d6g8k5htny-coder · Oct 2, 2026 · Oct 2, 2026 · Oct 2, 2026
@@ -0,0 +1,237 @@
+# Collisions below the pin scale have vanishing weighted mass
+
+Object OA-SUBPIN-COLLISION-20261002-v1. Author-side conditional corollary,
+OpenAI/Codex, 2 October 2026. Scientific effect NONE. This is a deduction from
+the exact imported results below, not a new acceptance of their proofs.
+
+## 1. Observable and source boundary
+
+Fix a dimension d >= 2 and a torus side L > 0. Use the centered variance-one
+periodized Gaussian field, the pin pair M = -ru/2, S = ru/2 with heights
+b and b-kr^3, k > 0, and both gradients zero. Q_r is the canonical regression
+on these pins. Retain the original typed determinant weight and FULL normalizer
+
+    W_r = F_d(H_M) F_(d-1)(H_S),
+    Z_r = E_Qr W_r,             P_r = (W_r/Z_r) Q_r.
+
+F_j is the absolute Hessian determinant times its index-j indicator. Let N_r
+count all additional critical points in the open window (b-kr^3,b), excluding
+M and S. All distances below are torus geodesic distances. The count is finite
+almost surely under the imported regularity and moment statements.
+
+Let epsilon:(0,r_*) -> (0,infinity) be deterministic and Borel measurable, with
+epsilon(r) -> 0. Define the ORDERED distinct-pair count
+
+    C_r = sum_(x != y, additional window critical points)
+                         1{dist_T(x,y) <= r epsilon(r)}.          (1)
+
+There is no adjacency, persistence or gradient-connection condition on x,y.
+The endpoints of (1) are additional witnesses, not the original pins.
+
+All source identities and exact consumed portions are in SOURCES.json, at
+Math- commit a808ba25de9db22110db8b7555a64852f38e031d:
+
+- [TS] TWO_SCALE_LAW.md, Theorems T and L, (T5), (T7), (L1), (L3).
+  Its complete configuration-space limit and ordered-pair law are needed;
+  convergence of the one-point intensity alone would not suffice.
+- [SC] spectral_cluster_closure/PROOF.md, as consumed by TS: finite nonempty
+  cubic measure, pointwise simple roots, and the first/count-moment bounds.
+- [C6] c6_palm_route/PROOF.md, Theorem Q: the original weighted factorial
+  moments of orders q >= 2, uniform on compact b,k and frames. The uniform
+  first-moment upper bound is [DL](1.7), explicitly used in the proof of C6
+  Section 7, Corollary P. With that first moment, the Stirling expansion gives
+  E_Pr N_r^q <= C_q r^3 for every positive integer q on such compact sets.
+- [DL] d5_dimension_lift/PROOF.md, Theorem G_d (1.7), with its stated
+  dimension-lift/regional dependencies retained. This is the compact-uniform
+  first-moment interface; SC's pointwise statement alone is not substituted.
+- [P] the lifetime parent, read with its required reconciliation/errata and
+  replacement Section 9. We use its full normalizer Z_r/r^2 -> z_0 > 0,
+  its marked Kac-Rice identity, and (10.2)-(11.2), including the compact-uniform
+  bound on A_r. The original parent Section 9 is not an unrepaired premise.
+
+TS has a source-bound nonauthor acceptance in review5360227991 and its current
+REVIEW_RECORD.md. That acceptance remains conditional on the exact SC/CUB/RM/C6/P
+interfaces stated in TS. We neither remove those conditions nor re-review them.
+No conclusion here is conditional on the open PR242, PR243 or PR244.
+
+## 2. The sub-pin-scale theorem
+
+**Theorem D.** For every fixed b, k > 0, frame, d and L, and every epsilon above,
+for each fixed nonnegative integer p,
+
+    E_Pr [N_r^p C_r] = o(r^3),
+    P_r(C_r > 0) = o(r^3),
+    E_Qr [W_r N_r^p C_r] = o(r^5).                         (2)
+
+N_r^0 is defined to be 1, including at N_r = 0. The little-o is pointwise in
+the fixed marks. No uniform little-o, algebraic rate, numerical constant,
+growing dimension/volume, or k -> 0 assertion is supplied.
+
+**Proof: remove pairs outside the microscopic component.** Choose a deterministic
+cutoff delta_r = sqrt(r) for small r; only delta_r -> 0 and delta_r/r -> infinity
+are used. Let N_in count the points at distance < delta_r from the pin midpoint.
+TS(L1), with q = 2, gives
+
+    E_Pr [(N_r)_2 - (N_in)_2] = o(r^3).                   (3)
+
+Every ordered pair not wholly inside that ball, including close pairs far from
+the pins, is counted in (3). Thus no near-diagonal covariance bound uniform in
+the midpoint location is being silently assumed.
+
+**Proof: the microscopic limit puts no mass on coincident spatial positions.**
+For the pairs wholly inside the delta_r ball, let nu_r be r^-3 times their
+expected ordered-pair measure on
+
+    E x E,        E = R^d x (0,1) x {0,...,d},
+
+using positions x/r, downward normalized height, and index. TS(L3) gives weak
+convergence of FINITE measures nu_r -> nu, including convergence of total mass.
+The limiting measure is the integral, on two-root cubic configurations, of
+the two ordered atoms (z_1,z_2) and (z_2,z_1) against the finite nonempty
+spectral measure. The cubic roots have distinct spatial positions almost
+everywhere on this measure. Therefore the spatial coincidence diagonal D has
+nu(D) = 0. This is stronger than finite-r simplicity alone.
+
+For eta > 0 define the bounded continuous cutoff on E x E
+
+    phi_eta(z,z') = min(1, max(0, 2 - |pos(z)-pos(z')|/eta)).       (4)
+
+It equals 1 when the distance is <= eta and vanishes at distance >= 2 eta.
+Eventually epsilon(r) <= eta. Also 2 delta_r < L/2, so the torus distance
+between any two points in the delta_r ball equals the Euclidean distance
+between their chosen local lifts. Consequently
+
+    limsup_(r->0) r^-3 E_Pr [C_r with both points inside]
+        <= lim_(r->0) integral phi_eta dnu_r
+         = integral phi_eta dnu.                              (5)
+
+As eta decreases to zero, phi_eta tends to 1_D. Finiteness of nu and dominated
+convergence give integral phi_eta dnu -> 0. Combining (3) and (5) proves
+E_Pr C_r = o(r^3). The order of limits is r -> 0 first for fixed eta, then
+eta -> 0. No finite-r density estimate, spatial total variation, or choice of
+root ordering is needed.
+
+**Proof: multiplicities and the original numerator.** C_r is either zero or
+at least two, since both orientations of a close pair are included. Hence
+
+    P_r(C_r > 0) <= E_Pr C_r/2 = o(r^3).                  (6)
+
+Pathwise, N_r^p C_r <= N_r^(p+2) 1{C_r>0}. Cauchy-Schwarz and the imported
+ordinary moment of order 2p+4 give
+
+    r^-3 E_Pr[N_r^p C_r]
+      <= (r^-3 E_Pr N_r^(2p+4))^(1/2)
+                                      (r^-3 P_r(C_r>0))^(1/2) -> 0.  (7)
+
+This is why an event probability alone is not substituted for counted mass.
+Finally the definition of P_r is an exact identity:
+
+    E_Qr[W_r N_r^p C_r] = Z_r E_Pr[N_r^p C_r].            (8)
+
+Since Z_r/r^2 -> z_0, (8) proves the o(r^5) numerator. There is exactly one
+endpoint determinant weight and one full normalizer. QED.
+
+## 3. Compact-mark lifetime consequence in the actual intensity units
+
+Fix compact intervals B in R and K = [k_-,k_+] with k_- > 0, a fixed d,L,
+and all directions u in S^(d-1). Define
+
+    m_p(r,b,k,u) = E_Pr[N_r^p C_r].
+
+The mark is jointly Borel: on a compact spatial domain away from the pin and
+pair diagonals, critical-point counts with the Borel distance/height indicator
+are measurable under the canonical regression kernels; exhaust these domains.
+The inherited weighted marked Kac-Rice identity extends by monotone truncation
+to this nonnegative count mark. Its integrability follows from C6's ordinary
+moment bound. This uses the repaired [P] Section 9, not continuity of C_r.
+
+Per unit midpoint volume, the candidate intensity carrying this multiplicity is
+
+    d beta_p = r A_r(b,k,u) m_p(r,b,k,u) dr db dk d sigma(u),       (9)
+    A_r = 12 pi_r (Z_r/r^2).
+
+Thus every pin density, spatial/height Jacobian and original determinant
+normalization is already present in (9). sigma is ordinary spherical surface
+area, not normalized Haar probability. The pin roles order the maximum/saddle
+pair; there is no new factor 1/2. C_r itself counts ordered witness pairs.
+
+**Corollary E.** Under the exact height-gap map ell = k r^3, beta_p has a
+Lebesgue density satisfying
+
+    rho_p(ell) = o(ell^(2/3)),
+    beta_p{0 < ell <= t} = o(t^(5/3)).                     (10)
+
+This statement concerns the collision-marked candidate measure (9). Its
+submeasure obtained by inserting the actual global elder indicator satisfies
+the same estimates. It is not an identification of another proxy population
+with the persistence bars, nor a claim that all selection failures satisfy (10).
+
+**Proof.** For ell < k_- r_*^3, put R = (ell/k)^(1/3). Tonelli and the exact
+one-variable change of variables give the density version
+
+    rho_p(ell) = ell^(2/3)/3 * integral_(B x K x S^(d-1))
+               k^(-5/3) A_R(b,k,u) [m_p(R,b,k,u)/R^3]
+                                                db dk d sigma(u).       (11)
+
+Indeed r dr/dell = k^(-2/3) ell^(-1/3)/3 and R^3 = ell/k.
+For every fixed b,k,u the bracket in (11) tends to zero by Theorem D.
+It is bounded uniformly on the compact mark domain because
+m_p <= E_Pr N_r^(p+2) <= C r^3 there. A_R is also uniformly bounded by [P].
+The remaining k^(-5/3) factor is integrable because k >= k_- and the domain
+has finite measure. Dominated convergence proves the density assertion.
+Integrating its absolute nonnegative bound gives the cumulative assertion:
+for any a > 0, rho_p(ell) <= a ell^(2/3) for all sufficiently small ell for
+this density version, so beta_p(0,t] <= (3a/5)t^(5/3).
+Any other version obeys the corresponding almost-everywhere bound. QED.
+
+Pointwise convergence in the marks plus this uniform integrable majorant is
+enough; no compact-uniform little-o is being imported. The Borel epsilon is
+the same function of r across the mark domain and need not be monotone.
+
+## 4. What the hypotheses prevent
+
+These are abstract marked-count examples, not counterexamples to the Gaussian
+field. Their purpose is to make the logical boundaries testable.
+
+1. **Moments alone do not remove subscale collisions.** With probability r^3
+   place two witnesses at scaled positions 2 and 2+r, otherwise none. Then
+   every fixed moment is 2^q r^3, but for epsilon(r)=sqrt(r), E C_r=2r^3.
+   The limiting pair measure charges the diagonal. The simple limiting roots
+   in TS(L3), not merely the global factorial bound, exclude this example.
+2. **No algebraic rate follows from weak convergence.** In a rare event of
+   probability r^3, place the two scaled points at 2 and 3 with probability
+   1-a_r, and at 2 and 2+r with probability a_r, where a_r=1/log(1/r)
+   for r<exp(-1). The finite pair measure converges to the simple unit-separated
+   pair. Nevertheless E C_r=2r^3/log(1/r) for epsilon(r)=sqrt(r). This is
+   o(r^3) and is not O(r^(3+gamma)) for any gamma>0.
+3. **An event estimate alone does not control multiplicity.** Let r_n=2^-n,
+   N_n=r_n^-1 with probability r_n^5 and zero otherwise, and cluster those
+   witnesses within the distance cutoff. Then P(C_n>0)=r_n^5=o(r_n^3),
+   but E C_n = r_n^3(1-r_n), which is not o(r_n^3). Its fourth count moment
+   is r_n, violating the moment hypothesis used in (7).
+4. **Scale-r doublets remain.** Replacing epsilon(r)->0 by a fixed positive
+   cutoff can retain positive limiting pair mass. TS's positive two-root
+   coefficient and all consequences about non-singleton clusters remain intact.
+
+The exact controls check finite count inequalities, these examples' rational
+arithmetic and the coarea exponents. They do not prove weak convergence,
+Gaussian conditioning, the imported parent theorems, or mathematical acceptance.
+
+## 5. Relation to the completion directive
+
+The earlier proposed compact-mark O(r^3) task was already supplied by C6
+pointwise in b,k and frame; [P](11.2) already uses height-disintegrated units.
+The missing implication in a generic marginal example cannot be attributed to
+these sources. Likewise the existing globally elder-marked endpoint route
+identifies actual finite bars at its own conditional scope; it does not require
+auxiliary-witness uniqueness. This note leaves that route and its review
+requirements unchanged. It supplies the stricter sub-pin-scale conclusion (2).
+
+This result removes subscale extra-witness collisions from a compact-mark
+ell^(2/3) coefficient in any argument whose error is explicitly dominated by
+the measure (9). It does not prove that every bad geometric configuration has
+that domination. It supplies no O(ell^(3/4)) remainder, no uniform k->0 tail,
+no rate in an expanding annulus, and no resolution of PR242 Conjecture 7.
+It does not concern a witness colliding with a pin, shared endpoints between
+different candidate pairs, or the whole class of scale-r doublets. No status,
+historical predicate, normalizer, scientific gate or independence credit changes.
@@ -0,0 +1,30 @@
+# Witness collisions below the pin scale
+
+Pairs of additional critical points can occur with probability of order r^3
+when their separation is comparable to the pin distance r. This conditional
+corollary shows that pairs separated by o(r) carry only o(r^3) weighted mass.
+On compact positive gap marks, their collision-marked lifetime density is
+o(ell^(2/3)). The original determinant weight and full normalizer are retained.
+
+Read [the proof and boundaries](PROOF.md), [the exact source map](SOURCES.json),
+and [finite negative controls](controls.py). This is a new deduction from the
+reviewed two-scale pair law, not a new Gaussian collision calculation or a
+whole-project completion certificate. No convergence rate is claimed.
+
+From the repository root, run:
+
+```sh
+python3 -B -S frontiers/subpin_collision_20261002/controls.py
+python3 -B -O -S frontiers/subpin_collision_20261002/controls.py
+```
+
+The output should match [RESULTS.json](RESULTS.json). The script checks exact
+algebra and logical counterexamples; the proof's analytic imports require their
+own source-bound reviews. Hosted downstream/formal checks have a different scope.
+
+Author-side OpenAI/Codex work, with root and named analytical subagents.
+Dylan Roy delegated the work; personal reading remains PENDING.
+Organizational independence 0; scientific effect NONE.
+
+[Public mathematical reading map](../../PUBLIC_READING_MAP.md)
+· [Two-scale configuration source](../two_scale_cluster_geometry_20260929/TWO_SCALE_LAW.md)
@@ -0,0 +1,51 @@
+{
+  "coarea": {
+    "cumulative_power": "5/3",
+    "gap_power": "-5/3",
+    "lifetime_power": "2/3",
+    "radial_power": "4"
+  },
+  "diagonal_mass_ratios": [
+    "2",
+    "2",
+    "2",
+    "2",
+    "2",
+    "2",
+    "2",
+    "2",
+    "2"
+  ],
+  "not_certified": [
+    "analytic limits",
+    "Gaussian parent proofs",
+    "uniform little-o",
+    "algebraic rate",
+    "scientific acceptance"
+  ],
+  "rare_high_multiplicity_ratios": [
+    "3/4",
+    "7/8",
+    "15/16",
+    "31/32",
+    "63/64",
+    "127/128",
+    "255/256",
+    "511/512",
+    "1023/1024"
+  ],
+  "simple_witness_graphs": 33868,
+  "slow_convergence_ratios": [
+    "1",
+    "2/3",
+    "1/2",
+    "2/5",
+    "1/3",
+    "2/7",
+    "1/4",
+    "2/9",
+    "1/5"
+  ],
+  "source_files_verified": 13,
+  "status": "PASS_EXACT_FINITE_CONTROLS_ONLY"
+}