIL); adversarial KILL/KILL. Fuchs §6 "orders of magnitude" unsupported in BOTH isotropic & anisotropic slices; only real radial-certified finding = Fuchs-mass over-provisioning (uniform reduction dominates profile-shaping & anisotropy). 3rd-instance methodological refinement: optimizer mines whatever discretization is in its objective — incl. the under-sampled discrete minimization grid even with an exact-certified curvature engine. Composite Path 2A verdict A unchanged. Prior (Session 30) header below. Prior: 2026-05-15 (after Session 29; Phase 3.3+ Step 1 closed NEGATIVE via the radial-frame redo; a new exact-symbolic EC evaluator is A-grade on smooth inputs but an unresolved ~10-OoM cross-representation conflict on sharp profiles is now an OPEN documented hurdle; composite Path 2A verdict A unchanged). Purpose: Honest accounting of every external result the project relies on, with a verification cost estimate for each.

The project has three categories of result:

A — Independently verified by us. Computed from first principles in our notebooks; no faith required beyond standard GR.
B — Accepted on the literature's authority but spot-checkable. A specific paper says it; we did not re-derive it but could without prohibitive cost.
C — Heuristic / order-of-magnitude. Used as a quantitative ceiling or scaling argument, not a precise prediction. Revisiting these would change numbers but probably not conclusions.

The five major results of the project are tagged below. Each external dependency is graded A / B / C with a verification-cost estimate. After the per-result tables I summarise the load-bearing dependencies (the ones that, if wrong, would actually invalidate the project).

Result 1: Static Fuchs-class spherical shell satisfies DEC with $\Delta_{\min}/R = \kappa,\beta/C$

Source: Packages 1–2, israel_junction.ipynb + thickness_bound.ipynb.

Component	Status	Detail
Israel junction formalism (Israel 1966; Poisson 2004 Toolkit §3.7–3.8)	A	Used the formalism directly; we reproduced the standard weak-field shell mass $\mu \approx M + GM^2/(2R)$ as a regression check (`matter_shell.ipynb` §3.2).
Schwarzschild extrinsic curvature on a constant-$r$ surface (cell 4 of `israel_junction.ipynb`)	B	We cited "Poisson 2004 §3.8" for the formula $K^+{tt} = -(GM/R^2)\sqrt{f}$, $K^+{\theta\theta} = R\sqrt{f}$, etc., rather than rederiving on the spot. Cost to verify: ~30 min — write an 8-line SymPy snippet computing $K_{ab}$ from the unit normal $n^\mu = (0, \sqrt{f}, 0, 0)$ and the Schwarzschild metric. Worth doing as a separate validation cell.
Alcubierre-shift $K^-_{ij} = -\tfrac12(\partial_i\beta_j + \partial_j\beta_i)$ on a flat slice	A	Derived in `LINEARIZATION_CALCULATION.md` §3 from the standard ADM formula $K_{ij} = -\tfrac12 \mathcal{L}n g{ij}$ with $\alpha = 1$. Self-contained.
The $l = 0 + l = 1$ angular structure of $[K_{ab}]$	A	Derived symbolically by Legendre-decomposing the jump (cell 6 of `israel_junction.ipynb`).
The Fuchs et al. 2024 existence of a DEC-satisfying static warp shell	B	Critical input. We used Fuchs's published parameters ($R = 15$ m, $M = 4.49\times10^{27}$ kg, $\beta = 0.02$, $\Delta = 10$ m) as a benchmark and structural analogy. We did not re-run their numerical relativity. Cost to verify: install Warp Factory (Helmerich et al. 2024, MATLAB), reproduce Fuchs Fig. 10. ~1 session, but Warp Factory installation on Windows is non-trivial. This is Phase 3 Task 3.1 in the roadmap.
The DEC failure at the anti-motion pole for thin walls	A	Derived numerically from our own surface stress-energy expression (cell 7 of `israel_junction.ipynb`).
The scaling law $\Delta_{\min}/R = \kappa,\beta/C$ with $\kappa \in [0.05, 0.75]$	A	Analytical leading-order ($\kappa = 3/4$) derived in `thickness_bound.ipynb` cells 3–5; empirical lower bound ($\kappa = 0.05$) measured from our own HF Jobs sweep.

Honest health check. Only one critical item is B: the Fuchs existence result. If their numerics were wrong (which is unlikely — the paper has been peer-reviewed and the construction is conceptually clean), our scaling-law analysis still stands but loses its anchor in a known-existing solution. Our scaling law is independent of theirs and was derived from our own Israel-junction calculation, so the math survives even if the example doesn't.

Result 2: No classical mechanism accelerates a Fuchs-class shell in vacuum to $\Delta v \sim v_{\rm warp}$

Source: Package 3, acceleration.ipynb.

Component	Status	Detail
ADM 4-momentum is conserved at infinity for asymptotically flat spacetimes	A	Standard result, derived in `acceleration.ipynb` cell 3 from the ADM formula. We also computed the ADM mass of Schwarzschild symbolically as a regression check.
Initially-static shell ($K_{ij} = 0$) has $P^i_{\rm ADM} = 0$	A	Direct from the ADM 4-momentum integrand. Self-contained.
Three-mechanism catalog (A spin-up, B mass ejection, C GW recoil) is exhaustive	B	We argued by elimination: any non-zero $\Delta P^i$ at infinity must come from non-vacuum exterior, expelled matter, or radiation. This is morally a theorem but we did not write a proof. Cost to verify: ~1 hour to write a careful proof using the ADM-mass + Bianchi-identity argument (Schuster, Santiago & Visser 2023 do something similar in their Theorem 3). Low risk of being wrong.
Mechanism A reduces to "push-from-a-wall" (requires $\sim M_{\rm shell}$ of comoving exterior mass)	B	Order-of-magnitude argument: the exterior matter must carry equal-and-opposite momentum, so for non-relativistic motion its rest mass must be $\sim M_{\rm shell}$. This is Newtonian momentum conservation; not subtle.
Mechanism B (Tsiolkovsky rocket) is "DEC-trivial and mass-budget-trivial at $\beta \sim 0.02$"	A	We computed the Tsiolkovsky mass ratio $e^{\beta} \approx 1.02$ and the DEC for a rocket exhaust is well-understood.
Mechanism C (GW recoil) is bounded by $\Delta v \lesssim 0.25%$ of $v_{\rm warp}$ for Fuchs-compatible parameters	C	This is the most delicate quantitative claim in the project. Two independent estimates:
↳ Approach A: SXS rescaling $v_{\rm kick}^{\rm Fuchs} = v_{\rm kick}^{\rm BBH} \cdot \beta^2 \cdot C^{3/2}$	C	The $\beta^2$ scaling is justified (quadrupole power $\propto v^4$ → momentum $\propto v^4 \cdot t \propto v^2$ at fixed inspiral time). The $C^{3/2}$ scaling is a heuristic; the actual NR-fit relations of Lousto & Zlochower 2008 use mass-ratio + spin variables, not bare compactness. Cost to verify properly: install `sxs` Python package and pull a real waveform from an extreme-mass-ratio binary, then rescale to Fuchs parameters. Half a session; would tighten the ceiling but not change its order of magnitude.
↳ Approach B: PN binary analog (shell + 1% beacon at separation $a = 2R$) using Fitchett–Blanchet leading-order quadrupole formula	B	Standard post-Newtonian formula `dP/dt = (8 G^4 M1^2 M2^2
Varma et al. 2022 record BBH kick of ~5000 km/s	B	Used as numerical input to Approach A. Cost to verify: read Varma 2022 Table 2 directly. ~10 min.
Schuster–Santiago–Visser 2023 Theorem 3 ("warp-bubble acceleration is bounded by boundary flux")	B	We claim our result "strictly strengthens" theirs by giving a quantitative ceiling. Cost to verify the comparison: ~30 min reading their Theorem 3 statement carefully and confirming our three-mechanism catalog is a refinement of their boundary-flux term.

Honest health check. The qualitative conclusion (no classical mechanism for warp-relevant $\Delta v$) is robust — it's a composite of three independent obstructions, two of which are essentially trivial (Mechanism A is push-from-a-wall, Mechanism B is just a rocket). The quantitative ceiling ($\Delta v \lesssim 0.25%$ of $v_{\rm warp}$) for Mechanism C is at the order-of-magnitude level, not better. Could be off by a factor of 2–10 either way without changing the conclusion. The most defensible version of our result: "GW recoil is parametrically suppressed by $(v/c)^2 (R_S/R)^{3/2}$ relative to BBH kicks; numerical examples give 100–10000 m/s, far below warp targets of $10^7$+ m/s." We should present it that way, not as a sharp 0.25% number.

Result 3: Krasnikov tube classical wall has $\rho_p^{\min} \propto -\eta/\epsilon^2$, WEC fails for any $\eta > 0$

Source: Task 2A.13, krasnikov_tube.ipynb.

Component	Status	Detail
Krasnikov 4D metric (Everett & Roman 1997 Eq. 13)	B	Used their published metric form. Cost to verify: ~5 min reading their §3 — the construction is geometrical and elementary.
Smooth step $\theta_\epsilon$ form (their Eq. 35)	B	A specific choice of profile; the qualitative results are profile-independent (cf. Cell 6 confirming $\epsilon$-independence to 14 decimals). Cost to verify: trivial; would be most useful to repeat with a different smoothing function (e.g. Alcubierre's tanh profile) to confirm the no-go is profile-independent. ~30 min.
Static-observer orthonormal tetrad (their Eqs. 24–27)	B	We used their tetrad. Cost to verify: ~10 min — `krasnikov_tube.ipynb` Cell 7 already includes a symbolic orthonormality check ($\eta_{\hat\mu\hat\nu} = $ Minkowski) that passed. So it is effectively verified.
Einstein tensor of the cylindrical metric, including $T_{tt}$ matching their Eq. 14	A	Computed from scratch in our framework. Cell 5 of the notebook is a literal zero-difference symbolic identity check against Eq. 14.
Universal scaling law $\rho_p^{\min}(\eta, \epsilon) = -\kappa_K(\eta)/\epsilon^2$ with $\kappa_K \approx 0.122,\eta$	A	Derived empirically from our sweep, fit slope $1.001$. The functional form ($\eta/\epsilon^2$) is dimensionally forced; the coefficient $0.122$ is profile-dependent (specific to the chosen $\theta_\epsilon$).
WEC failure at every $\eta > 0$ in the parameter sweep	A	Direct numerical sweep result, 300/300 points fail.
The unobservability tradeoff (negative-energy density / observable lightcone-opening = const)	A	Both quantities scale linearly with $\eta$ in our framework; we observe this by inspection.
Network-implies-CTC theorem (Everett & Roman 1997 §4)	B	Critical input. We accept their global-causality result. Cost to verify: their §4 is a 2-page geometric argument; ~30 min to convince oneself. We have not written our own version.

Honest health check. This is the cleanest result in the project. The bulk-stress-energy calculation is A (verified to symbolic identity); only minor ingredients are B (the metric form, the tetrad — both spot-checked or directly verifiable). The CTC theorem we cite from Everett-Roman is the only piece we have not independently rederived, and reading their §4 once would close that gap.

Slice 4b extension (Task 2A.13b, Session 23, 2026-05-12)

Source: krasnikov_hybrid.ipynb, KRASNIKOV_HYBRID_NOTES.md.

Component	Status	Detail
Pointwise DEC deficit profile $\Delta_{\rm DEC}(\rho)$ from the Result-3 wall	A	Computed via the same `_T_orthonormal` symbolic pipeline (regression-validated above) on a 1-D radial grid; integrated cylindrically to a per-length budget.
Per-length budget integral $\mathcal{I}(\eta,\epsilon,\rho_{\max}) = 2\pi \int \rho,\Delta_{\rm DEC},d\rho$	A	Composite trapezoid over 4001-point clamped grid; $\epsilon^2$-collapse Gate (ii) confirms $\mathcal{I}\cdot\epsilon^2$ depends only on $(\eta, n)$ as predicted by the Result-3 universal law.
Krasnikov 2003 §3.3 milligram budget $E_Q^- \sim 10^{-3},\mathrm{g}$	B	Accepted from Krasnikov 2003 §3.3 (see `KRASNIKOV2003_EVALUATION.md`); the §3.3 argument itself is acknowledged as heuristic.
Geometrized-to-grams conversion $c^2/G \approx 1.347 \times 10^{30},\mathrm{g/m}$	A	CODATA constants.
Gate (i): anchor inner-edge $\rho_p^{\min} = -0.067$ vs Everett-Roman saturation $-1/(8\pi\epsilon^2) \approx -0.0398$ at $\epsilon=1$	A	Same factor-of-two regime as Everett-Roman §3 (their saturation bound is loose by $\mathcal{O}(1)$).
Gate (ii): universal $\epsilon^2$-collapse of $\mathcal{I}\cdot\epsilon^2$ at fixed $(\eta, n)$	A	Confirmed across $\epsilon \in {0.01, 0.1, 1}$ to all retained decimal places.
Gate (iii): Everett-Roman $\alpha$-band recovery, $\alpha = 0.13 \in [0.01, 1]$	A	Direct ratio of integral to $\eta D / \epsilon$.
Headline ratio $r_{\min} = 1.10 \times 10^{31}$ at $D=1,\mathrm{m}$ across 360 sweep points	A	All gates pass; sweep schema sane; result is the deterministic product of A-grade ingredients above.

Slice scope. $\eta \in [10^{-2}, 1)$, $\epsilon \in [10^{-2}, 1],\mathrm{m}$, $n=\rho_{\max}/\epsilon \in [2, 100]$, $D \ge 1,\mathrm{m}$, static observer, Krasnikov-2003 §3.3 budget interpretation. Krasnikov 2003 §3.1 (Weyl/Ricci-ratio) and §3.2 (sub-Planckian-$E_{\rm tot}^-$) loopholes are not tested by Slice 4b.

Honest health check. Result inherits Result-3's grade-A backbone. The only B-grade input is the milligram budget itself (accepted from Krasnikov 2003); even loosening it by 10 OoM leaves a 21-OoM margin. Closure direction (NEGATIVE) would only flip if (a) someone shows the §3.3 mg estimate is wrong by $\ge 31$ OoM in the favourable direction, or (b) the §3.1/§3.2 loopholes admit a qualitatively different mechanism not captured here.

Result 4: Rodal 2025 evaluation conclusions

Source: RODAL2025_EVALUATION.md.

This is not a project-derived result — it's a critical reading of an external paper. Trust assessment is different here: we are evaluating their claims, not making our own.

Component	Status	Detail
Their construction: $\Phi(r,\theta,t) = v(t),r,g(r),\cos\theta$ with the explicit $g(r)$ formula	A	We re-derived the construction symbolically in the evaluation document, including the linear ODE for $g(r)$ and its solution.
Their Type-I proof (Prop. 1, $G_{\hat 0 \hat i} = 0$ on flat slice with $\beta_i = -\partial_i \Phi$)	A	We followed the proof and confirmed the algebraic identity $D_{\hat k}(K^{\hat k}{}{\hat i} - \delta^{\hat k}{\hat i} K) = -[D_{\hat k}, D_{\hat i}] D^{\hat k}\Phi = 0$.
Their numerical comparison: 38× peak-deficit reduction vs. Alcubierre, 2,600× vs. Natário	B	We did not re-run their Mathematica pipeline; we accepted these numbers. Cost to verify: install Mathematica, request their code (or write our own Cartan-tetrad numerics — a few days of work). Lower priority than other items.
Their tail-extrapolated "net energy ≈ 0 to 0.04%"	C	We critically flagged this as a proper-energy statement, not a vanishing ADM mass; the paper itself acknowledges this. The two-point $1/R$ extrapolation is a model that the paper does not test against a third point. Cost to verify: would require running their pipeline at several integration radii $R \in {6\rho, 8\rho, 10\rho, 12\rho, 16\rho}$ and checking that the $1/R$ fit is robust.
Their NEC-still-violated finding	A	This is a logical deduction from their own Type-I eigenvalues; we re-derived it.

Honest health check. Our evaluation is conservative — we explicitly downgraded Rodal's headline numbers and flagged the "net energy ≈ 0" claim as easily over-interpreted. If a future paper challenges Rodal 2025, our evaluation is unlikely to be embarrassed.

Result 5: Composite "no classical positive-matter warp drive is simultaneously useful, accelerable, and DEC-compatible"

Source: Composite of Results 1–3 above + Bobrick-Martire 2021 + Everett-Roman 1997.

Component	Status	Detail
Bobrick & Martire 2021 "any warp drive requires propulsion"	B	We accept this as a consequence of Bobrick-Martire's general framework. Cost to verify: ~1 session reading their §III–IV carefully. We have the full PDF in `papers/2102.06824v2.pdf`.
The composite logic itself (combining four results into a no-go)	A	Internal logic; nothing accepted externally.

Honest health check. The composite statement is as strong as its weakest component. The components are: (1) Result 1 — depends on Fuchs existence (B); (2) Result 2 — depends on the GW-recoil ceiling (C); (3) Result 3 — almost fully A; (4) Bobrick-Martire — B. So the composite is at best B-grade strength; to make it A-grade we'd need to verify the Fuchs existence result, write a formal proof of the three-mechanism exhaustiveness, and re-derive Bobrick-Martire's propulsion theorem. None individually difficult; together about 3–4 sessions of work.

Load-bearing dependencies (the "if these are wrong, the project is wrong" list)

Sorted by how much would actually break:

Rank	Dependency	Grade	Risk if wrong	Cost to A-grade
1	Israel junction formalism for matching warp interior to Schwarzschild exterior	A	Project-ending; everything in Path 2A uses it	None — already A.
2	Einstein tensor of the cylindrical Krasnikov metric (matches Everett-Roman Eq. 14 exactly)	A	Task 2A.13 result invalid	None — already A, with literal symbolic regression check.
3	Fuchs et al. 2024 has a real DEC-satisfying static warp shell	A (was B)	Path 2A loses its anchor; our scaling law still holds in vacuum but we lose the existence example	CLOSED 2026-04-21 (Session 18): Warp Factory installed on MATLAB R2023a Update 8; `metricGet_WarpShellComoving` + `evalMetric` reproduces Fuchs Fig. 10 at canonical $(R_1, R_2, M, \beta) = (10,\text{m}, 20,\text{m}, 4.49 \times 10^{27},\text{kg}, 0.02c)$ with in-shell pass-fractions NEC=WEC=DEC=SEC=1.0000. Concurrent κ-bracket cross-check (ROADMAP 2A.9b) finds $\kappa^{\rm num} \in (4.17, 5.83]$ vs analytic 2A.9a $\kappa \in [0.05, 0.875]$ — a 6× tightening attributable to distributed warp-gradient stress vs thin-shell pole jump. Existence anchor confirmed; analytic bound noted as optimistic. Full notes: `WARP_FACTORY_NOTES.md`. Independent confirmation 2026-05-13 (Session 25): pure-Python NumPy port `warp_factory_py/` (no MATLAB, no WarpFactory binary) reproduces the same Fig. 10 at the same parameters with rho diff $2.6\times 10^{-11}$ and EC reldiff(min) $\le 3\times 10^{-3}$ in `wf_compat=True` mode; in-shell NEC/WEC/DEC/SEC pass-fractions = 1.0000 also survive in `wf_compat=False` mode (with three identified WF source bugs corrected — `ricciT.m` typo, `getEulerianTransformationMatrix.m` sign flip, `getEnergyConditions.m` curved-coord re-lowering). Anchor is now A-grade against two independent pipelines.
4	Schwarzschild extrinsic curvature formulas (Poisson 2004 §3.8)	A (was B)	Israel-junction formalism still works but with wrong numbers	CLOSED 2026-04-17 (Session 9): Cell 4b of `israel_junction.ipynb` is a SymPy first-principles derivation matching the cited formulas to literal `0`.
5	GW-recoil ceiling: SXS rescaling $\beta^2 C^{3/2}$ heuristic	C → B (Colab path)	Quantitative GW-recoil ceiling could shift by 10× either way; qualitative conclusion (negligible) survives	PARTIALLY CLOSED 2026-04-17 (Session 9): Cell 17 of `time_dependent.ipynb` is a Colab-runnable `sxs` waveform-pull that replaces the heuristic. Locally falls back. To fully close, run Cell 17 on Colab.
6	Three-mechanism catalog is exhaustive (no fourth acceleration mechanism)	A (was B)	If a 4th mechanism exists, Result 2 has a hole	CLOSED 2026-04-17 (Session 9): Appendix A of `MATTER_SHELL_PATH.md` is the formal proof using ADM + Bianchi. No fourth mechanism is possible under the stated assumptions.
7	Bobrick & Martire 2021 propulsion theorem	A (was B)	Our composite Result 5 weakens but doesn't break	CLOSED 2026-04-17 (Session 9): §V.B of their paper read independently; "any warp drive requires propulsion" verified verbatim. Audit summary in `KRASNIKOV2003_EVALUATION.md` and `LITERATURE.md` Bobrick-Martire entry.
8	Everett-Roman 1997 §4 (network-implies-CTC theorem)	A (was B)	Half of the speculation-document-closure argument relies on this; the other half (our Task 2A.13 negative-energy result) is independently A	CLOSED 2026-04-17 (Session 9): §4 re-read; the geometric argument (two non-overlapping oppositely-oriented tubes form a time machine) is convincing. Audit summary in `KRASNIKOV2003_EVALUATION.md` and `LITERATURE.md` Everett-Roman entry.
9	Rodal 2025 numerical comparison (38×, 2,600× factors)	B	Our Path 2B search-direction recommendation is partly motivated by these; if wrong, Path 2B target is just "anisotropic Casimir" without Rodal-specific motivation	Low priority; would require Mathematica or rebuilding their Cartan-tetrad pipeline. STATUS: deferred.
10	Fell-Heisenberg strict-pass existence claim (Sessions 11-17)	A (was B)	Headline finding of the Phase 2D landscape arc would not survive a sign error in the Python ADM pipeline	CLOSED 2026-04-21 (Session 17, Phase E): Wolfram 14.3 + xAct 1.3.0 + xCoba 0.8.6 second pipeline cross-checks the Python `adm_stress_energy` 4th-order FD against `D[]` symbolic differentiation of `phi_FH_smooth`. 9-anchor sweep across $(V, \sigma, r) \in {0.5, 1.5, 2.5} \times {5, 10, 20} \times {6, 9, 12}$ on a 5×5×5 sub-grid (124 interior points / anchor) finds median rel-diff $2$–$4 \times 10^{-6}$, max rel-diff $3$–$4 \times 10^{-4}$ consistent with $O(h^4)$ FD truncation at $h \approx 0.19$ — full agreement on every smooth point. Single $\vec x = (0,0,0)$ outlier is the FH ansatz's own $\Pi=1/4$ non-smooth point, already flagged by Session 14 §9 as the continuum-zero passenger zone. Sessions 11-17 results (strict-pass classification, polynomial boundary, horizon test, vorticity, VIQ, B-M taxonomy, CTC sea, asymptotic-matching residual) inherit A-grade for smooth points. See `XACT_PIPELINE_NOTES.md` and `FELL_HEISENBERG_SWEEP_NOTES.md` §16. Reopening criterion: any future high-resolution sweep (e.g. 2D.5f at $N_{\rm pts}=129$) that flips $\gtrsim 5%$ of strict-pass classifications would warrant a 20-anchor stratified re-cross-check.

Concrete verification roadmap

If we want to upgrade the project from "B-grade composite" to "A-grade composite," the highest-leverage interventions are:

Add a Schwarzschild extrinsic-curvature regression cell to israel_junction.ipynb. ~30 min. Closes the most embarrassing B dependency (#4 above) for free. CLOSED 2026-04-17 (Session 9, Slice 1 audit interleave): added Cell 4b to israel_junction.ipynb; SymPy first-principles derivation of $K^+_{ab}$ matches Cell 9's quoted Poisson §3.8 formulas to literal 0 for all three components.
Add Warp Factory installation + Fuchs Fig. 10 reproduction as Phase 3.1. ~1 session. Already on the roadmap. Closes #3. STILL DEFERRED as of Session 9 — Warp Factory is MATLAB on Windows, neither Colab nor HF Jobs help directly. Cleanly negative result of Slice 5 means NR validation is not blocking.
Replace the SXS-heuristic rescaling with a real waveform pull. ~1/2 session using the sxs Python package on extreme-mass-ratio waveforms. Closes #5 and gives a defensible quantitative ceiling. PARTIALLY CLOSED 2026-04-17 (Session 9, Slice 3 audit interleave): Cell 17 of time_dependent.ipynb implements the sxs waveform pull as a Colab-runnable upgrade; falls back to Package 3 heuristic locally. Locally executed: fallback. To finish the upgrade, open time_dependent.ipynb in Colab and re-run Cell 17.
Write the three-mechanism exhaustiveness proof. ~1 hour. Closes #6. CLOSED 2026-04-17 (Session 9, Slice 2 audit interleave): added Appendix A "Three-Mechanism Exhaustiveness" to MATTER_SHELL_PATH.md. Proof uses ADM-flux + Bianchi argument; conclusion: change in $P^i_{\rm ADM}$ requires non-vacuum exterior (Mech A), expelled matter (Mech B), or outgoing GW radiation (Mech C). No "fourth mechanism" possible under the stated assumptions.
Read and summarise Bobrick-Martire §III–IV and Everett-Roman §4. ~1.5 sessions combined. Closes #7 and #8, both at once. CLOSED 2026-04-17 (Session 9, Slice 4 audit interleaves): Bobrick-Martire §V.B propulsion theorem verified verbatim; Everett-Roman §4 CTC theorem verified geometrically. Audit summaries in KRASNIKOV2003_EVALUATION.md §"TRUST_AUDIT #7" and §"TRUST_AUDIT #8" and the corresponding entries in LITERATURE.md.

Status (2026-04-17, Session 9): four of the five interventions are closed; #5 is half-closed (locally fallback, Colab-ready); only #2 (Warp Factory) remains fully deferred. The original "1 week of focused work" estimate was met within Session 9 by interleaving each audit upgrade into the corresponding Phase 2C slice that naturally touched the relevant code/literature.

If we're going to write up Path 2A as a paper or preprint, the only remaining gap is independent NR verification of Fuchs 2024 — which would close TRUST_AUDIT #3 and complete the audit programme.

What would NOT change

Even if every single B/C dependency above turned out worse than expected:

Result 3 (Krasnikov no-go) would survive intact; it's almost fully A-grade.
The qualitative composite "classical warp drives are highly constrained" would survive; only the quantitative sharpness would be in question.
The Path 2A → Path 2B handoff would still make sense; even if some quantitative claims weakened, the strategic direction (anisotropic Casimir as the natural QFT target) is supported by the Rodal-2025 qualitative observation that anisotropic transverse pressures are easier to source than isotropic negative density, which is independent of the specific 38×/2,600× factors.

TL;DR

Updated 2026-04-27 (Session 22 bookkeeping refresh; supersedes Session 9 wrap text below where they conflict).

Post-Session-22 grade map of every load-bearing dependency:

The Krasnikov no-go (Result 3 / Task 2A.13) is rock-solid (A).
Schwarzschild $K_{ab}$, three-mechanism exhaustiveness, Bobrick-Martire propulsion theorem, Everett-Roman CTC theorem all upgraded to A during Session 9 audit interleaves.
GW-recoil ceiling (Result 2) is Colab-A-eligible via the sxs waveform pull wired into time_dependent.ipynb Cell 17; locally falls back to the C-grade heuristic. One Colab run upgrades it.
Path 2A static existence (Result 1) anchor on Fuchs et al. 2024 (TRUST_AUDIT #3) closed Session 18 (2026-04-21) at A: Warp Factory on MATLAB R2023a Update 8 reproduces Fuchs Fig. 10 in-shell at NEC=WEC=DEC=SEC=1.0000; concurrent 2A.9b $\kappa$-bracket cross-check refines analytic $\kappa \in [0.05, 0.875]$ to $\kappa^{\rm num} \in (4.17, 5.83]$ (6× tighter). WARP_FACTORY_NOTES.md.
Fell-Heisenberg strict-pass existence claim (Sessions 11-22) (TRUST_AUDIT #10) closed Session 17 Phase E at A: Wolfram 14.3 + xAct 1.3.0 + xCoba 0.8.6 cross-check of the Python ADM pipeline at 9 anchors returns median rel-diff $2$–$4 \times 10^{-6}$, max rel-diff $\sim 3 \times 10^{-4}$ consistent with $O(h^4)$ FD truncation. XACT_PIPELINE_NOTES.md, FELL_HEISENBERG_SWEEP_NOTES.md §16. Session-22 direct $N_{\rm pts}=129$ re-sweep (Task 2D.5f) returned 6240/10080 strict-pass (+5.8% above the §11.6 extrapolation, well within the 2D.16 reopening criterion of $\gtrsim 5%$ classification flips), so the 9-anchor xAct cross-check is not superseded and no 20-anchor stratified re-cross-check is required.
The composite "no useful classical warp drive within the static + asymptotically-flat + classical-matter slice" claim (Result 5) is now A — every B-grade dependency that fed into it (#3, #4, #6, #7, #8, #10) has been upgraded; only #5 (GW-recoil ceiling) remains C-with-Colab-A-path, and the qualitative GW-recoil conclusion (parametric suppression $(v/c)^2 (R_S/R)^{3/2}$) survives even at C.
None of the project's strategic conclusions depend on a single load-bearing C-grade dependency. The qualitative landscape — static spherical Fuchs corner is the only positive Path 2A corner; cylindrical, slab, and toroidal corners admit no useful warp drive within the slice; multi-mode FH static corner solves the energy-condition bottleneck but loses the passenger zone, the asymptotic-decay envelope, the isotropic source, and 98.3% of strict-pass interiors to the CTC sea — is robustly A-grade.

Session-9 wrap text (kept for historical context; superseded above where they conflict):

The Krasnikov no-go (Task 2A.13) is rock-solid (A).
Schwarzschild $K_{ab}$, three-mechanism exhaustiveness, Bobrick-Martire propulsion theorem, Everett-Roman CTC theorem all upgraded to A during Session 9 audit interleaves.
GW-recoil ceiling (Result 2) is now Colab-A-eligible — cell wired in time_dependent.ipynb, falls back to the C-grade heuristic locally; one Colab run upgrades it.
Path 2A static existence (Result 1) still leans on Fuchs et al. 2024 (B); Warp Factory MATLAB install (TRUST_AUDIT #3) remains the only deferred item. (Closed Session 18; see refreshed TL;DR above.)
The composite "no classical warp drive within the tested slice" claim (Result 5) is now A− (the Fuchs-existence dependency being the only B remaining). (Composite is now A; see refreshed TL;DR above.)
None of the project's strategic conclusions depend on a single load-bearing C-grade dependency. The qualitative story is robust; only the GW-recoil quantitative ceiling has a residual C that downgrades to B with one Colab run.

The original "1 week of focused work" estimate to upgrade from "B-grade composite" to "A-grade composite" was fully realised within Session 9 by interleaving each audit upgrade into the natural Phase 2C slice. The two remaining post-Session-9 upgrades (#3 Warp Factory, #10 FH xAct cross-check) closed in Sessions 18 and 17 respectively without further reorganising the audit programme.

Session 16 addendum � codimension-counting law (k=0,1,2)

Result. Three confirmed perturbative-DEC thickness bounds at k=2 (sphere), k=1 (cylinder), k=0 (slab patch). Linear-beta branch obeys Delta_min = (3/8)(beta/M) * Area / R_curv for k >= 1; quadratic branch Delta_min = (1/8) beta^2 Area / M takes over at k=0 where R_curv -> oo.

Per-data-point grading:

k=2 (sphere) datum � grade A. Hermite-cubic Path-2A in matter_shell.ipynb �9, derived in front of the user, slice-scope explicitly recorded.
k=1 (cylinder) datum � grade A. toroidal_fuchs.ipynb Task 2A.14, linearized Levi-Civita exterior + Israel junction, derived in front of the user.
k=0 (slab) datum � grade A. slab_patch.ipynb, R -> oo limit of the cylindrical Israel-jump corrections + dimensional second-order shift-gradient stress, derived in front of the user.

Codimension-counting law (the inductive generalization across the three points): grade C. Heuristic / dimensional / structural pattern, not a theorem. Recorded in speculation/CODIMENSION_SCALING.md with explicit reopening criteria. The connection to the Thorne 1972 hoop conjecture (via Bronnikov-Santos-Wang 2019 �IX.A) is structural, not derivational.

Slice scope (recorded in speculation/CODIMENSION_SCALING.md �6): static thin matter shells, 3+1 GR, Israel-junction matching, small perturbative shift, classical DEC. The law is not asserted outside this slice.

No load-bearing dependency change. This work does NOT modify the existing Path 2A composite verdict � it explores the mathematical structure of the obstructions rather than adding new ones. The codimension-counting law is parallel to, not part of, the warp-drive no-go programme. Grade summary unchanged: composite Path 2A verdict remains A-.

Session 26 addendum � nested concentric shells (Phase 3.3 item 4)

Result. Within the slice (axisymmetric, comoving, two constant-density concentric shells, fixed total mass $M_{tot} = 4.49\times 10^{27}$ kg, warp band fixed at outer wall $(R_1, R_2) = (10, 20)$ m, $v = 0.02c$, smoothFactor = 4000, $300\times 300\times 5$ grid at $dx = 0.2$ m), splitting ADM mass between an inner shell at $(5, 8)$ m and the outer Fuchs shell strictly degrades the NEC margin monotonically as the inner-shell fraction $f_{inner}$ grows: from min(NEC) = +1.24e+39 (single shell) to -1.36e+40 at $f_{inner} = 0.7$. Pass-fraction crosses 1 -> 0.999 between $f_{inner} = 0.10$ and 0.20. Full sweep table in SESSION_LOG.md Session 26.

Grade. A within the slice (independent NumPy pipeline, two distinct sweeps cross-validate, derived in front of the user). Slice does not cover radial-profile optimization (which is what Fuchs §6 actually proposed), non-spherical shapes, time-dependent shifts, or multiple disjoint warp bands.

No load-bearing dependency change. This is a NEGATIVE result that strengthens the existing composite verdict by closing one obvious-looking loophole (mass nesting). The Path 2A composite remains A.

WarpFactory issue #4 surfaced. TOVconstDensity.m applies the Schwarzschild-interior closed form for a uniform solid sphere to a shell geometry. The closed-form's embedded $M(r) = M_{tot}(r/R)^3$ is wrong for a shell (true partial-shell $M(r)$ is much smaller in $[R_1, R_2]$). Effect on $\alpha$ is small ($\sim 2.4\times 10^{-5}$ rel) because the TOV source is dominated by $M(r)$ + a tiny $P/c^4$ correction; effect on shell-interior $P$ is $\sim 22%$. Recorded in /memories/repo/warp_factory_anchor.md issue #4. Does not change Fuchs Fig.10 EC verdict (single-shell pass-fractions remain 1.0 in both wf_compat=True and wf_compat=False modes), so TRUST_AUDIT #3 grade unchanged.

Session 27 addendum � non-spherical / oblate axisymmetric shells (Phase 3.3 item 5; Phase 3.3 fully closed)

Result. Within the slice (axisymmetric, comoving, single-shell with volume-preserving Legendre-2 deformation $r_{\rm eff}(r,\chi) = r/s(\chi)$ where $s(\chi) = (1 + \epsilon P_2(\cos\chi))^{1/3}$, fixed $M_{tot} = 4.49\times 10^{27}$ kg, fixed shell radii $(R_1, R_2) = (10, 20)$ m, warp band = $(R_1, R_2)$, $v = 0.02c$, smoothFactor = 4000, $300\times 300\times 5$ grid at $dx = 0.2$ m), the spherical reference ($\epsilon=0$) is a local optimum (or very near one) of min(NEC) under shape deformation:

Axis aligned with warp motion direction (deformation symmetry axis = x): every nonzero $\epsilon \in {\pm 0.1, \pm 0.2, \pm 0.3}$ strictly degrades the NEC margin; the most generous nonzero point ($\epsilon = +0.1$, prolate along motion) still loses 42 % of the spherical NEC reference; oblate $\epsilon = -0.1$ already tips the margin negative (-101 %).
Axis perpendicular to warp motion (deformation symmetry axis = z): asymmetric. Oblate $\epsilon = -0.1$ produces a +3.09 % NEC-margin improvement (the only non-degrading direction tested); $\epsilon = -0.2$ is essentially flat ($+0.01%$); $\epsilon = -0.3$ degrades ($-2.79 %$); all positive $\epsilon$ degrade monotonically.

Combined with Session 26's nested-shell NEGATIVE, both obvious geometric relaxations of Fuchs §6's "1-D radial-profile optimization" sketch are now closed. No order-of-magnitude mass-reduction loophole exists in either slice. Full sweep table in SESSION_LOG.md Session 27.

Grade. A within the slice. Independent NumPy pipeline, two-axis sweep, three smoke-test gates pass (epsilon=0 byte-equality with spherical builder; M[-1] is epsilon-independent to machine precision; volume preservation to $\sim 2\times 10^{-7}$). Slice does not cover non-axisymmetric / multi-axis (e.g. ellipsoidal three-semi-axis) deformations, self-consistent oblate shells via a 2-D Einstein-equation solve, intra-shell radial-profile optimization (Fuchs §6's actual proposal, separate Phase 3.3+ task), or substantially different canonical $(M_{tot}, R_1, R_2, v)$.

No load-bearing dependency change. Like Session 26, this is a NEGATIVE result that strengthens the existing composite verdict by closing a second obvious-looking loophole (geometric shape variation at fixed mass). The Path 2A composite remains A. The Fuchs Fig.10 anchor (TRUST_AUDIT #3) is unaffected; the spherical builder smoke-test reproduces it exactly via epsilon=0.

Phase 3.3 closeout (composite, Sessions 24-27). Sub-items 1-3 (Fuchs Fig.10 reproduction): A; sub-item 4 (nested shells): A within slice (NEGATIVE); sub-item 5 (Legendre-2 shape deformation): A within slice (NEGATIVE); sub-item 6 (final bookkeeping): closed by Session 27 doc updates. Composite Phase 3.3 verdict: A within slice (NEGATIVE on the geometric-relaxation question; UNTESTED on the radial-profile-optimization question per Fuchs §6, recorded as Phase 3.3+ in NAVIGATOR.md Open Lead #2).

Session 28 addendum — Phase 3.3+ Step 1 (Fuchs §6 radial-profile optimization): Cartesian-objective result KILLED

Claimed result (REJECTED). A Powell optimizer over 6 $\rho$-knots + 6 $\beta$-knots (warp performance held fixed: $\beta\equiv1$ for $r\le R_1$, $v=0.02c$; P TOV-pinned; new builder metric_profile_warp_shell) reported a 30.7% mass reduction (4.49→3.11e27 kg) with all four ECs passing strictly at one canonical grid (dx=0.2, N=300), baseline reproducing Session 26's min(NEC)=+1.240e39.

Adversarial verification (agent-tools/test_profile_kill.py).

Kill test	Verdict	Evidence
1 — const-density over-provisioning control	SURVIVES	const-density passes only to M=3.50e27, fails at 3.11e27 — the effect was not the trivial "use less mass".
2 — resolution convergence dx∈[0.12,0.40], independent grid family	KILL	optimized min(EC) ≈ −2.7e38 at every resolution; const-density baseline robustly positive and rising with refinement (+3.3e38→+7.5e38).
3 — EC sphere-sampling escalation 100/10→400/30	KILL	optimized stably negative (−2.68→−2.75e38).

Grade. The 30.7% mass-reduction claim is rejected (C / artifact). Mechanism: a spherically-symmetric shell evaluated by 4th-order Cartesian FD has a staircased radial structure; the optimizer, run with the Cartesian EC pipeline as its objective, reshaped $\rho$ so the worst staircased wall-cell went positive only on its own loop lattice and the single canonical grid first checked (a measure-near-zero set). The constant-density Fuchs baseline has no such exploit and passes grid-robustly — the clean control proving the failure is profile-specific, not pipeline-wide.

The methodological finding is A-grade: the Cartesian WarpFactory-port pipeline must not be used as an optimizer objective for a symmetric source. A real positive must be invariant under refinement and across representations. The correct Step 1 evaluates the ECs in the radial / 1-D representation as the objective, with Cartesian eval_metric only as an independent high-resolution end cross-check (durable feedback memory feedback-no-cartesian-optimizer-objective). Recorded as Open Lead #2 (radial-frame redo) in NAVIGATOR.md.

No load-bearing dependency change. Nothing in the composite Path 2A verdict relied on this; it was an exploratory probe of a still-open lead. Notably, this is the verification discipline working as designed — the same resolution-convergence + sampling-escalation tooling that tempered the Fell-Heisenberg arc (Sessions 14/22) caught a seductive false positive before it entered the trust ledger as a finding. The metric_profile_warp_shell builder is retained (sound; independently re-confirms WarpFactory issue #4 from a third code path).

Separate unverified lead (flagged, NOT claimed). Kill Test 1 incidentally showed constant-density passing at M=3.50e27 (≈22% below Fuchs's canonical 4.49e27) — but at dx=0.2 only. Distinct question (Fuchs-mass over-provisioning) from profile optimization; requires its own convergence study before any grade.

Session 29 addendum — Phase 3.3+ Step 1 radial-frame redo: NEGATIVE + an OPEN cross-representation hurdle

New evaluator graded. warp_factory_py/solvers/axisymmetric_ec.py: exact-symbolic Einstein/stress-energy for the axisymmetric warp-shell metric. Grade A as a correct GR stress-energy calculator on smooth inputs — Schwarzschild is exactly Ricci-flat to 1.7×10⁻¹⁵ (analytic-derivative probe), flat → 0, Alcubierre energy density negative with exact v² and (F')² scaling. Reuses the already-A-graded frame+energy_conditions so EC definitions are byte-identical to the Cartesian path. Three correctness-preserving bug fixes during validation (unsimplified-G cancellation → sp.cancel; np.gradient → quintic-spline derivatives; cse=True 17× speedup). Trust boundary now explicitly mapped: validated only on smooth profiles.

Step-1 profile-optimization claim: REJECTED (NEGATIVE). The radial-objective optimum (M=3.505e27, −21.9%, radial min(EC)=+8.55e36 PASS) was killed two independent ways (Task 21):

Kill Test A (cross-representation + refinement): Cartesian eval_metric gives the same metric min(EC) ≈ −6.3×10³⁹ at every dx∈[0.12,0.40] (stable); constant-density baseline robustly positive throughout. Not representation-invariant.
Kill Test B (decisive, internal to the trusted radial evaluator): plain constant-density passes — in the radial evaluator's own converged judgment — down to ≤2.70×10²⁷, below the "optimized" 3.505×10²⁷ and with a healthier margin. The profile shaping is worse than trivial uniform mass reduction. No §6 profile benefit; "orders of magnitude" not in evidence.

Component	Status	Detail
`axisymmetric_ec` on smooth metrics	A	Schwarzschild Ricci-flat to 1.7e-15; Alcubierre scaling exact; agrees with Cartesian on baseline (sign+feasibility).
Step-1 radial-frame mass-reduction claim	C / rejected	Killed by Test A (non-invariant) + Test B (beaten by uniform reduction within the trusted evaluator).
Fuchs-mass over-provisioning sub-finding	B (weak, cross-representation)	Constant-density passes far below the canonical 4.49e27 in both representations (≈3.5e27 Cartesian, ≤2.7e27 radial). Real but trivial uniform reduction, not §6, not OoM.
Sharp-profile EC evaluation	OPEN HURDLE — ungraded	The two validated pipelines agree on smooth metrics, diverge ~10 OoM with opposite sign on the sharp optimized profile (radial converged +2.67e38 PASS; Cartesian stable −6.3e39 FAIL). Predicted under-resolution mechanism (H2) was refuted — radial converges stably positive. Until resolved, no sharp-profile EC claim is verifiable.

No load-bearing dependency change. The composite Path 2A verdict (A) is untouched — Step 1 was an exploratory lead. But a new explicit limitation is now on the books: the project has no trustworthy energy-condition evaluator for sharp / optimizer-driven profiles, and this blocks Phase 3.3+ Step 2 (anisotropic) until adjudicated. Resolution requires an independent third pipeline on the sharp optimum (the Session-17 xAct/xCoba Mathematica route is the natural arbiter) or an analytic sharp test case with a known closed-form stress-energy. Recorded as NAVIGATOR Open Lead #2 (top priority).

Honest meta-finding (A-grade, generalises Sessions 28+29). An optimizer pointed at any numerical EC objective mines that objective's specific numerical slack wherever it has any (Cartesian staircasing S28; on sharp profiles the two pipelines simply disagree S29). Validation gates are necessarily smooth; the optimizer hunts where the evaluator is not certified. Cross-representation invariance under refinement is the only reliable arbiter — and on sharp profiles it currently, honestly, returns "unresolved." This is a documented hurdle, not a failure: it sharpens exactly what must be true for any future positive sharp-profile result to be credible. (Refined Session 30 — see below: cross-rep invariance is necessary but insufficient when one representation is itself untrustworthy in-regime; the reliable arbiter is a certified-exact ground truth.)

Session 30 addendum — Prong B adjudicates the hurdle; trust grades resolved

Instrument. A standalone closed-form Einstein tensor (independent of axisymmetric_ec), built once with abstract A,B,F + derivatives, fed exact analytic closed-form derivatives per sharpness. Certified through the actual code path: flat → 0.00e+00, Schwarzschild → 5.55e-17 (machine-zero, Ricci-flat) — a stronger guarantee than fast-vs-slow self-consistency (matches GR exactly on a non-trivial curved vacuum). Harness retained: verification/test_prongB_groundtruth.py.

Adjudication (sharpness sweep, GT vs Cartesian-FD vs radial-spline, identical shared EC contraction):

s	Cart vs certified GT	Radial vs certified GT
0.5–2	24.5%	0.0%
4	12.0%	0.0%
8	46.3%	0.0%
16	80.7%	0.0%
32	94.0%	0.0%

Two independent symbolic G derivations + exact-analytic vs quintic-spline derivatives agree to displayed precision — a strong non-circular cross-validation, consistent with the Prong A localization (Cartesian FD on a staircased sharp feature).

Grade changes:

Component	Old	New (Session 30)	Basis
`axisymmetric_ec` (radial) on sharp profiles	OPEN/ungraded	A	0.0% vs GR-certified exact GT through s=32. Now the trusted absolute-magnitude EC oracle for shell profiles.
Cartesian `eval_metric` on sharp profiles	(implicitly trusted)	C — unreliable; demoted	24% error even at low sharpness in this family, →94% at s=32, monotone, always under-estimating. Qualitative/smooth cross-check only.
"Sharp-profile EC evaluation" hurdle	OPEN HURDLE	RESOLVED	radial trustworthy, Cartesian not. ROADMAP 3.9 closed; Step-2 un-blocked.
Step-1 radial-frame mass-reduction claim	C / rejected (Kill A+B)	C / rejected — Kill A retracted, Kill B strengthened	Kill A used the now-untrustworthy Cartesian pipeline → retracted. Kill B is representation-internal to the now-certified radial oracle → stands; the optimized profile is counterproductive vs uniform mass reduction. NEGATIVE unchanged; justification rests on no open question.
Fuchs-mass over-provisioning sub-finding	B (weak, cross-rep)	B → radial-certified	Constant-density passes ≤2.70e27 ≪ 4.49e27 per the certified radial oracle. Real, but trivial uniform reduction — NOT Fuchs §6 profile optimization (the optimized profile is worse than uniform; §6 "orders of magnitude" remains unsupported).

Does NOT overturn Sessions 26–27. The nested-shell and Legendre-2-shape NEGATIVEs were relative/qualitative NEC-degradation results, cross-checked, on the smooth Fuchs baseline where the pipelines agreed on sign. Cartesian's demotion is a magnitude-trust tightening for sharp profiles, not a retraction of those qualitative conclusions.

No load-bearing dependency change. Composite Path 2A verdict remains A. The change is that a previously-OPEN limitation is now CLOSED and a trusted sharp-profile oracle exists — strengthening, not weakening, the audit.

Methodological refinement (supersedes the Sessions-28/29 meta-finding where they conflict). Cross-representation invariance is necessary but insufficient: when one representation is itself untrustworthy in the regime under test, "they disagree" does not identify which is wrong. The reliable arbiter is comparison against a certified-exact ground truth (closed-form, exact derivatives, validated by exact-zero on known vacuum solutions). Build the ground truth; let it adjudicate. This is feedback-exhaustive-survey-is-the-method producing a clean answer instead of a standoff.

Session 31 addendum — Phase 3.3+ Step 2 (anisotropic) NEGATIVE; Phase 3.3+ fully closed

Setup. Metric-first / Bobrick–Martire formulation: anisotropy automatic via independently-free, decoupled $\alpha(r)$ and $m(r)$ (no anisotropic-TOV solver). Correctness gate PASSED — the Fuchs isotropic baseline is representable in the free-(α,m,β) family and reproduces the Step-1 isotropic in-shell min(EC) to 5.7%, sign-consistent (one global C2 spline; the initial piecewise-C0 splice's −5.9e42 kink was fixed). Optimization scored against the Prong-B-certified radial evaluator only (Cartesian Prong-B-demoted).

Result graded:

Component	Grade	Detail
Step-2 anisotropic mass-reduction claim	C / rejected (KILL/KILL)	Optimizer plateaued at M_opt/M_ref = 0.9632 (3.7%); full-res min(EC) = −3.72e39 (DEC FAIL). Test A: DEC violation converges genuine under refinement (522 r coarse "+2.9e35 PASS" → 2088 r −3.43e39 → 4175 r×120θ×na160 −3.72e39 stable) — coarse-loop pass = discrete-minimization-grid under-sampling mirage. Test B (decisive, representation-internal): constant-density passes (certified radial) to ADM ≈ 2.79e27 while the anisotropic optimum FAILS DEC at 4.46e27 — anisotropy counterproductive.
Fuchs §6 "orders of magnitude" claim	rejected, both slices	Unsupported in isotropic (Step 1) AND anisotropic (Step 2).
Fuchs-mass over-provisioning sub-finding	B → radial-certified, strengthened	Constant-density passes to ADM ≈ 2.79e27 (≪ canonical 4.49e27) on the certified evaluator — but trivial uniform mass reduction, which DOMINATES both profile-shaping (Step 1) and anisotropy (Step 2). NOT §6 profile optimization.
`verification/aniso_step2*` harnesses	A (kept)	Shared parameterization, gate, optimizer, adversarial battery — tracked, reusable.

Slice scope (honest). NEGATIVE for this parameterization family (global-C2-spline free α,m,β), this optimizer (Powell, 28-dim, 700 evals, plateaued), this canonical config (R₁,R₂,v)=(10,20,0.02c). Not a proof no anisotropic shell can do better — but the decisive kill (Test B) is representation-internal to the certified oracle; across the whole Phase-3.3+ arc nothing approached beating constant-density-at-2.79e27.

No load-bearing dependency change. Composite Path 2A verdict remains A — strengthened (another exploratory loophole closed NEGATIVE; the verification discipline caught a coarse-mesh mirage before it was recorded).

Methodological refinement (A-grade, 3rd distinct instance — supersedes prior where they conflict). An optimizer mines whatever discretization is in its objective: Cartesian staircasing (S28); Cartesian-untrustworthy-for-sharp (S29); and even with an exact-certified curvature engine, the under-sampled discrete (r,θ,direction) minimization grid (S31). The reliable arbiter is jointly: cross-representation invariance + a certified-exact ground truth + a converged objective sampling mesh — with the optimum-plus-adversarial-battery as the catch (it caught all three). Recorded in feedback-no-cartesian-optimizer-objective.

Session 32 addendum — Task 3.10 certified minimal-mass map (positive quantitative closure)

Setup. Constant-density Fuchs shells, minimal EC-passing mass bisected per (R1, R2, v) cell against the Prong-B-certified radial evaluator (axisymmetric_ec) at the Session-31 full-res accept/reject tier; coarse scout tier proposes brackets only (S28–S31 converged-objective rule). Grid mirrors 3.2's axes (18 cells). Gate battery: S31 anchor regression (ADM to 0.01%), canonical-threshold consistency (floor must sit at/below the mass already shown to pass), FULL↔CONF mesh-escalation stability at the threshold — ALL PASS, plus stability spot-checks at the two extreme-κ cells and the narrow-window cell.

Results graded:

Component	Grade	Detail
Canonical floor M_min = 2.568e27 nominal / 2.650e27 ADM (over-provisioning 1.75× / 1.69×)	A (within slice)	Bisected on the certified oracle, rel_tol 0.5%; classification stable under mesh escalation both sides; reconciles S29/31 numbers (nominal 2.7e27 ↔ builder-ADM 2.786e27 — same probed point, two bookkeepings; both were upper bounds, not the floor).
Certified-radial κ surface: 4.64 ± 0.57 over 14 thresholds (range [3.61, 5.40], rising with R2)	A (within slice)	Canonical κ = 4.77 inside the 2A.9b Cartesian-era bracket (4.17, 5.83]; consistent with & refining 3.2's 18% spread. Protocol note: 2A.9b bisected Δ at fixed M (MATLAB WF, Cartesian era); 3.10 bisects M at fixed Δ on the certified evaluator. Thin-cell (Δ=3.75 m) κ inherits the ~5 m canonical smoothing length — convention, flagged.
Linear-in-β scaling of M_min (1–4% across map)	A (within slice)	Direct ratio check at fixed geometry; matches the 2A.5/2A.7 scaling-law form.
Binding condition at the floor: null (NEC) 13/14; strong (SEC) at the near-cap narrow-window cell	A (within slice)	From min_by_cond at the bisected threshold, full-res tier.
4 null-configuration cells at v = 0.05c (no EC-passing constant-density mass)	A (within slice, one stated structural assumption)	Golden-section peak of min(EC) over the horizon-valid mass range robustly negative (−1.6e39 … −1.0e40) per cell. Verdict inherits the unimodality assumption (min-EC vs M = rising NEC-support margin ∧ falling high-compactness margin); assumption recorded in module + ledger reopening trigger (ii). Certified-radial confirmation of 3.2's "null configuration" phenomenon.
`hf_jobs/sweeps/mmin_map.py` + `verification/test_mmin_map_gate.py` + `sweeps/mmin_map_full_concat.parquet`	A (kept)	Sweep module, 3-gate battery, negation-tracked 18-row artifact.

Slice scope (honest). Constant-density ρ(r) on [R1, R2]; TOV-pinned isotropic pressure; ℓ=1 dipole Alcubierre shift (canonical compact sigmoid); smooth_factor 4000 at fixed canonical radial sample spacing (physical smoothing length matched to Sessions 25–31 across cells); EC minima over the in-shell mask; radial representation only (Cartesian demoted per 3.9). Says nothing about non-constant profiles (Phase 3.3+ closed those NEGATIVE separately) or other shift/topology families.

No load-bearing dependency change. Composite Path 2A verdict remains A. The map quantifies the static-slice Fuchs family (over-provisioning = 1.75×, NOT orders of magnitude — Fuchs §6 stays unsupported in all three tested senses: profile, anisotropy, and now uniform-mass headroom).

Methodological refinement (A-grade — the DUAL of the S28–S31 lesson). Sessions 28–31 established that optimizers manufacture false POSITIVES by mining whatever discretization is in their objective. Session 32 adds: search/bracketing logic manufactures false NEGATIVES through structural blind spots, and "no result here" records need the same adversarial treatment as positive claims. Two in-session instances, both caught by kill-style spot-checks before recording: (1) a narrow passing window fell between scout-ladder rungs squeezed against the horizon wall — the cell was mis-recorded as horizon-capped while M = 7.41e27 passes cleanly; (2) the first fix still missed it because min-EC vs M is unimodal, not monotone — the window sits between two EC-FAIL rungs (fails low on NEC support, high on compactness). Correct pattern: search for the peak of the feasibility margin (golden-section maximize), and record the located peak value/location as the explicit basis of any no-pass verdict. Recorded in feedback-exhaustive-survey-is-the-method (refinement: the battery applies to negatives too).

Session 33 addendum — Task 2D.11 Phase 3 (FH-form multi-mode A) NEGATIVE; Task 2D.11 fully closed

Setup. Per-component FH-form vector potentials (gradient-normalised, independent radii; asymmetry/exponent inherited from the FH anchor), curl added to the irrotational shift, evaluated through the same adm_stress_energy_from_N pipeline as Phases 1-2 (bit-exact baseline regression at the canonical anchor). Two previews: perturbative (N_vort <= 0.5) and non-perturbative (N_vort 1.5-5.0), 2914 rows, 0 errors.

Results graded:

Component	Grade	Detail
Phase-3 negative (no improvement on any gate criterion; 100% strict degradation of both slacks; passenger zone unchanged)	A (within slice)	2912 augmented points; monotone collapse in total amplitude; 1184 strong-amplitude rows develop NEW WEC violations. Pipeline cross-checked by bit-exact baseline regression + Phase-1/2 comparability (identical record schema, same anchor).
Task 2D.11 composite verdict (three vortical families all NEGATIVE; irrotational restriction not the driver of the all-wall-no-interior pathology)	A (within slice)	The three families are structurally independent (rotating-frame axisymmetric; fixed-direction Cartesian; FH-form multi-mode). Phase 3 has maximal structural overlap with the violating regions — the Phase-1/2 "unhelpful overlap" caveat is closed, not dodged. Slice: canonical anchor, static smooth-N, tested envelope/radius/amplitude ranges.
Spin-2 no-cavity bridge (Task 1.11)	C (unchanged)	Reinforced, not proven: consistent with, but not derived from, the Costa-Natario catalog. The 2B.8 assessment is the right instrument to test it.
`fell_heisenberg_vortical_multimode.py` + two preview configs	A (kept)	Tracked, reusable; parquets reproducible in ~2 min each.

No load-bearing dependency change. Assumptions-table row 1's vorticity clause tightens from "does not lift it (Session 15)" to "does not lift it (Sessions 15, 33 — three families incl. FH-form multi-mode A)". Composite Phase-2D reading unchanged: every structural test degrades the warp-drive interpretation; none restores it.

Methodological note. The gate-driven preview discipline (decision gate stated in the config _comment before the run; definitive-at-anchor negative closes without dispatching the full sweep) held for the third consecutive vortical family — total Phase-3 compute cost ~4.5 min local, $0 remote.

Session 34 addendum — Task 2B.8: Path 2B closed as a physical mechanism

Setup. Literature assessment (web-verified primary sources + one review) + a tracked even-if arithmetic harness (verification/test_2b8_casimir_gap.py). Canonical record: QUANTUM_CLASSICAL_BRIDGE.md §8.

Results graded:

Component	Grade	Detail
Ordinary matter cannot reflect/absorb GWs (impedance $Z_G \sim 2.8\times10^{-18}$ SI mismatch; Dyson $\sim 10^{-41}$ cm²/g absorption; single-graviton detectors collapse to BHs)	B	Literature, uncontested across camps (the impedance statement appears in the pro-mirror school's own papers). Spot-checkable numbers; not re-derived here.
Superconductor H-C loophole: speculative, contested, unobserved	B	One programme + adopters; Quach 2015 explicitly conditional on H-C (erratum = units fix); 2022 review catalogs 55 years of contradictory results, no experimental support; originating programme's own 2022 refinement undercuts the mechanism. Not a refutation-theorem — a status assessment.
Even-if magnitude bound: perfect mirror + favourable sign ⇒ $	\rho_C	\le \hbar c/d^4$ ⇒ 63.5–69.6 OoM short of every radial-certified target; required spacing sub-proton ($\sim 10^{-16}$ m) through a metre-scale wall
Composite verdict: Path 2B CLOSED as physical mechanism (static sliver + acceleration supplement)	A (within slice)	Decisive leg is the magnitude bound; legs 1–2 close the mirror question itself. Mode-decomposition math (2B.1–2B.6) survives as classification tool; Claim (a) and the §5 effective-boundary (Path 2A) reading untouched.
Task-1.11 spin-2 no-cavity bridge	C (unchanged)	Consistent with 2B.8 (no cavity-forming boundary for gravitons in known physics) but still not derived from a formal catalog.

Slice scope (honest). The closure is within 4D semiclassical gravity ($G_{\mu\nu} = 8\pi G \langle T_{\mu\nu}\rangle$) with standard QFT Casimir scaling at macroscopic boundary scales. It does NOT exclude: modified gravity (Phase 2E.2 / Slice 6b), exotic field content or trans-QI vacuum states (Phase 2E.3), or horizon-based constructions incompatible with subluminal shells (none known). Those remain where they were — deliberately deferred with reopening criteria.

No load-bearing dependency change to Path 2A. Composite Path 2A verdict remains A; 2B.8 strengthens the overall landscape statement: within 4D semiclassical GR there is now no known candidate for a vacuum+DEC+dynamical warp realisation.

Methodological note. The decisive step deliberately repeated the Slice-4b pattern (grant the contested physics, bound the magnitude): the assessment does not need to win the contested superconductor argument — the loophole-independent bound closes the path at ~2× the 31-OoM Slice-4b standard, so the verdict is robust to any resolution of the mirror literature.

Session 35 addendum — pre-Phase-2E deep audit + remediation: grade corrections and disclosures

Setup. Four-thread audit of the whole closure record (closure-basis classification; retroactive evidence review vs the Session-30/32 lessons; harness code review + battery re-runs; cross-doc numeric/logic tracing), then same-session remediation. Full chronology in SESSION_LOG Session 35. Nothing below flips a recorded verdict; the corrections are magnitude/provenance/wording-level.

Disclosures and grade impacts:

Item	Impact	Detail
R1 mask bug (one-cell in-shell-mask misalignment, all Cartesian-path harnesses)	Corrects Session-30 Prong-B recorded numbers; verdict unchanged	Corrected sweep: Cartesian-vs-GT 9.1% → 94.0% over s=0.5→32 (recorded: 24.5% → 94.0%); radial = 0.0% throughout, no sign flips. The Cartesian demotion stands; its smooth-end penalty was overstated ~2.7×. The Session 26–27 marginal numbers (nested sign-flip threshold, oblate +3.09%) remain inside the corrected ~9% smooth-baseline band — still not trustworthy as magnitudes (unchanged reading, tighter band).
`mmin_map` S1/S2 latent bugs (stale `adm_hi`; horizon point as bisection endpoint)	No grade change	Verified never-fired on the recorded 18-cell map (parquet ADM/nominal ∈ [1.0037, 1.0347]); GATE-2 regression after fix reproduces the canonical cell to ≤1e-4 relative. Map remains A (within slice).
3.10 null-configuration cells	Grade reading clarified: "A (within slice, one stated structural assumption)" should be read as conditional-A — the assumption is untested exactly where it is load-bearing	The stated unimodality assumption is really per-condition monotonicity of each EC margin in M (which implies the ∧-shape the golden-section search presupposes). It is untested in the null cells (where no window was observed — precisely where observation cannot confirm it), and the SEC-binding near-cap cell shows a third margin curve participates. Reopening trigger unchanged (ledger).
κ = 4.64 ± 0.57	Definition recorded (was undefined in every doc)	±0.57 is the 1σ cross-cell dispersion of a κ that varies systematically with R₂ — a spread over the 14 located thresholds, not a random error bar on a universal constant. Per-threshold bisection precision is separately rel_tol 0.5%. NAVIGATOR's "over the 18-cell grid" descriptor corrected to the 14 thresholds.
2B.8 Leg 1 wording	Scope corrected; verdict unaffected	"No gravitational conductor exists in known physics" is a claim about matter (impedance + Dyson absorption). The §4 taxonomy omitted curvature-based partial confinement (QNM-style quasi-bound modes); not separately assessed, subsumed by Leg 3's perfect-mirror grant. QCB §4/§8 + ROADMAP re-scoped; triggers (ii)/(iii) sharpened (theorem-grade counterexample counts for (ii); (iii) reopens Legs 1–2 only). Grades unchanged (Legs 1–2 B, Leg 3 A within slice).
Slice-1 "0/140" (2C.1)	Evidence-quality caveat recorded; verdict stands pending kill-test	Preview grid only; R₀ frozen at 5.0; `shift_families_full.json` never dispatched; free-form j₁ ridge (0.94) unrefined — Session-32 false-negative shape. Symbolic Natário dismissal unaffected. Kill-test queued (ROADMAP Session-35 audit queue).
Hybrid-wall "0/480" (2C.2)	Provenance corrected; verdict stands pending kill-test	Sweep varied (η, δ_M, w_M) only — ε, n frozen at 1.0, 100; `hybrid_wall_full.json` never dispatched. ROADMAP description corrected.
2D.11 Session-33 closure	Scope caveat recorded	The augmented anchor's baseline is itself DEC-violating at Npts=49, so the strict-pass gate was uninformative; surviving claim = 100% strict slack-degradation at that anchor; anchor V=0.5 vs Session-11 winner V=1.5 unexplained; no strict-pass anchor ever augmented. NAVIGATOR anchor qualifier restored. Multi-anchor kill-test queued.
2A.10 GW-recoil ceiling	Standing C-grade re-flagged	TRUST_AUDIT #5 Colab pull still pending since Session 9; `gw_recoil_full.json` never run. Qualitative acceleration no-go (symbolic) unaffected. Queued.
thickness_bound provenance (2A.7)	Provenance corrected	The κ = 0.05 empirical calibration used the 600-row local preview; the ~1.3e5-point full config was never dispatched (SESSION_LOG Session 6 phrasing overstates). Superseded in practice by 2A.9b/3.10.
Stale doc numbers	Corrected in place	Four stale [0.05, 0.75] brackets → [0.05, 0.875] w/ supersession notes; README Path-2B/Phase-2C statuses; LANDSCAPE_SYNTHESIS §6/§8 stale + self-contradictory items; phantom task 2D.16 retroactively defined (ROADMAP Phase 2D).

Methodological note (fourth instance of the searcher-honesty family): a verification harness is itself code and can carry a systematic bug (R1) for five sessions while every verdict it reports remains directionally correct — because the battery's verdicts were sign/trend-based with large margins. The discipline that caught it was auditing the evidence (masks, grids, provenance) rather than the conclusions. Corollary recorded in the ROADMAP audit queue: negative results whose grids froze axes or whose full configs never ran (W1/W2) are now flagged in place in the claims record, not just in session logs.

Session 36 addendum — Slice-1 (2C.1) frame-projection bug: grade corrections

Setup. Block 2(a) of the Session-35 audit queue (kill-test of the Slice-1 "0/140") escalated: the pre-run diagnostic violated an exact identity, and adjudication found the Slice-1 evaluator itself defective. Full chronology in SESSION_LOG Session 36; adjudication harness verification/test_shift_families_frame_adjudication.py (20/20 gates, certification mode).

Disclosures and grade impacts:

Item	Impact	Detail
Slice-1 evaluator frame projection (`shift_families.ipynb` Cell 3 + sweep module, Sessions 9→36)	All recorded Slice-1 numbers superseded as wrong-observable; verdict unchanged and upgraded	Tetrad legs stored as matrix columns but contracted as rows (M T Mᵀ for Mᵀ T M); recorded scalar = coordinate −T_tt, not ρ_E (deviation 2.0–8.6× at the recorded single-point). Affects: single-point table, 140-pt fractions, "best 0.94 free-form ridge" (does not exist on the corrected observable: max 0.0027), Q_zz quadrupole-proxy table (regions selected by the defective observable).
Slice-1 corrected result	Upgraded to A (analytic, within slice)	Four profile-independent identities close all four families for every parameter value: (i) z-shift ρ_E = −b′²sin²θ/32π ≤ 0 (alcubierre, freeform, any radial multi-mode); (ii) Natário ∇·β ≡ 0 ⟹ ρ_E ≤ 0 (concordant with the Session-15c FH Phase-3b proof, which the recorded table had contradicted since Session 9); (iii) irrotational ∫ρ_E dV = 0 (with analytic 1/r⁴ dipole tail −v²C²/6R, verified to 1.1e-08) ⟹ WEC-everywhere forces ρ_E ≡ 0. Corrected sweeps: 0/140 preview, 0/2496 full config (first dispatch), corrected observable certified against warp_factory_py anchor chain (median 1.6e-07).
Irrotational domain truncation	Provenance corrected	sympy `log(1±tanh)` antiderivative overflowed for \|r−R₀\| ≳ 19/σ; recorded irrotational fractions were computed on the silently truncated finite subset. Fixed via equal-constant log-cosh form (`_LogCosh`, float64-safe).
Session-15c / LITERATURE / MATTER_SHELL_PATH "dismissed as special case of Slice 1" chains	No grade change; consistency restored	Those dispositions rested on the FH Phase-3b identity (independent pipeline, unaffected). The Slice-1 table was the outlier; corrected numbers now agree with the identity chain.
Krasnikov-class frames (`krasnikov_tube.py`, `hybrid_wall.py`)	No change	Audited: their metric has g_tt = −1 and their tetrad rows ARE orthonormal legs consistent with the row-wise contraction (algebra closes: g(u,u)=1, g(∂t,u)=0). The transpose defect is confined to shift_families. 2C.2's kill-test (Block 2(e)) remains a coverage question only.
Fell-Heisenberg pipelines (Sessions 10–17, 33)	No change	FD-based on FH's closed-form ρ_E decomposition; no shared frame code (grep + Session-35 review).

Methodological note (fifth instance of the searcher-honesty family): a result pipeline can manufacture plausible positive structure (the 0.94 ridge) inside a directionally-correct negative, and an already-proven in-repo identity (Session 15c) falsified the recorded table for 14 sessions unnoticed. When a sweep table and a symbolic identity coexist, run the cross-check at closure time — the identity is the cheaper, stronger audit.

Session 37 addendum — TRUST_AUDIT #5 CLOSED (C → B): GW-recoil SXS anchor verified conservative

Item #5 (GW-recoil ceiling: SXS rescaling β²C^{3/2} heuristic, C-grade since Session 8; Cell 17 Colab path wired Session 9, never run). Closed 2026-07-05 by verification/test_sxs_kick_pull.py (4/4 gates), which pulls the SXS data over plain HTTPS — Zenodo per-record metadata for SXS:BBH:1937 cross-checked against the collaboration's catalog.zip (identical) — with no sxs package or Colab required.

Findings:

Finding	Detail
Cell-17 design defect	The wired comparison targeted SXS:BBH:1937 as the "high-mass-ratio kick record per Varma 2022" and expected its remnant kick to confirm 5000 km/s within 1.5×. The simulation is actually q = 4.0 aligned-spin non-precessing (χ₁⊥ ≈ 9e-7) with remnant kick 93.6 km/s — 53× below the expectation; aligned-spin systems cannot superkick. The success branch of that cell has never been correct; every recorded execution took the fallback path. Superseded by the harness.
Anchor verified conservative	Catalog-wide max remnant kick over 2021 public SXS simulations = 3119.1 km/s (SXS:BBH:0662, q = 1.33, χ₁⊥ = 0.80; top-5 all near-equal-mass precessing). The Package-3 input 5000 km/s upper-bounds every public NR simulation (1.60× headroom) ⟹ the recorded Mechanism-C ceiling is conservative.
Full-config sweep	First `gw_recoil_full.json` dispatch (4320 pts; preview regression bit-exact on all 1200 recorded rows). Ceiling Δv/(βc) ≤ 0.58% for physical C ≤ 0.5 (1.41% only at the unphysical (0.99, 0.9) corner); canonical Fuchs point unchanged (~1e-4). Recorded "max at β=0.9, C=0.5" corrected to C=0.3 (label error; the ratio 0.25% was computed correctly). M-axis of the grid is analytically degenerate in both formulas.

Grade: C → B (within slice: single-bubble Fuchs-class shells, quadrupole-order rescaling). Not A: the remnant velocities are accepted from SXS collaboration metadata rather than derived by integrating waveform momentum flux ourselves. Reopening trigger: an NR result or surrogate prediction exceeding 5000 km/s for astrophysically admissible spins, or a shell-specific radiation channel outside the β²C^{3/2} rescaling class.

Methodological note: dormant verification code (a wired-but-never-run success branch) can encode a wrong expectation while conferring an appearance of rigor. When closing long-dormant audit items, re-derive the check from the primary data source rather than finally executing the recorded button.

Session 38 addendum — 2D.11 evidence re-based (W3 kill-test): verdict upheld; Session-33 strict-pass gate superseded

Setup. Session-35 audit item W3: the 2D.11 Phase-3 closure augmented one anchor whose baseline was DEC-violating at the run's Npts=49 (gate vacuous) with an off-grid V=0.5. Block 2(c) kill-test executed 2026-07-05 (Session 38).

Disclosures and grade impacts:

Item	Impact	Detail
Session-33 anchor baseline	Diagnosis corrected: resolution artifact, not anchor choice	The anchor structure (σ=10, m₀=3, a=0.05, ℓ=4, r=9) is certified strict-pass at Npts=65 for every V on the Session-11 grid; the recorded dec_slack −7.74e-2 was an Npts=49 under-resolution effect. V-choice immaterial: slacks scale exactly as V² (strict-pass signs V-invariant; hence the sweep's exactly-234-per-V strict-pass split).
Session-33 "0 strict passes among 2912 augmented points"	Superseded (baseline-inherited)	At four certified strict-pass anchors × the same vortical grid at Npts=65 (7280 augmented points), 100% retain strict-pass at preview amplitudes — including a +8.3e-5-margin anchor under vortical fields 3× the FH amplitude.
2D.11 verdict (vorticity not the driver of "all wall, no interior")	UPHELD; evidence upgraded to informative gates	0/7280 improve either slack (universal degradation confirmed, ∝ V·V_A, ≤0.26% of margin at $\|V_A\| \le 0.3$); passenger_zone_radius = h for all 7285 rows including baselines. Baseline regression vs certified sweep rows ≤4.2e-5 rel (A3: ~1.3e-7 absolute). Within slice: static smooth-N, FH-form multi-mode $\vec A$ with inherited exponents, $\|V_{A,i}\| \le 0.3$ at Npts=65.
Session-33 absolute numbers (dec_slack −7.74e-2 baseline etc.)	Flagged as resolution-contaminated	Differential (same-grid) claims unaffected; absolute Npts=49 slack values at this anchor family should not be quoted as physical.

Methodological rule recorded: a perturbation study's baseline must pass its decision gates at the study's own resolution, else those gates are vacuous — the FD sibling of the Session-30 regime-validity lesson, applied to study design.

Session 39 addendum — nested-shell (Phase 3.3 sub-item 4) REVERSED: first verdict flip of the audit programme

Setup. Session-35 audit item W5 (first half): the Session-26 nested-shell mass-split threshold sat inside the demoted Cartesian pipeline's error band. Block 2(d) kill-test executed 2026-07-05 (Session 39) via the new tracked harness verification/test_nested_shell_radial_ladder.py (identical physical configuration; certified evaluate_axisym_ec; mmin_map resolution tiers; staged RES_CONF confirmation).

Disclosures and grade impacts:

Item	Impact	Detail
Session-26 recorded ladder (monotone degradation, flip in (0.10, 0.20))	SUPERSEDED — verdict reversed	Certified radial: min(NEC) rises from +2.8764e38 (f=0) to +2.3299e39 (f=0.10) — an 8.1× improvement — before declining; certified sign flip $f^* \in [0.6234, 0.6312]$ (RES_CONF, df ≤ 0.01; RES_FULL↔RES_CONF plateau agreement ≤1%). The recorded numbers came from the near-equatorial thin-slab Cartesian convention (1, 300, 300, 5); the certified evaluator minimises over the full (r, θ) mesh. GATE 1: nested builder at f=0 ≡ single-shell profile builder to 0.0 relative.
"Fuchs single-shell locally optimal under mass redistribution" (Phase 3.3 composite verdict, was A within slice)	Refuted within slice; composite verdict revised	The nesting leg of the composite is reversed; the corrected statement (improvement plateau + $f^* \approx 0.63$) is A (within slice: two-shell constant-density, fixed radii (5,8)/(10,20), outer-wall warp band, v = 0.02c, fixed $M_{tot}$ = 4.49e27). Session-26's physical reading inverted the enclosed-mass argument (inward mass-splits increase $M(r)$ throughout the band at fixed $M_{tot}$).
Session-27 oblate +3.09% (W5's other half)	Still unpinned	Same thin-slab Cartesian provenance; NOT re-tested this session. Until re-run radially, treat the oblate numbers as direction-unknown (the nested reversal demonstrates the band can hide sign and structure errors, not just magnitude).
New candidate lead	Recorded (unranked)	Nested-variant minimal-mass map: does the ~8× margin improvement translate into a lower certified $M_{\min}$ than 3.10's 2.568e27? ROADMAP "Unranked candidate (Session 39)".

Methodological note: first demonstration in this programme that the demoted pipeline's error band hid a wrong-shape record (non-monotone → recorded as monotone) and a wrong-window threshold (0.63 → recorded as 0.1–0.2), not merely imprecise magnitudes. "Kill-test the negatives" is not ceremonial: three strengthenings and one reversal in four items.

Session 40 addendum — 2C.2 hybrid-wall coverage kill-test (W2): negative strengthened

Item (W2). The 2C.2 "0/480" had frozen axes ($\epsilon$, $n$) and an undispatched full config. Closed 2026-07-06 (Session 40) by the first hybrid_wall_full.json dispatch: 82,944 points, all five axes swept, preview regression bit-exact on all 480 recorded rows.

Item	Impact	Detail
2C.2 "0/480"	Strengthened to 0/34,560 (WEC, η ≥ 0.1) and 0/82,944 (DEC, anywhere); grade A within slice	The unfrozen $\epsilon \in [0.3,5]$ × $n \in [10,200]$ axes hide nothing (per-cell best-fraction flat at each η plateau); best functional-tube wec_fraction 0.891. Slice: single-bump $\delta_M B_{w_M}$ perturbation of $k(\rho)$, static, 1-D radial evaluation in the static orthonormal frame.
Raw-artifact caveat	Recorded	236 rows with wec_fraction ≥ 0.999 exist at η ≤ 0.004 — the tube-off trivial limit (near-flat + matter bump). Not a loophole; flagged in ROADMAP/NOTES so the parquet is not quote-mined.
Krasnikov-class frame	Hand-check upgraded to scripted certificate	The Session-36 audit's algebra (tetrad rows orthonormal for the $g_{tt}=-1$, $g_{tx}=(1-k)/2$, $g_{xx}=k$ metric class) verified symbolically for generic $k(\rho)$: all 16 components of $e_a^\mu e_b^\nu g_{\mu\nu} - \eta_{ab}$ simplify to 0. The hybrid_wall/krasnikov_tube observables are genuine static-frame components.
Config description	Provenance note	`hybrid_wall_full.json`'s "~25000 points" comment undercounts the actual 82,944-point grid 3.3× (the gw_recoil config had the mirror-image defect — overcounting; config `_comment` size estimates are evidently unmaintained — treat as non-authoritative).

Session 41 addendum — Δ-ladder certified (W4 closed); Session-19 κ statistics re-based; Block 2 complete

Item (W4, thickness direction). Closed 2026-07-06 by verification/test_delta_ladder_radial.py (all 27 Session-19 cells through the certified mmin_map.min_ec path).

Item	Impact	Detail
Session-18 anchor κ ∈ (4.17, 5.83]	CONFIRMED and tightened; A within slice	Certified: κ ∈ (4.479, 4.583] (Δ_min ∈ (5.375, 5.500] m), NEC-binding. The 6× gap to the analytic 2A.7/2A.9a upper (0.875) stands.
Session-19 surface statistics (mean 5.3, std 1.0, "κ ∈ (3, 7]")	Superseded — mixed populations	The recorded stats pooled genuine crossings with floor artifacts ("NaN-lower" rows) and cap artifacts. Certified re-basing: κ = 4.93 ± 0.44 over the 12 genuine crossings (range [4.36, 5.87]); rising-with-R₂ trend is real (not wall-resolution, as S19 suspected) and matches 3.10's mass-direction trend; MATLAB sweep-resolution values ran ~13–20% high. Grade: certified cells A within slice.
Recorded MATLAB thin-wall failures (β=0.005 rows etc.)	Pipeline-discrepant	12 cells pass down to a 0.5 m floor radially (margins to +1.8e39) where MATLAB recorded failures at Δ ≈ 2–3 m. Convention caveat: at SF=4000 the ~5 m smoothing width exceeds these Δ — nominal thin walls are wide low bumps; Δ < ~5 m is not a physical thickness in this construction (both pipelines share the convention; only the radial one evaluates it correctly).
Nulls	3 confirmed; 2 candidate new nulls REFUTED (harness's own artifact)	β=0.05/C=1/6 nulls verified at the scout peak (−4.7e39…−6.8e39). The harness's first run declared (1/3,15,0.05) and (1/3,20,0.05) null via a coarse ladder + wrong-point (cap) verification; fine scans found genuine narrow windows (κ lower crossings (4.44,4.67] and (4.67,4.83]). Null path fixed.
3.10 cross-check	Cross-direction consistency	Δ-direction κ = 4.93 ± 0.44 vs mass-direction 4.64 ± 0.57 — overlapping, same R₂ trend, same NEC-dominant binding. The certified κ surface is now pinned from two independent directions.

Methodological note (sixth in the searcher-honesty family): a kill-test harness manufactured its own false negatives (coarse rungs straddling a narrow window + verifying the wrong point) and was caught only by applying the S32 discipline recursively — fine-scanning the scout peak before accepting any null. Rule: null verdicts require full-resolution verification at the most favourable probed point, not at a boundary.

Block 2 complete (Sessions 36–41): four strengthenings, one reversal (nested-shell, first flip of the programme), one confirmation-with-re-basing. Grade ledger net effect: Slice-1 → analytic A; #5 → B; 2D.11 → informative-A within slice; Phase-3.3 nesting leg → reversed (new candidate lead); 2C.2 → A within slice at 72× coverage; Δ-ladder/κ → certified A within slice, two-direction pinned.

Session 42 addendum — Task 2D.5e: closed-form FH pressures; the strict-pass record re-scoped to its evaluation box

Setup. The Session-14c Hard Fix verdict ("the FH potential is structurally too complex for closed-form principal pressures; fundamental property of the ansatz, not a SymPy limitation" — §12.4) and the entire Sessions-11–17 strict-pass record are affected. Battery: verification/test_fh_axisym_closed_form.py, 9/9 gates.

Item	Impact	Detail
§12.4 "no closed form exists" verdict (Session 14c)	Corrected	The adopted FH ansatz is exactly axisymmetric about Z; the y=0-plane block decomposition gives closed-form eigenvalues everywhere (SymPy literal zeros for the off-block components; eigenvalues ≡ eigvalsh to 4.4e-16). The det() wall was symmetry-blindness, not ansatz complexity. Grade of the closed-form pressures: A (symbolic certificates + machine-precision cross-checks + FD validation ≤ 2.1e-4 at certified anchors).
Sessions 11–17 "strict-pass" statistics (1404/6818/6240 counts, connectivity, CTC fractions, VIQ, xAct anchors)	Re-scoped: correct as L=12-box statements; withdrawn as global-EC statements	Closed-form far field: every tested $a>0$ configuration violates WEC+DEC beyond finite equatorial $R^$ (17–200 over probes; V-invariant), with diverging magnitude. FD cross-check: A1 wec_slack_min +0.0374 (L=12) → −0.848 (L=45)* through the same pipeline that produced the record. The swept grids used $a \ge 0.05$ throughout ⟹ within the adopted $m,n$ concretization, the globally-EC-passing subset of the swept family is empty. $a=0$ inconclusive at the 1e-5 level.
NAVIGATOR load-bearing row 1 ("single-mode axisymmetric — load-bearing AND broken")	"Broken" withdrawn	The multi-mode counterexample was box-scoped. Post-S36 (single-mode analytic closure) + S42: no tested shift family passes WEC+DEC globally within the explored slices.
§11 boundary resolution-flakiness (47% pass→fail flips)	Structurally explained	Marginal cells are configurations whose $R^*$ falls near the box's corner reach; their classification depends on exactly which far-field cells the grid samples.
Session-38 kill-test conclusions (2D.11)	Differential content unaffected	Same-box comparisons (augmented vs baseline at L=12); its "genuinely strict-pass anchors" language inherits the box scope. Vorticity's universal degradation and the passenger-zone results stand as stated within slice.

Methodological rules recorded: (i) inspect symmetries before declaring symbolic intractability; (ii) fixed-box pointwise-EC evaluation of a non-decaying ansatz requires an explicit far-field gate — "pass" within a window is not "pass." (Seventh and eighth entries in the searcher-honesty family.)

Session 44 addendum — 2E.4 Π-exponent axis closed NEGATIVE (A within slice)

Item	Impact	Detail
"Π = 1/4 non-smoothness drives the passenger-zone pathology" (standing hypothesis since Session 14)	Refuted	Passenger zone = single voxel at every Π ∈ [0.125, 1] across three certified anchors; central \|N\| 13–22 throughout.
2E.4 first axis (exponent variation)	Closed NEGATIVE; A within slice	Dual-box protocol (L=12 + L=45 far-field gate), Npts=65, Π=0.25 baselines regress exactly. Box strict-pass only on Π ≲ 0.3; far-field violation at every Π (the R-linear φ growth carries no Π — same mechanism as the §18/S42 closure). Slice: adopted m,n concretization, anchors A1/B1/S12, Π ∈ [0.125, 1]. Remaining 2E.4 sub-axes (topology; joint vortical+Π) open.

Session 45 addendum — Lentz 2020 closure upgraded from analogy (C) to class-level certification (A within slice)

Battery: verification/test_lentz_full_wec.py, 5/5 gates.

Item	Impact	Detail
Lentz 2020 closure basis ("special case of the FH irrotational family" — analogy, effectively C)	Upgraded to A within slice	Class-level theorem: every compact ℓ¹ member has a purely unidirectional outermost wavefront on which ρ_E ≡ 0 (marginal) while the (u,z) stress block carries a traceless ±λ pair, λ ∝ (front curvature)² — full WEC and DEC strictly violated at every amplitude (exact quadratic scaling certified). Closed-form quadrant reduction cross-validated against the 3D ADM pipeline (median 4e-5). Slice: published ℓ¹ ansatz class, any v_h, any source.
Lentz 2020's own Eulerian-positivity claim (B — accepted from the paper)	Re-graded: unverifiable from the published record	The digitised Fig.-1 source under the correct 1+1 propagation gives ~37% ρ_E < 0; no per-station rescaling (343 tested) recovers positivity. The property lives in per-chord fine structure the figure does not carry. Not asserted false — asserted not reproducible, upgrading Bobrick–Martire's process complaint to a structural fact. Moot for our verdict: even exact Eulerian positivity is marginality on the fronts, where the class-level violation lives.
Session-43 "non-decaying wake" observation	Withdrawn (artifact)	The S43 3D march solved the isotropic 2D-transverse equation; Eq. 18 is the 1+1 kernel. Correct propagation: wake decays, Fig.-2 signatures reproduced.
Methodological rule (ninth in the searcher-honesty family)	Recorded	A machine-exact residual on the wrong PDE passes every internal gate — identify which equation the paper's kernel actually is (here: from the figure's characteristic slopes) before trusting a reconstruction.

Session 46 addendum — Garattini–Zatrimaylov 2025: literature grade B → A within slice (reproduced exactly); sharpenings recorded

Battery: verification/test_gz_desitter_reproduction.py, 9/9. Record: GARATTINI_ZATRIMAYLOV2025_EVALUATION.md.

Item	Impact	Detail
G-Z 2025 Eqs. 6/14/17/18/20/23/24–27/30–31 (B — accepted from the paper, behind NAVIGATOR assumption 5b)	Upgraded to A within slice	Every equation verified at machine precision, incl. against the full 4D Einstein tensor of the exact time-dependent moving-bubble metric (2e-15); Eq. 6 ≡ Hamiltonian constraint; sign convention pinned ($K_{ij} = +\tfrac12(\partial_iN_j+\partial_jN_i)$); exact mass-conserving rearrangement; unmatched-trajectory control shows the Hubble matching is load-bearing. First external construction in the programme to survive reproduction.
G-Z underdensity theorem (Eqs. 32–40)	A for the paper's mass-conserving class, via a repaired proof step	Eq. 38 ("the only way… is Q ≡ 0") overreaches: WEC forces only $Q \ge 0$; explicit counterexample families to the proof step recorded (theorem conclusion untouched — they have no fast-frame underdensity). Repair: Eq. 41 mass conservation ⟹ $\int Q,d^3r = 0$ ⟹ $Q \equiv 0$.
Session-46 sharpenings (ours): comoving lock; ANEC violated on every wall-crossing ray tested; local violations scale-locked to $\hat\rho$ ($1/L^2$, shape-set multiple)	A within the stated probe scope	Plain (not achronal-restricted) ANEC; meridional-plane ray family at one parameter set; compact-support wall required (flat-patch past-incompleteness makes ANEC ill-defined for non-compact tails).
NAVIGATOR 5b row basis	Re-based lit-B → reproduction-A; "no useful-warp loophole"	The 5b qualifier is real but describes a comoving vacuum void with zero transport content; pointwise WEC/DEC still violated in the wall; ANEC violated.
Probe-integrity instances (S32 "search logic can manufacture false no-results" family)	Recorded	(i) fixed-point launch = probe that cannot reach its target silently reads as a null result; (ii) integrating past a finite-affine patch boundary manufactures signal; (iii) non-compact profile tails amplified by horizon blueshift make an averaged quantity formally divergent. All three baked into the battery as asserted integrity gates.

Session 47 addendum — nested minimal-mass map: canonical floor re-based; S39 reversal quantified as a mass lever

Battery: verification/test_mmin_nested_map.py (audit 4/4; RES_CONF certify on 3 rows, all gates PASS). Map: sweeps/mmin_map_nested_full_concat.parquet.

Item	Impact	Detail
Nested/graded-wall minimal-mass map (100 points, 4 cells)	New, A within slice	Derived end-to-end in our harness through the certified radial oracle; the certified `find_mmin` search logic reused unchanged (oracle injection; refactor regression bit-exact); f=0 baselines EXACT vs the S32 map at all 4 cells; bracket honesty on all 96 pass rows; RES_CONF escalation on the three headline rows (walk-up ≤ 1 step of rel_tol).
Canonical-cell certified floor (Task 3.10 headline number)	Re-based 2.568e27 → 2.2256e27 nominal (−13.3%)	Two-component graded wall, f=0.10, inner (7, 9.5) m; κ_nominal 4.77 → 4.13; over-provisioning vs the canonical 4.49e27 configuration: 1.75× → 2.02×. The single-shell 2.568e27 remains the correct single-shell floor; cite the pair with their wall classes.
Session-39 reversal (margin ~8× at fixed M)	Quantified as a mass lever: real but strongly sublinear	~8× margin surplus converts to −13.3% (canonical) / −17.8% (R₂=15) certified mass reduction; margin collapses steeply as M drops. Thin-wall Δ/R₂=0.25 cell: NO improvement at any (f, geometry) — monotone degradation, structural negative.
jaga as compute provenance	Accepted (validated pipeline + exact cross-checks)	Same code path as local (PYTHONPATH dispatch of repo modules); pipeline validated bitwise on thickness_bound the same day; S47's own f=0 baselines reproduce S32 local results exactly through the full nested code path. OOM-sizing lesson (7.6 GB/point peak, 28-worker cap) recorded in the machines cookbook.

Session 48 addendum — graded-wall map NEGATIVE (A within slice); S47 mechanism reattributed

Battery: verification/test_mmin_graded_map.py (audit 4/4, certify 3/3, adjudicate 2/2). Map: sweeps/mmin_map_graded_full_concat.parquet.

Item	Impact	Detail
Continuous graded-wall family (82 points, 2 cells)	New, A within slice — bounded NEGATIVE	0/80 members beat the S32 single-shell floor; best member (minimal extension) +1.2% above ref; negative RES_CONF-certified; d=0 baselines EXACT vs S32 through the graded rho_of_r path.
S47 "graded wall" interpretation	CORRECTED (verdict and numbers unchanged)	Cross-builder discriminator at identical nominal ρ: per-shell-TOV floor 0.8824× ref vs single-TOV floor 1.1409× ref. The S47 lever is the two-body pressure ansatz (per-component P=0 surfaces shaping the lapse), not density grading. Both families honestly evaluated (T_μν from G_μν); S47's 2.2256e27 EC-passing configuration stands.
S47 mask robustness	Kill-tested, PASS	Union vs contiguous mask min(EC) identical to 0.00e+00 rel at the corrected floor; worst point mid-wall (r≈12.5 m), nowhere near the standoff sliver.
jaga NUMA dispatch	Measured: pinning no-op; policy recorded	680.1 s vs 681 s A/B on identical points; bandwidth ceiling is intra-node; default dispatch stays unpinned with RAM-sized pools (machines cookbook).

Session 49 addendum — spin-up margin surface (2E.1 first leg): A within slice

Battery: verification/test_spinup_margin.py (audit 4/4, exact 2/2). Surface: sweeps/spinup_margin_full_concat.parquet.

Item	Impact	Detail
Time-dependent axisymmetric evaluator (`evaluate_axisym_ec_timedep`)	New solver capability, A	Exact symbolic Einstein tensor of the rigid-profile v(t) family; static limit exact (9.5e-14 lambda-level, ≤6e-14 per-row); structure facts exact in symbolic arithmetic (no v̈ anywhere; vd in diagonal + flux rows only).
Quasi-static corridor (three certified configs)	A within slice — CLEAN	min-EC(v, vd=0) ≥ 0 for every intermediate v; corridor minimum at v_target. The S32 M_min(v) monotonicity assumption kill-tested and upheld on an 8-point ladder.
Spin-up rate bound τ*	A within slice, order-of-magnitude precision	24.5/50.4/47.9 ns (0.37/0.76/0.72 R₂/c) for canonical/floor/two-body; every v-row caps (bounds real); DEC binds near the floor; log-grid interpolated crossings (not bisected) — stated.
S32 cross-anchor drift 7.2e-6	Explained, within the S35 band	The S32 parquet predates the Session-35 behaviour-preserving mmin_map fixes (≤1e-4 band); masses matched exactly in S47, values drift inside the band. Battery GATE 3 tolerance documents this.
2E.1 status	First leg closed POSITIVE-within-slice	EC bookkeeping does not obstruct rigid-profile inflate-then-coast; remaining 2E.1 content (dynamical restructuring, worldline acceleration, superluminal transitions) stays deferred with original criteria.

Session 50 addendum — 2E.4 closed: far-field coefficients exact; topology + joint axes negative

Battery: verification/test_2e4_residual_axes.py (topology 4/4; joint 5/5). Map: sweeps/fell_heisenberg_joint_2e4_full_concat.parquet. Record: FELL_HEISENBERG_SWEEP_NOTES §20.

Item	Impact	Detail
Far-field saturation φ_sat with Π symbolic	A (exact)	GATE T1 residual literal 0; saturation certified vs exact φ to 6.3e-15 at Π-scaled radii (T4); R-linear coefficient −V√(σπ) independent of every family parameter (T2); anisotropic growth ∝ a·tanh(Z/ℓ) (T3). Upgrades the §18.3/§19 mechanism statements from derived-numerics (B) to symbolic-exact (A).
Topology sub-axis of 2E.4	CLOSED NEGATIVE, A within slice (analytic)	Compact quotients admit no family member (linear term ⟹ no closed-manifold potential; T³ harmonic freedom = constant boost; lens spaces none); interior surgeries leave the finite-R* exterior violation untouched. Slice: the adopted flat-space family; novel compact-topology ansätze outside it not asserted.
Joint vortical+Π sub-axis of 2E.4	CLOSED NEGATIVE, A within slice (180/180)	Far-field gate negative at every cell incl. per-component exponent splits; passenger zone single-voxel everywhere; 0/168 box-slack improvements; baseline regression exact vs S44. Normalisation subtlety handled: A-fields normalised once on the L=12 grid and reused at L=45 (a per-grid renormalisation would have silently changed the ansatz between boxes).
Task 2E.4	CLOSED in full	All three sub-axes negative (S44 + S50). Reopening criteria unchanged.

Session 51 addendum — f(R) matter evaluator certified; quadratic-f(R) first physics

Battery: verification/test_fr_matter.py (6/6). Evaluator: warp_factory_py/solvers/fr_matter.py + generated fr_correction_generated.py. Design record: MODIFIED_GRAVITY_LIT.md §6b.

Item	Impact	Detail
f(R) = R + αR² Jordan-frame matter evaluator	New solver capability, A	Split form T = G_certified + αC (GR part byte-identical to the certified pipeline); C exact-cancelled and code-generated; anchors: Schwarzschild vacuum persistence 6.8e-16, de Sitter Einstein-space 6.0e-15 with R = 12/L² exact, α→0 rel 0.0, machine-precision small-α linearity, regeneration cross-check 5.6e-7.
Alcubierre non-rescue	A within slice	Best improvement over α ∈ ±[1e-2, 1e5] m² is 0.036%; violation amplified ~linearly beyond
Fuchs-floor degradation	A within slice	Viable direction strictly shrinks margins (slope −6.47e37/m²); EC window αR ≲ 1e-3 (dynamically negligible); helpful direction tachyonic + destructive beyond α ~ −10.
Slice caveat (standing)	Recorded	Jordan-frame test on the loophole's own terms; Einstein-frame conformal reading dissolves the loophole; quadratic f only — designer-f leg (with the single-function/non-monotonic-R overdetermination question) still open for S52.

Session 52 addendum — 2E.2 closed: α-map uniform negative; designer-f pointwise LP theorem

Batteries: verification/test_fr_matter.py map (4/4); verification/test_fr_designer_lp.py (4/4). Artifacts: sweeps/fr_alpha_map_full_concat.parquet; modules hf_jobs/sweeps/fr_alpha_map.py, hf_jobs/analysis/fr_designer_lp.py.

Item	Impact	Detail
α × geometry map (161 pts, 7 configs)	A within slice — uniform negative	No rescue of any EC-violating config at any α (best 0.2%); best viable α = 0 for every EC-passing config; floor EC window αR < 1.4e-5 at RES_FULL; α=0 baseline rel 3.8e-8 vs certified; jaga/local cross-determinism exact.
Designer-f NEC feasibility LP	A within slice — theorem-grade negative on the Alcubierre class	NEC linear in (f′,f″,f‴) per R value; 24/24 level-set bins infeasible both modes; pointwise form: 23.4% of wall points individually infeasible (kills every f(R) with f′>0, no global structure needed). Row machinery certified vs the quadratic evaluator at 1.15e-10; HessR recovered exactly from the generated correction; ∂R spline vs FD cross-check 3.8e-4. Control: Fuchs floor 0/24 (GR feasible there, as required).
Slice-6 canonical table row	Narrowed: f(R) corner closed	"4D Einstein gravity" assumption: the f(R) family (any f, f′>0) is closed for the tested warp-wall classes; Horndeski/f(R,T)/EGB untested — recorded in MODIFIED_GRAVITY_LIT.md.
fr-evaluator compute class	Ops: 11.2 GB/point	New sweep-family peak measured (rule followed belatedly after one OOM); jaga cap 20 workers; cookbook updated.

Session 53 addendum — standoff axis retired (A within slice)

Battery: verification/test_mmin_nested_map.py audit 4/4 on sweeps/mmin_map_nested_standoff_concat.parquet.

Item	Impact	Detail
Two-body standoff axis (9 gaps × 3 f, canonical cell)	Retired — no floor change	g ≲ 0.5 m is a bit-identical plateau at the S47 winner (2.21451e27 RES_FULL); monotone degradation beyond; worse than single-shell by g = 5 m. The S48 contact-vs-gap hint was a thickness confound (d = 3 vs 2.5 m). Certified floor 2.22558e27 unchanged.

Sessions 54–55 addendum — final optional residues retired

Batteries: verification/test_oblate_full3d_retest.py; verification/test_lentz_kink_sheets.py (5/5).

Item	Impact	Detail
Session-27 oblate +3.09% (last thin-slab-provenance number; W5 second half)	REVERSED — the oblate ε=−0.1 'improvement' never existed: the full-3D re-test gives Δ = −0.92% (N=97) / −1.26% (N=129), sign-consistent and resolution-stable (spread 0.34 pp), agreeing in sign with the thin-slab reconstruction (−3.1%); the recorded +3.09% was a sign error in the lost scratch. Session 27's headline conclusion (spherical local optimality under shape deformation) is thereby STRENGTHENED — the full-3D ε<0 ladder is monotone with no improvement pocket (−0.3: −5.0%, −0.2: −2.7%, −0.1: −1.3%, +0.1: −15.3% at N=129).	Provenance probe: four driver reconstructions incl. an exact mask-count match all give Δ ≈ −3.1 to −3.4% — recorded magnitude, opposite sign (sign error in the lost scratch). Full-3D re-test at three resolutions with certified ε=0 calibration renders the standing verdict. W5 fully discharged.
Lentz ℓ¹ kink sheets	Characterized: finite Israel-type layers, strictly violating (A within slice)	Peak ~ ε⁻¹ (smeared δ), integrated stress converges to ≈1.06 A²-units, negative at every ε; the a-priori dipole-layer hypothesis REFUTED by the measured scaling and recorded. Third independent EC failure of the class; S45 verdict unchanged and sharpened.

Session 56 addendum — Natário 2002 reproduced exactly (Task 3.8)

Battery: verification/test_natario2002_reproduction.py (10/10, 73 s); record NATARIO2002_EVALUATION.md.

Item	Impact	Detail
Natário 2002 Sections 1–2 (class formalism, Theorem 1.7, zero-expansion construction, ρ formula)	B → A within slice — reproduced exactly	Symbolic-exact for arbitrary profile f (construction, six K_ij, tr K = 0, Hamiltonian-constraint ρ); full 4D Einstein tensor with time-dependent v(t) at 40-digit precision (worst 7e-44); Prop 1.3 geodesics (6e-42). ρ carries no v̇ — instantaneous law (S49-parallel), so ρ_E ≤ 0 along any rigid spin-up.
Zero-expansion wall characterization (new; not in the paper)	A within slice	WEC violated at 100% of wall points incl. axis (no marginal directions); Hawking–Ellis Type IV at 98.4% of wall (v=0.1) / 77.4% (v=1) — flux ∝ v dominates density ∝ v² as v→0; exact v² scaling of ρ_E (4.000000000000); wall ergo-band sup‖X‖/v = 5.53 at (R₀,σ)=(5,4) ⟹ v* ≈ 0.181, v* ~ 4/(R₀σ).
Rodal 2025 Table-2 Natário/Alcubierre comparison ratios	B — externally anchored two-sidedly	Our exact pipeline reproduces his plot-derived ratios at matched (5,4): ρ ratio 64.9 (his ≈67), NEC ratio 65.7 (his ≈60); Type IV pockets confirmed for both families. Anchors our machinery against an external computation AND upgrades confidence in RODAL2025_EVALUATION's "Natário is unusually bad" reading.
Natário 2002 Section 3 (superluminal optics: horizon, refraction, infinite blueshift)	Not graded — not reproduced	Outside the subluminal slice; noted as geometric-optics precursor of FLB 2009 (fence crack #1). Reopens with the 2E.1 superluminal-transition leg.

Session 57 addendum — Phase 4 no-go confrontation (overnight autonomous)

Battery: verification/test_nogo_confrontation.py (12/12, 92 s); record NOGO_CONFRONTATION.md.

Item	Impact	Detail
SSV 2021 algebraic spine (Eqs 4.3–4.5, 4.7, 4.12/4.13)	A within slice — verified exact	Generic-flow symbolic + two concrete-flow full-4D Einstein cross-checks (residuals ~1e-18). Their Eq 4.8 curl-curl flux form is opposite-signed vs their own Eq 4.7 (sign slip recorded; 4.7 is what the argument uses).
Alcubierre–Lobo NEC identity G_zz = 3 G_nn	B → A, and UPGRADED to a class identity	Published as an explicit computation for the Alcubierre drive with "no obvious geometrical reason"; certified here symbolically exact for ANY z-directed flow including time-dependent. Consequence ρ + T_zz = 4ρ ≤ 0 is identity-level for the whole family.
SSV NEC/SEC global theorems	Remain B (premises + instance A)	Not re-proven in generality; the §7.4 restoration argument instantiated end-to-end on the canonical profile — NEC obligation violated on both wall crossings (+6.98e-2), SEC +2.76e-2; PF passage phenomenology reproduced to 1e-6.
PF 1997 classical chain (Eqs 5, 7, 8, 16, 26, 28)	A within slice — verified exact	Eq 8 == repo identity 1; Δ–σ relation certified at 30 digits (tanh double-angle defeats CAS simplify); magnitudes at stated OoM (their g-vs-L_P conversion carries ~2× rounding).
Ford–Roman QI premise (PF Eq 9)	Remains B — external, scope contested	Krasnikov 2003 loosening stands (Slice 4); our classical no-gos are QI-independent.
Everett–Roman 1997 Krasnikov algebra + causality bookkeeping	A within slice — verified exact	Single structure: no CTC (t_E = Dδ > 0). Two opposite superluminal structures: loop closes (antitelephone pattern). Everett 1996 itself stays B (no local copy; mechanism-carried).
Hiscock 1997 2D chain	A within slice — verified exact	Static form, R_2D = −A″, anomaly+conservation RSET, profile-independent near-horizon divergence, T_H. Subluminal gating is an identity: A ≥ 1−v₀² > 0. The 2D trace-anomaly METHOD is standard external (B); his flagged 4D question remains open.
Natário off-axis ergo-band vs axis reductions	New observation: geometry A; semiclassical implication C (flagged)	Axis sup‖X‖/v = 1.000 exactly vs equator 5.53 at (5,4) ⟹ first stationary-limit surface off-axis at v* ≈ 0.181 — invisible to Hiscock-style 2D reductions; a concrete sub-luminal target for 2E.3 / fence crack #1.

FilesExpand file tree

TRUST_AUDIT.md

Latest commit

History

TRUST_AUDIT.md

File metadata and controls

Trust Audit — What We Derived vs. What We Accepted

Result 1: Static Fuchs-class spherical shell satisfies DEC with $\Delta_{\min}/R = \kappa,\beta/C$

Result 2: No classical mechanism accelerates a Fuchs-class shell in vacuum to $\Delta v \sim v_{\rm warp}$

Result 3: Krasnikov tube classical wall has $\rho_p^{\min} \propto -\eta/\epsilon^2$, WEC fails for any $\eta &gt; 0$

Slice 4b extension (Task 2A.13b, Session 23, 2026-05-12)

Result 4: Rodal 2025 evaluation conclusions

Result 5: Composite "no classical positive-matter warp drive is simultaneously useful, accelerable, and DEC-compatible"

Load-bearing dependencies (the "if these are wrong, the project is wrong" list)

Concrete verification roadmap

What would NOT change

TL;DR

Session 16 addendum � codimension-counting law (k=0,1,2)

Session 26 addendum � nested concentric shells (Phase 3.3 item 4)

Session 27 addendum � non-spherical / oblate axisymmetric shells (Phase 3.3 item 5; Phase 3.3 fully closed)

Session 28 addendum — Phase 3.3+ Step 1 (Fuchs §6 radial-profile optimization): Cartesian-objective result KILLED

Session 29 addendum — Phase 3.3+ Step 1 radial-frame redo: NEGATIVE + an OPEN cross-representation hurdle

Session 30 addendum — Prong B adjudicates the hurdle; trust grades resolved

Session 31 addendum — Phase 3.3+ Step 2 (anisotropic) NEGATIVE; Phase 3.3+ fully closed

Session 32 addendum — Task 3.10 certified minimal-mass map (positive quantitative closure)

Session 33 addendum — Task 2D.11 Phase 3 (FH-form multi-mode A) NEGATIVE; Task 2D.11 fully closed

Session 34 addendum — Task 2B.8: Path 2B closed as a physical mechanism

Session 35 addendum — pre-Phase-2E deep audit + remediation: grade corrections and disclosures

Session 36 addendum — Slice-1 (2C.1) frame-projection bug: grade corrections

Session 37 addendum — TRUST_AUDIT #5 CLOSED (C → B): GW-recoil SXS anchor verified conservative

Session 38 addendum — 2D.11 evidence re-based (W3 kill-test): verdict upheld; Session-33 strict-pass gate superseded

Session 39 addendum — nested-shell (Phase 3.3 sub-item 4) REVERSED: first verdict flip of the audit programme

Session 40 addendum — 2C.2 hybrid-wall coverage kill-test (W2): negative strengthened

Session 41 addendum — Δ-ladder certified (W4 closed); Session-19 κ statistics re-based; Block 2 complete

Session 42 addendum — Task 2D.5e: closed-form FH pressures; the strict-pass record re-scoped to its evaluation box

Session 44 addendum — 2E.4 Π-exponent axis closed NEGATIVE (A within slice)

Session 45 addendum — Lentz 2020 closure upgraded from analogy (C) to class-level certification (A within slice)

Session 46 addendum — Garattini–Zatrimaylov 2025: literature grade B → A within slice (reproduced exactly); sharpenings recorded

Session 47 addendum — nested minimal-mass map: canonical floor re-based; S39 reversal quantified as a mass lever

Session 48 addendum — graded-wall map NEGATIVE (A within slice); S47 mechanism reattributed

Session 49 addendum — spin-up margin surface (2E.1 first leg): A within slice

Session 50 addendum — 2E.4 closed: far-field coefficients exact; topology + joint axes negative

Session 51 addendum — f(R) matter evaluator certified; quadratic-f(R) first physics

Session 52 addendum — 2E.2 closed: α-map uniform negative; designer-f pointwise LP theorem

Session 53 addendum — standoff axis retired (A within slice)

Sessions 54–55 addendum — final optional residues retired

Session 56 addendum — Natário 2002 reproduced exactly (Task 3.8)

Session 57 addendum — Phase 4 no-go confrontation (overnight autonomous)

Result 3: Krasnikov tube classical wall has $\rho_p^{\min} \propto -\eta/\epsilon^2$, WEC fails for any $\eta > 0$