The resolved theory

Bounded degree plus two-sided isoperimetry gives a doubling measure and a precompact family of continuum limits — unconditionally. If the limit is smooth and a bounded-degree recovery sequence exists, discrete Ollivier curvature converges to Ricci, the discrete action to Einstein–Hilbert, and general relativity is recovered. The recovery sequence now exists (gravity, Theorem C4; matter, Theorem 05.D). The one cited input is a shape theorem for bounded-degree random geometric graphs — a published theorem, not an open problem: Yao–Chen–Guo (J. Appl. Probab. 48 (2011) 154–172) for the continuum instance both sectors consume, Cox–Durrett (1981) for the lattice, with the D-NN transfer proved in C4.1b. Discharged in CITATIONS.md §SD.

It cannot close by the mechanism the theory proposed. The causal 1-Wasserstein cost is convex rather than a metric, so no contraction — hence no synthetic timelike Ricci lower bound — follows from a discrete timelike Ollivier bound. This is a signature fact, proved, independent of degree. The Lorentzian action is reachable only by Wick rotation.

A QSN is a countable set N with a relation ≺ whose transitive closure is a partial order, such that the interval [n,m] is finite for all n,m, and the edge set E = {(n,m) : n ≺ m, no k with n ≺ k ≺ m} gives a locally finite graph.

A function T : E → [0, ∞). The tension is edge data, posited rather than derived — a fact that will matter (Theorem 03.A).

For an edge e = (n,m), with μ_n, μ_m the out-neighbour measures and d_G the graph distance, K(e) = 1 − W₁(μ_n, μ_m) / d_G(n,m). With the sink convention, K(e) ∈ [−2, 1].

d_eff = infimum over paths γ of Σ_e∈γ T(e)⁻¹.

E[Q] = Σ_e∈E T(e) K(e), with gradient-flow dynamics.

Each edge carries a self-adjoint H_e with discrete spectrum, gap Δ(e) > 0, and

No derivation from L1–L5 is known, and the v2 attempt (KMS, modular flow) is permanently withdrawn: it confused a property of a state with a property of a spectrum, and modular flow is trivial in the trace state of a Type II₁ factor.

Under L1, L2, L4 with 0 < T_min ≤ T(e) ≤ T_max, the identity map is a bi-Lipschitz equivalence between (N, d_G) and (N, d_eff) with constant T_max/T_min.

ProofAlong any path, Σ T⁻¹ lies between T_max⁻¹·len(γ) and T_min⁻¹·len(γ). Take infima over paths. ∎

Under L1, L2 with T_max/T_min bounded, H1′ (degree bound D, bilateral tension bounds) and H4′ (two-sided isoperimetry, fixed isoperimetric dimension N₀ ≥ 3), there is C_dbl = C_dbl(D, N₀, T_max/T_min) with μ(B(x,2r)) ≤ C_dbl μ(B(x,r)) for all x, r.

ProofThe five-step proof uses only H4′ and H1′. No curvature hypothesis enters: Step D states "H2 is not used." Classical inputs are Coulhon–Grigor'yan (Nash inequality), Carlen–Kusuoka–Stroock (Nash-to-ultracontractivity), and Davies–Delmotte (Gaussian bounds), each applied with bounded geometry, not curvature. ∎

Correction to v3. v3's dependency matrix listed this lemma as consuming H2 (K ≥ 0). That is wrong — the proof does not use H2. The error mattered: it made the strongest unconditional result look curvature-dependent and hid the existence of U3.

Let (Q_n) be QSNs satisfying L1–L5, H1′ with fixed D and fixed T_max/T_min, H3 (finite d_eff-diameter after rescaling) and H4′ with fixed N₀, and |N_n| → ∞. Then the rescaled metric measure spaces (N_n, ε_nd_G, μ_n) are precompact in the measured Gromov–Hausdorff topology, and any limit is a compact doubling metric measure space.

ProofU2 gives doubling with a constant independent of n; H3 gives uniformly bounded diameter; Gromov precompactness (with the measured refinement of Fukaya and Cheeger–Colding) gives mGH precompactness. ∎

What it is not. U3 asserts the existence of a limit. It asserts nothing about what the limit is — it need not be a manifold, need not have the right dimension, and carries no curvature bound. This is the honest unconditional continuum limit, and it is stronger than anything v3 stated unconditionally, because it needs no curvature hypothesis.

On a finite QSN with the L5 energy, the Wetterich equation admits an exact solution; no truncation is involved. The one place in the renormalisation sector where the flow is exact rather than truncated.

The mGH limit (X, d_X, m) of Theorem U3 is a smooth N₀-manifold with a C³ Riemannian metric g, and m is a positive multiple of the Riemannian volume. This hypothesis is false as stated — see §4.

Assume H10, and assume the sequence carries a tension tensor isotropic in the sense H12, with the bias hypothesis H13 and the recovery hypothesis H14. Then the rescaled Ollivier curvatures converge pointwise to the Ricci tensor:

along edges whose endpoints converge to a common point x with unit direction v_e.

ProofOllivier's pointwise asymptotic expansion on a Riemannian manifold (Proposition 6; Example 7 for the normalised ball, coefficient 1/(2(N₀+2))), combined with the mGH convergence of U3, which supplies that the discrete neighbourhoods converge to the manifold neighbourhoods. H12, H13, H14 supply isotropy, the bias bound, and the recovery sequence. The citation carries the continuum step only — the discrete-minus-continuum discrepancy's coefficient is H13(b), which Classical E fixes in order but not in form (CITATIONS.md §E, stop 1). ∎

Correction to v3. v3 attempted to pass an Ollivier bound to the limit as a synthetic RCD condition, and correctly proved that fails. R1 does not attempt that. It uses Ollivier curvature only in its valid role: a pointwise expansion on a smooth target. The hypothesis that makes this work is H10, not a synthetic curvature condition.

Under H10, H12, H13, H14, the rescaled QSN action converges:

with α_d = 2(N₀+2)/(16πG) fixed by R1. The first variation gives the Einstein equations with matter source.

Where H12 does its work. The average of Ric(v,v) over directions equals R/N₀ only if the edges sample all directions equally. A bounded-degree graph with an anisotropic local law samples a finite set of directions and converges to a direction-dependent — frame-dependent — functional. This is the same obstruction that makes the Lorentzian sector fail, and it is why H12 appears in both R2 and L4. In the Riemannian sector H12 can be achieved by an isotropic local law; in the Lorentzian sector with positive tensions it cannot. That asymmetry is the difference between the two signatures.

Conditional on H10, H12, H13, H14, the Riemannian sector reproduces general relativity in the continuum limit, with the cosmological constant Λ_b a boundary condition of the flow.

Assume the limit satisfies RCD(K, N₀). Then X decomposes as X = R ⊔ S with R open dense and a topological N₀-manifold, S closed with Hausdorff dimension ≤ N₀ − 2, and every point of R has a neighbourhood bi-Hölder homeomorphic to a Euclidean ball.

Why R4 does not close the gap. H-BE is not derivable from L1–L6. Three independent obstructions: (a) no theorem converts a graph Ollivier bound to a graph Bakry–Émery estimate — the notions are incomparable on general graphs; (b) Mosco convergence of Dirichlet forms controls integrated energies, not the carré du champ; (c) no finite-dimensional CD(K,N) follows from Ollivier curvature — the weighted manifold (M^N₀, g, e^−Vvol) with Ric + Hess V ≥ K and non-constant V is a counterexample. H-BE is assumed, and R4 is strictly weaker than R1–R3: it gives a manifold structure on the regular set only, without smoothness.

There exists a sequence of QSNs satisfying L1, L2, H1′ (bounded degree), H3, H4′, and H12 whose mGH limit is a metric space that is a topological 2-manifold but not a smooth Riemannian manifold: a flat cone.

ConstructionThe discrete cone over a 5-cycle: the triangular lattice with one vertex removed and its five neighbours identified pairwise. Degree 6 away from the apex, degree 5 at it; both bounded. The six triangular directions satisfy Σ v̂ ⊗ v̂ = 3·I (a tight frame). Away from the apex the cone is locally Euclidean, so H12 holds exactly there; at the apex the five remaining directions give a tensor not proportional to I — but the apex is one point of measure zero, so the integral H12 holds. The rescaled distance converges to the flat cone metric with total angle 5π/3, deficit angle π/3. The apex has no Euclidean neighbourhood: a small circle of radius r has circumference (5π/3)r, not 2πr. Hence not a Riemannian manifold, while homeomorphic to ℝ². ∎

A real singularity. The scalar curvature is a distribution: R = (2π/3)·δ_apex. The Einstein–Hilbert action ∫R dvol = 2π/3 is finite and nonzero, concentrated at one point — the same kind of defect as a cosmic string, not a coordinate artifact.

A singular set of Hausdorff codimension ≥ 2 has d-dimensional measure zero, and the number of edges meeting it within a ball of radius r scales as r^d−2 against r^d edges in the ball, so its contribution to the H12 sum is a fraction r⁻² → 0. H12 holds even though the limit is not smooth. The theory's only isotropy hypothesis is blind to exactly the defects that make the limit non-smooth.

Let (X, d) be the mGH limit of a QSN sequence satisfying L1–L5, H1′, H3, H4′, H12, and suppose it satisfies RCD(K, N). Then exactly one of: (a) X is a smooth Riemannian manifold; or (b) X has a non-empty singular set of Hausdorff codimension ≥ 2. The bound is sharp — codimension 2 is attained (02.A).

Why this ends O1. O1 asks for a proof that (a) holds. 02.A shows (b) is consistent with every hypothesis the theory has. So O1 as stated is false, not merely open. The theory has no axiom that selects (a).

v3 asserted that bounded degree forbids isotropy. It does not. Bounded degree forbids isotropy only for a deterministic finite-stencil local law. The square lattice satisfies H12 in its strongest form — tension tensor M = Σ_{±e₁,±e₂} v̂ ⊗ v̂ = 2·I, exactly isotropic at every node — and still has an ℓ¹ limit whose unit ball is the square {|x|+|y| ≤ 1}. A finite stencil's asymptotic cone is a polytope, never round. The step from "a fixed finite stencil is not isotropic" to "bounded degree forbids isotropy" silently assumed the local law is a fixed finite stencil.

The emergent norm at p is the Minkowski functional of K_p = conv(supp ν_p). It is Riemannian iff K_p is round, iff supp ν_p is not confined to finitely many directions. H12 (second-moment isotropy) is strictly weaker: a square and a circle inscribed in it share the second moment 2I but have different hulls. The correct hypothesis is H12′ — ν_p is the uniform measure on the sphere S^d−1.

There exists a sequence satisfying L1–L5, H1′, H3, H4′, H12′, H13 whose mGH limit is isometric to a prescribed (M, g). The construction: a Poisson point process of intensity ε_n^−d, joined to its D nearest neighbours, with tension T(e) = θ·d_g(p,q). H12′ is proved here, not assumed: the rescaled neighbourhood is O(d)-invariant, and the union of O(d)-invariant D-tuples over a metric neighbourhood fills the sphere as n → ∞.

The one cited step, now named — and it is a published theorem. The convergence ε_n d_G → c·d_g is the chemical-distance / first-passage-percolation theorem for random geometric graphs: (i) linear growth of graph distance (subadditive ergodic theorem); (ii) the time constant μ(θ) is direction-independent because the local law is O(d)-invariant, and the only O(d)-invariant functions on the sphere are constants. Isotropy is exactly what upgrades "a norm" to "a constant times d_g." Part (ii) is elementary and proved here. Part (i) is a named, published theorem: for the continuum Poisson geometric graph C4 uses, Yao–Chen–Guo (J. Appl. Probab. 48 (2011) 154–172) established the time constant's existence in the supercritical regime; Gouéré–Théret proved it positive and continuous, and the lattice analogue is Cox–Durrett (1981). So the cited input is a classical theorem, not an open problem — the same epistemic status as the Coulhon–Grigor'yan input to U2. What remains cited-but-not-formalized is its analytic content (a formalization gap, not a mathematical one). The planar lattice instance is machine-checked in the openai_math corpus (result 212). The transfer from the Boolean model to C4's D-nearest-neighbour graph is an exact sandwich — G_bool(r_n⁻) ⊆ G_DNN ⊆ G_bool(r_n⁺), reversing into a chemical-distance bound and forcing the time constants equal by Poisson concentration — proved in CONSTRUCTION.md C4.1b; the classical input is discharged in CITATIONS.md §SD. See sources/INDEX.md.

Let W₁^τ be the Lorentzian causal 1-Wasserstein cost, W₁^τ(ν,ν′) = sup over causal couplings π of ∫ τ dπ. Then (i) it is convex, being a supremum of affine functions over a convex constraint set; (ii) consequently no contraction estimate follows from a uniform lower bound on the timelike Ollivier curvature — the Riemannian proof of contraction needs the gluing lemma, which needs W₁ to satisfy the triangle inequality as a metric, and W₁^τ does not; (iii) consequently no synthetic timelike Ricci lower bound in the TCD sense follows.

ProofThe Riemannian contraction glues two couplings along a geodesic; the gluing step uses the triangle inequality for W₁. W₁^τ has no triangle inequality, being a cost supremum rather than an infimum over all couplings. ∎

Scope. This rules out the contraction route, not every conceivable route. What is proved is that the mechanism v3 used is unavailable — which forces the one-sided hypothesis H6⁺ and the Wick-rotation conclusion.

Any hypothesis intended to encode a timelike Ricci lower bound must be an upper bound on the timelike Ollivier curvature. The v3.0 hypothesis H6, stated as a lower bound, is inconsistent with the geometry and is withdrawn.

On a smooth Lorentzian manifold the shell Ollivier curvature satisfies

with the sign opposite to the Riemannian expansion. The inequality is one-sided; the matching equality is Conjecture E-Lor, which needs a new timelike MTW-type regularity hypothesis and is not load-bearing (the proved inequality already suffices downstream).

Provenancev3 Lemma 7.4, retained (Erratum C22), with the four-step proof in O'Neill conventions and the de Sitter sign check. Not a citation to a published theorem: the Lorentzian expansion has no published form. This is the theory's own computation, and its hypotheses — a smooth Lorentzian background, g ∈ C³, the shell radius below the injectivity radius — are those of the construction, not of an external result.

Under bounded degree H1′, the Lorentzian recovery sequence requires a tension tensor isotropic in the sense M^μν ∝ g^μν, which is impossible with positive tensions: a positive combination of causal outer products is positive semidefinite, and no positive semidefinite form is proportional to an indefinite one. Consequently the Lorentzian Einstein–Hilbert action is obtained only through Wick rotation, and the physical content at bounded valence is a frame-dependent gravity with parameters (κ_D, a_D, b_D).

The Lorentzian sector of Lach-Space, at bounded degree, is not general relativity. It is a frame-dependent gravity with three phenomenological parameters per degree bound, recovering general relativity only in the double limit of large degree and Wick rotation. This is a proved limitation.

Under H3′ and the named inputs Combes–Thomas, integral representation of the sign function, and Neuberger (discharged in CITATIONS.md §I, J, K), the overlap operator is γ₅-Hermitian, satisfies the Ginsparg–Wilson relation, is exponentially local with ξ_loc = ½ log(1 + m_W²/(2C_H²)), and has the correct index. Doubler-freedom holds under H16. Nielsen–Ninomiya is evaded because the graph is not translation-invariant.

Under the named inputs Kazhdan–Lusztig and Andersen–Paradowski, the category of tilting modules of U_q(su(3) ⊕ su(2) ⊕ u(1)) at q = e^iπ/ℓ is a UMTC with the stated S-matrix; the abelian factor is at M = 6ℓ + 1 (odd, modular); anomaly cancellation holds exactly, with all five perturbative coefficients zero and the SU(2) doublet count even.

The one instrument. The exact-arithmetic verification of this sector is the only machine-checked content. It has already caught one error (the ℓ = 5 loop counterexample that refuted a v3.0 claim). The classical inputs (Kazhdan–Lusztig, Andersen–Paradowski) are discharged in CITATIONS.md §N, O.

Modulo Reuter's truncation, the truncated Wetterich flow has a non-Gaussian fixed point with (g*, λ*) = (0.7073208810, 0.1932005072) and complex critical exponents θ = 1.4753024 ± 3.0432058i. Under the Standard-Model-shifted truncation the fixed point moves to (0.8632468, −1.2393818).

Discharge. Reuter's truncation is the named input; the fixed-point values are certified in Lean (certificates/) and the ODE recomputed in scripts/03. CITATIONS.md §N, O covers the quantum-group inputs to M2 that M3's SM-shift consumes.

The inductive limit of finite edge-mode algebras is a hyperfinite Type II₁ factor, with K₀ ≅ ℤ[1/n]. The CPW crossed-product isomorphism and edge-mode completeness remain conjectures; the Page-curve results are conditional on H9.

Discharge. The classical input is Murray–von Neumann (hyperfinite II₁ uniqueness), discharged unconditionally in CITATIONS.md §T — hyperfiniteness is automatic from the inductive-limit construction, needing no Connes injectivity.

Let {û_k} be a tight frame: D directions equally spaced by 2π/D. Then for every rotation angle θ,

ProofBoth sums are rotation-covariant, and R((D/2)I)R^T = (D/2)I, R·0 = 0. Evaluate at θ = 0. ∎ (Verified to machine precision at θ = 0, 0.37, 1.9, 2.718.)

The whole correction. H15(d) is a condition on the second moment, and a tight frame's second moment is (D/2)I for every orientation. So a graph can satisfy H15(d) at every node while its frames point in completely different directions at different nodes. Per-node isotropy constrains the shape of each node's direction set, not its orientation. The first pass treated "isotropic stencil" as "the same stencil everywhere," and that is what manufactured the D-gon.

If the stencil is held fixed, the mGH limit is the norm whose unit ball is the convex hull of the directions — a regular D-gon, not a Euclidean disk for any finite D. The roundness defect is sec(π/D) − 1 = π²/(2D²) + O(D⁻⁴), verified numerically (D = 4 → 0.414, 6 → 0.155, 10 → 0.051, 20 → 0.0125, 100 → 0.0005). This is real — but it is a statement about a fixed stencil, and nothing in H15(d) requires one. The obstruction is proved directly (scripts/05, 08); it is a term of art in the frame-orientation class, discharged in CITATIONS.md §BB (the asymptotic-cone step).

Place a tight frame at each node with iid random orientation θ(x), uniform tension, by the C4 prescription. Then H15(d) holds exactly at every node, H12′ holds, and the mGH limit is (M, g) — Riemannian, round.

Why it is not a new conditionalC4's limit proof uses only (i) linear chemical-distance growth and (ii) direction-independence of the time constant, which follows from O(d)-invariance of the local law. Input (ii) holds a fortiori here — the frame graph's local law is exactly uniform, where C4's is uniform only in the neighbourhood-union limit. Input (i) is a shape theorem of C4.1's class for a bounded-degree geometric graph. The frame graph is a different graph, so this is the same class of cited theorem with its own instance, not C4.1 verbatim. ∎

Verified (script 09). The angular profile f(φ) = mean d(0,x)/|x| is 10-fold modulated (amplitude ≈ 0.051) for the fixed stencil and flat for the random-orientation graph. Amplitude vs L: fixed 0.0556 → 0.0385 (constant, the D-gon defect); random 0.0088 → 0.0047 (shrinking, fitted exponent ≈ −0.33; pure L^−1/2 is −0.5, a D-gon is 0.0). The cited theorem and its D-NN transfer are discharged in CITATIONS.md §SD.

• H12′ — the marginal direction law is exactly uniform (θ ~ U[0,2π) shifted by fixed frame angles). Exact.
• H1′ — the frame directions realize as distinct nearest lattice vectors; degree bounded. Exact.
• Connectivity — a single giant component covering ~100% of nodes for D ≥ 4. Verified.
• H15(d) — exact for the ideal frame (05.1).

Degree, corrected. Item 01's threshold D ≥ d(d+1)/2 = 10 governs tuning tensions on generic directions. A tight frame is isotropic at uniform tension, so the count does not apply. The minimal centered frames are the regular simplex (D = d+1) and the cross-polytope (D = 2d) — in 4D, D = 5 or 8, below 10. The matter sector closes at D = 5.