SG-3 — Three generations χ=−3: full dossier — rendered package. Rendered from DOSSIER_SG3_FULL.md; frozen technical content unchanged by rendering.

SG-3 — Three generations χ=−3: full dossier

Status (binding): DERIVED-GIVEN-E. STATUS-UPGRADES:0. Frozen 13D branch dcc66f1b2685 / manifest meta a5b1e6f9d951 READ-ONLY; spin/twisted-spin bundle data 0fd19c9ae0c1; orbifold parity ledger ac4d2df3e708. Nothing in this dossier is mutated, applied, or deployed.

The one sentence to carry. The number of matter generations — a brute fact the Standard Model has input by hand for fifty years — comes out of this geometry as a rigid whole number with no dial: |χ(K₆,E)| = 3 — computed with the observed Standard-Model content E as input (given-E ≠ derivation of E). The magnitude is genuine and convention-independent; the chirality sign is not pinned by the frozen record.

Published-history note (current board status). The status language in this dossier is a frozen working audit captured on the 2026-06-27 / 2026-06-29 pre-ratification branch, preserved verbatim as the contemporaneous record of the closure work. On the current board (33 requirement-gates: all 33 RESOLVED at +0 · 0 anchored · 0 open, ratified 2026-07-08; the live /gates/ ledger + per-gate dossiers are the source of truth), SG-3 (three chiral generations (χ=−3)) is RESOLVED at +0 (DERIVED-GIVEN-anchor). The frozen per-hole work items and projected endpoints below are the discipline that produced that closure, shown openly — read them as history, not as the current grade.

Table of contents

  1. Executive summary + honest status
  2. The community gap
  3. The construction — rigorous math
  4. The insights we used
  5. Evidence & reproducibility
  6. Open gaps + closure path — the specialist work plan
  7. Honest ceiling & scope

1. Executive summary + honest status

1.1 Headline

Why are there exactly three generations of matter — three copies of the same quark-and-lepton pattern, no more and no fewer? It is one of the oldest unanswered questions in particle physics. The Standard Model does not derive the number 3; it inputs it. Experiment pins the count of light neutrino species to N_ν = 2.984 ± 0.008 (LEP/SLD), but that is a measurement, not a mechanism. No accepted theory closes this.

On the frozen 13D K₆ branch, the family count is not a fitted parameter — it is a topological index. Two complementary counting theorems agree:

This is a rigid, deformation-proof integer: no continuous modulus can move it inside the declared search category (Atiyah–Singer rigidity). It is exactly the kind of object a family count should be — a winding number, not a volume knob. The engine underneath it (the zero-weight multiplicity rule m₀(p,q) and the ℤ₆ centre congruence) regenerates from scratch, target-blind, from the actual SM hypercharges. On the program's own grading this gate ties SG-2 as the strongest in the corpus, because an integer index is more rigid than a coupling-power ratio.

1.2 The honest grade

DERIVED-GIVEN-E (direction: held; no promotion).

The qualifier given-E is load-bearing and travels with every statement in this dossier. The index χ(K₆,E) = −3 is the index of a bundle E that is the SM chiral content itself (its first Chern class / weight is read off the observed particle charges). So the result certifies that this geometry+bundle yields three families, given the particles we observe. It does not claim three is forced across every conceivable geometry (that is a universal negative, unprovable for any theory — see §7), and it emphatically does not derive E. Using χ = −3 to "force" E would be circular, because E is the input.

1.3 What this dossier establishes and what it does not

Establishes (given-E, given the selected & frozen geometry+bundle): that the family count on this branch is a rigid integer index = 3, left-handed, no surviving mirror; that the rep-theory engine producing it reproduces target-blind; that the carrier K₆ is the same SU(3) flag used for colour at SG-2 (the strongest reuse point in the program); and that the genuine magnitude claim is a confident, falsifiable bet.

Does not establish: that 3 is unique across geometries (OPEN, no witness); that the bundle/weight returning −3 is the unique admissible one (OPEN — hole R3, highest leverage); that the +2-dimension preference of K₆ over the cheaper CP² carrier is target-blind (OPEN — hole R2, the SHAPE suite's named live smuggle); that the structure is literally "spin-ℂ" (REFUTED on-disk — hole R4; the genuine object is the twisted (Spin × G_SM)/ℤ₆); that the machine certificate independently re-derives the index (it is a consistency lint — hole R6); or that the chirality sign (left- vs right-handed) is pinned (it rides an unpinned Pin⁺/Pin⁻ convention bit whose default value comes out the wrong sign in the sister gate — hole R7).

given-E ≠ derivation of E. AXIOM-CLOSED ≠ proven. selection ≠ derivation. dissolved ≠ solved.


2. The community gap

2.1 The precise open problem

The Standard Model gauge group G_SM = SU(3)_c × SU(2)_L × U(1)_Y comes with a matter sector that repeats: each generation is one copy of

Q_L = (3, 2, +1/6), u_R = (3̄, 1, −2/3), d_R = (3̄, 1, +1/3), L_L = (1, 2, −1/2), e_R = (1, 1, +1) — 15 Weyl fermions per generation (16 with a right-handed neutrino).

The pattern occurs three times, with identical gauge charges and differing only in mass. The SM Lagrangian contains the number 3 as a multiplicity input; nothing in the SM explains why it is 3 rather than 1, 2, or 4. This is the family (generation) replication problem.

2.2 The state of the art / best bound

The strongest empirical statement is the LEP/SLD measurement of the invisible Z width, which counts light, weakly-interacting neutrino species: N_ν = 2.984 ± 0.008 (cited in the corpus as the admissibility-boundary anchor; see DOSSIER_SG3 §0 and Objection 4). This is a measurement of the number of light active neutrinos — a constraint, not a derivation. It excludes a fourth light generation but does not explain why the count is 3.

On the theory side there is no accepted derivation. Past program-level attempts to force the count fall into recognizable classes, each with a known obstruction:

2.3 Why each prior attempt falls short — and where this gate stands relative to them

The community gap is not "find a number that equals 3." Almost any framework can be arranged to give 3. The real gap is derive 3 without feeding in the answer — target-blind, with the matter content not assumed. Measured against that bar:


3. The construction — rigorous math

Full published derivation: GUT.html §6.4 (Gate-4 card), §5.3 (narrative), §3.4 (topological-count machinery), Appendix E (E.1 mechanism / E.2 three generations / E.3 mirror ledger / E.6 certificate), Appendix GP.3 (the counting theorems), Appendix C2 (K₆ dossier), C4 (S¹_Y/ℤ₂ dossier), A2.2 (bundle ledger). This section is the attack-grade reconstruction of that material, not a reprint; every number is cited to a corpus file read for this dossier.

3.1 The end-to-end pipeline

The family count is generated by two index computations on the frozen geometry, summarized in the corpus as "K₆ counts the families; S¹_Y/ℤ₂ makes them chiral by projecting out the mirror copy" (CR4.6). The forward chain (§5.3, Appendix E.1; DOSSIER_SG3 §1.1):

 K₆ = SU(3)/T²  +  frozen (twisted-)spin bundle  --BWB index-->  χ(K₆,E) = −3   -->  |Index| = 3 families
 S¹_Y/ℤ₂ fold   +  frozen boundary parity ledger  --APS index-->  (n_L, n_R) = (+3, 0)  -->  mirrors removed
 E_matter (chiral actors) + P_χ projector          --zero modes--> three left-handed SM generations

The load-bearing factors, named so each can be attacked independently:

3.2 The carrier K₆ = SU(3)/T²

K₆ is the complete flag manifold of ℂ³ — the space of full flags (line ⊂ plane ⊂ ℂ³). As a homogeneous space it is SU(3) modulo its maximal torus T² = U(1) × U(1):

The reuse point (strongest in the program). K₆ = SU(3)/T² is the same factor SG-2 uses to recover colour SU(3). The carrier that carries colour is the carrier that counts families — no new geometric term is introduced for the family count (DOSSIER_SG3 §1.1, factor 1; W3). The carrier's selection is itself defended at SG-2 by abelian-isotropy uniqueness: T² is the unique purely-abelian SU(3) isotropy subgroup that gives a clean SU(3) gauge sector, and the cheaper CP² = SU(3)/U(2) over-produces gauge content at Gate 2 (DOSSIER_SG3 §4.3 firewall table, row 1: PASS — K₆ is the unique clean carrier for gauge, data-independently). Attack handle: the carrier is forced for gauge; the bundle/weight read on it for the family count is selected (residuals R2/R3).

3.3 The matter bundle and the projector

The chiral actors live in the matter bundle E_matter, the ⊗-layer object on which the chirality projector

P_χ = ½(1 + γ₅ Γ₈)

and the index operator act (Appendix C7; P_χ derived from the frozen hashes ac4d2df3e708 + 0fd19c9ae0c1; DOSSIER_SG3 §1.1 factor 4). Without E_matter neither the projector nor the index has a domain. The surviving content is exactly the SM set: the right-handed singlets (u_R, d_R, e_R and the neutrino-sector mode) arise from the conjugate sector through the chamber projectors with the same generation count. The generation module is

𝒢_gen = span{g₁, g₂, g₃}, dim = 3,

matched to the family index −3 (A1.13), and consumed downstream by the F⁺ flavor chamber (SG-8 inherits this dimension; DOSSIER_SG3 §1.1 factor 5).

3.4 Borel–Weil–Bott on K₆ and the index χ = −3

On a homogeneous Kähler space G/T, Borel–Weil–Bott computes the cohomology of a line bundle L_λ associated to a weight λ in closed form: H*(G/T, L_λ) is either zero (when λ + ρ is singular, ρ = half-sum of positive roots) or carries a single irreducible G-representation in a degree fixed by the length of the Weyl element making λ + ρ dominant. The holomorphic Euler characteristic

χ(K₆, L_λ) = Σ_i (−1)^i dim H^i(K₆, L_λ)

is therefore a closed-form, integer-valued, representation-theoretic quantity — not a numerical fit. For the frozen bundle E on K₆, this Euler characteristic returns

χ(K₆, E) = −3 (Appendix E.1; bundle ledger A2.2; GP.3; frozen R1.4 datum, hash 0fd19c9ae0c1).

The interpretation (DOSSIER_SG3 §1.1, factor 2; §1.2 genuine-content table):

Honest structural caveat (R4). Kähler manifolds are canonically spin-ℂ, and the manuscript currently calls this object the "spin-ℂ Borel–Weil–Bott index." For pure SM content that wording is REFUTED (§3.8). The numerical value |χ| = 3 survives the correction unchanged — a count is not a structure label — but the genuine forced object is the twisted (Spin × G_SM)/ℤ₆ bundle, and the BWB-on-line-bundle computation is the symbolically re-derivable leg of it.

3.5 The zero-weight multiplicity rule m₀(p,q) (the engine, target-blind)

Under the BWB machinery the relevant T²-invariant (zero-weight) subspace of an SU(3) irrep with Dynkin labels (p,q) has multiplicity given by

m₀(p,q) = min(p,q) + 1 if (p − q) ≡ 0 (mod 3), else 0

(W4; SHAPE_M0_DELTASPINC §1.1; verified target-blind via Freudenthal's recursion in the closure campaign — "m₀(p,q) re-derived from scratch via Freudenthal"). All eight tabulated values regenerate. The worked table (SHAPE_M0_DELTASPINC §1.1):

(p,q) Casimir C₂ m₀ dim deg = dim·m₀ contrib = deg·C₂
(0,0) singlet 0 1 1 1 0
(1,1) adjoint 3 2 8 16 48
(3,0) 1 10 10 (UV tower)
(2,2) 3 27 81 (UV tower)
(3,3) 4 64 256 3840 (UV)

The key target-blind output: the dominant low-lying triality-0 boson is the adjoint (1,1) with contribution 16·3 = 48; the singlet contributes 0. (The full unregulated tower sums to 4596, but that figure is a UV artifact — (3,3) alone is 3840 — removed by the heat-kernel regulator; it is discarded for the sign read. SHAPE_M0_DELTASPINC §1.1.) The point for SG-3 is not these specific weights but that the rule m₀(p,q) = min(p,q)+1 is regenerated from scratch, not reverse-engineered to any target — it passes the κ³/π anti-smuggle test (W4).

3.6 The ℤ₆ centre via the Tong congruence (target-blind, field-by-field)

The subgroup of the centre of G_SM that acts trivially on the SM matter content is computed directly from the actual SM hypercharges (REDUCTION_SPINTWIST_C §2; W5). Writing a centre element as (z₃, z₂, phase) in (ℤ₃ of SU(3)_c) × (ℤ₂ of SU(2)_L) × U(1)_Y, the kernel is exactly six elements,

ℤ₆ = {(0,0,0), (0,1,3), (1,0,4), (1,1,1), (2,0,2), (2,1,5)},

generated by ξ = ω(x) η(x) e^{2πi/6}. Equivalently, the Tong congruence

q = 6Y ≡ 3z₂ − 2z₃ (mod 6)

holds field-by-field for Q_L, u_R, d_R, L_L, e_R, and the Higgs (REDUCTION_SPINTWIST_C §2(B); Tong arXiv:1705.01853). This was computed from the SM charges, not reverse-engineered to a target (W5; passes the κ³/π anti-smuggle test).

3.7 The no-mirror fold S¹_Y/ℤ₂ and its APS index

A closed internal factor is handedness-neutral. This is corpus fact F2: on a closed manifold, left- and right-handed zero modes come in matched pairs, so the net handed count is zero — a bare circle S¹_Y mirrors every fermion, and the APS index reads (n_L, n_R) = (+3, +3) (DOSSIER_SG3 §1.1 factor 3). That is the wrong physics: the SM is chiral.

The repair is an edge (boundary). The orbifold S¹_Y/ℤ₂ folds the circle by the reflection θ ↦ −θ, with two fixed points {0, π} (freeze R1.3, hash ac4d2df3e708). The boundary parities admit one handedness and refuse its mirror. Under the frozen parity ledger the Atiyah–Patodi–Singer boundary index returns

(n_L, n_R) = (+3, 0) (Appendix E.1/E.3; A1.8; GP.3) —

three families with one surviving handedness, zero surviving mirror — i.e. chiral rather than mirror-symmetric. (Which handedness — "left" vs "right" — is the sign of the index, and that sign rides an unpinned Pin⁺/Pin⁻ convention bit whose default comes out the wrong way downstream; see §3.7-caveat and R7. The settled, convention-independent facts here are the magnitude 3 and the one-sidedness n_R = 0, not the left/right label.) A one-sided count of this kind is impossible on any closed factor — it requires the boundary. The contrast with the bare-S¹_Y control (+3, +3) proves the fold is load-bearing (W2; this is the cleanest no-go leg of the gate, and it is hand-checkable, see §5.3).

3.8 The spin-ℂ obstruction and the forced twisted-spin structure (R4)

The manuscript wording "spin-ℂ Borel–Weil–Bott index" is provably wrong for pure SM content. The computation (REDUCTION_SPINTWIST_C §2; reproduced target-blind in C:/tmp scripts cited there; SHAPE_M0_DELTASPINC §2):

Result. The global spin form forced by E's charges is spin-G_SM = (Spin × [SU(3)_c × SU(2)_L × U(1)_Y])/ℤ₆, with twist order n = 2 — not a product Spin × (gauge bundle), and not spin-ℂ. This is FORCED-GIVEN-E (it lands on E + the anomaly rulebook C_admiss + a Dai–Freed single-valuedness principle; all three falsifier conditions PASS — spin-from-E, lands-on-E, E-stays-primitive; REDUCTION_SPINTWIST_C §4). Crucially the value |χ| = 3 is unchanged by this correction — and the corpus is explicit that this is FORCED-GIVEN-E, not a forcing of E itself (T3 stays REFUTED).

3.9 What "closed" means here, and its three conditions

"Closed" for SG-3 is a rigid integer family-count mechanism under declared assumptions — strictly not a from-nothing inevitability of three, not a proof that three is unique across geometries, not a derivation of E. It is conditional on (DOSSIER_SG3 §1.2):

  1. given-E — the SM chiral content E (the bundle whose Chern class returns −3, and the hypercharge ledger fixing admissibility) is supplied as input. SG-3 does not derive E; using χ = −3 to force E is circular.
  2. given-the-selected-geometry+bundle — the carrier K₆ is forced for gauge (abelian-isotropy uniqueness), but the bundle/weight is selected inside a declared admissibility class, not proven unique.
  3. given-the-data-exclusion — the deformation that would change the index to 2 or 4 is excluded by the LEP N_ν bound (2.984 ± 0.008), i.e. by data. So the count's robustness against a fourth-family rescue is partly a data input at the class boundary, not a pure internal forcing.

The genuine, defensible content that survives this accounting:

Genuine output Mechanism
Family count = 3, a deformation-proof integer (no dial)
No surviving mirror (n_R = 0) APS one-sided boundary index on S¹_Y/ℤ₂ (F2 + the fold)
The integer is the right kind of number topology, not a tuned bundle modulus ("three by dial" fails even at 3)

The conditional / selection-in-disguise content:

Claim Honest reality
"Three families are derived" given-E — the bundle returning −3 is E; not a from-nothing derivation
"Three is unique across geometries" NOT claimed — only this geometry+bundle is certified
"K₆ is forced to carry exactly −3" the carrier is forced for gauge; the bundle/weight is selected (R2, R3)
"the spin-ℂ index" REFUTED wording for pure SM — the forced object is twisted (Spin × G_SM)/ℤ₆ (R4)
"no fourth family, by topology" robustness in-category is topological; exclusion of the 2/4 deformation leans on LEP N_ν

4. The insights we used

These are the moves that made the progress believable and reproducible — now shared at working-physicist depth.

4.1 A family count must be the right kind of number — an index, not a volume

The deepest insight is a type constraint, not a value. A continuous bundle modulus can be tuned to any real number; a count tuned that way is meaningless. The right object for "how many copies" is a topological index — an integer that is deformation-proof, that cannot slide under a continuous knob (Atiyah–Singer rigidity). The corpus encodes this as the candidate target-blind axiom AXIOM-COUNT-MUST-BE-INDEX: "an admissible family count must be a deformation-proof topological integer (an index), not a continuous bundle modulus." Stated this way, the criterion rejects "three by dial" because it is a dial, not because it is three — value-blind. This is what makes the gate's central claim more than "arranged to give 3."

4.2 Reuse the carrier — the colour flag is the family flag

K₆ = SU(3)/T² is not introduced for the family count; it is already present as the colour carrier at SG-2, selected there by abelian-isotropy uniqueness (T² the unique purely-abelian SU(3) isotropy giving a clean gauge sector; CP² over-produces gauge). Counting families on the same manifold, with no new geometric term, is the strongest economy in the program (DOSSIER_SG3 §1.1, W3; §4.3). It also localizes the open surface: the carrier debt is paid at SG-2; what remains genuinely open at SG-3 is only the bundle/weight on it.

4.3 Split the gate into "count" and "chirality," and use a boundary for the chirality

The two halves are physically distinct and the corpus keeps them apart: K₆ + BWB counts (magnitude), S¹_Y/ℤ₂ + APS makes them chiral (removes the mirror). The chirality half is a clean no-go: a closed factor cannot produce a one-sided count (F2), so the (+3, 0) result requires the orbifold edge — the bare-circle control (+3, +3) is the proof that the fold is load-bearing. Separating the halves means a soft spot in one (the convention sign, R7) does not contaminate the other (the rigid magnitude).

4.4 Compute the engine from the charges, not from the answer (the κ³/π discipline)

The rep-theory machinery underneath — the m₀(p,q) multiplicity rule (§3.5) and the ℤ₆ centre congruence (§3.6) — was regenerated from scratch: m₀ via Freudenthal's recursion, the ℤ₆ kernel field-by-field from the actual SM hypercharges. Neither was reverse-engineered to a target. This is the κ³/π falsification test in action: a proposed axiom or computation counts only if it would be written without knowing the answer. Passing it is what separates a genuine reproduction from a fit dressed as a derivation.

4.5 When a label is wrong, say so — and check the value survives

The "spin-ℂ" wording is technically false for pure SM (Davighi–Gripaios–Lohitsiri). Rather than defend it, the corpus computed the obstruction (empty all-odd-U(1) scan, with a B−L test-the-test confirming the scanner discriminates) and identified the genuine forced object (twisted (Spin × G_SM)/ℤ₆ via the Tong ℤ₆ and (−1)^F = SU(2) 2π rotation). The discipline: a refuted label is a correctness gain, not a downgrade — provided you confirm the value (|χ| = 3) is a count, not a structure label, and so survives. It does.

4.6 Name the smuggle out loud — the +2-dimension payment

The most honest move in the suite is to flag its own live smuggling surface. K₆ (6D) beats the cheaper CP² (4D) on the family count only by paying +2 dimensions under an anti-fitting tie-break ("adjustable counts as fail") that is aligned with the 3-generation target (REVIEW_SHAPE_SUITE defect 4). The carrier-over-CP² choice is legitimately forced for gauge (SG-2), but the family-count preference leans on a target-aligned tie-break. Naming it as the suite's named live smuggle — rather than hiding it — is what keeps the gate's strong magnitude claim credible.


5. Evidence & reproducibility

5.1 The witness ledger

Each witness is graded hand-checkable / symbolic / machine-lane, with an honest reproduces? flag (DOSSIER_SG3 §2.1):

# Witness What it asserts Grade Reproduces?
W1 BWB index χ(K₆,E) = −3 (App E.1; A2.2; GP.3) family count = |−3| = 3, a rigid integer symbolic (closed-form BWB) Yes — closed-form BWB output on the frozen bundle, not a fit (§5.2)
W2 APS index (n_L, n_R) = (+3, 0) on S¹_Y/ℤ₂ mirror sector removed; one-sided count hand-checkable (toy) / symbolic (full) Yes. One-line interval toy is hand-checkable (§5.3); full count reads the frozen parity table ac4d2df3e708. Bare-S¹_Y control returns (+3, +3) — fold is load-bearing.
W3 χ(K₆) = 6 = |W(SU(3))| = |S₃|; dim K₆ = 6 carrier facts of the colour flag hand-checkable Yes — |W(SU(3))| = |S₃| = 6; dim = 8 − 2 = 6 (§5.3)
W4 m₀(p,q) = min(p,q) + 1 (if (p−q)≡0 mod 3, else 0) rep-theory engine reproducible machine-lane Yes — reproduced target-blind via Freudenthal; all 8 table values regenerate (SHAPE_M0_DELTASPINC §1.1).
W5 ℤ₆ centre via Tong congruence q ≡ 3z₂ − 2z₃ (mod 6) trivially-acting centre is exactly ℤ₆ machine-lane Yes — computed field-by-field from actual SM hypercharges (REDUCTION_SPINTWIST_C §2); not reverse-engineered (passes κ³/π).
W6 spin-ℂ obstruction for pure SM (DGL arXiv:1910.11277 Sec. 7) no U(1) ⊂ G_SM gives all-odd Weyl charges ⇒ no spin-ℂ machine-lane Yes — reproduced (empty all-odd-U(1) scan; B−L test-the-test makes it succeed ⇒ non-vacuous). Refutes the literal "spin-ℂ" label;
W7 selector elimination of CP² ("three by dial") cheaper 4D carrier fails: its count is a continuous modulus symbolic/audit Conditional. Elimination reproduces, but relies on the anti-fitting tie-break aligned with the 3-target (R2). Honest only if anti-fitting is a genuine prior pass/fail.
W8 freeze hashes 0fd19c9ae0c1, ac4d2df3e708, meta a5b1e6f9d951 every Gate-4 object content-addressed before comparison machine-lane Yes in principle (re-hash R1.3/R1.4; recompute meta); not re-run here.
W9 machine certificate certificates/G04_chirality/ parity-table lint + index-consistency against declared values machine-lane AUDIT — checks consistency against the declared −3 and (+3, 0) (CR4.11); a lint, not a from-scratch re-derivation (this is hole R6).

5.2 What reproduces, plainly

5.3 How a reader re-runs / re-derives

5.4 Frozen-object hash index (load-bearing; READ-ONLY)

Branch dcc66f1b2685 · manifest meta a5b1e6f9d951 · (twisted-)spin bundle data on K₆ (R1.4) 0fd19c9ae0c1 · ℤ₂ orbifold + boundary parity ledger (R1.3) ac4d2df3e708. Carrier K₆ = SU(3)/T², χ(K₆) = 6, and the index value χ(K₆,E) = −3 / APS (+3, 0) are frozen R1.3/R1.4 outputs. No new hash is introduced by this dossier; nothing is mutated.

Missing artifact (blocks R4/R6 from-scratch re-index): k6_dirac_spectrum.csvABSENT (closure-campaign batch1; the twisted-Dirac spectrum is not on disk). Do not fabricate it.


6. Open gaps + closure path — the specialist work plan

The most load-bearing section. Each open hole is a work-package a specialist can act on. Closure paths must be non-target-fitted: a proposed axiom or computation counts only if it would be written without knowing the answer (the κ³/π falsification test). The realistic ceiling per residual is AXIOM-CLOSED (name one explicit target-blind posit) or sharper-OPEN; DERIVED-CLOSED is rare and, where it depends on a universal negative, unreachable. Dispositions stay conservative. STATUS-UPGRADES:0.

Leverage ranking (attack order): R3 → (R1 via R3) → R2 → R4 → R6 → R5 / R7. The cardinal honest point: R1/R2/R3 are where the given-E conditioning lives. If a closure path introduces a new target-aligned criterion to force 3 (e.g. "pick the weight that gives three"), it has relocated the input, not removed it — the κ³/π falsification test applies in full force, and the SHAPE suite already flags the +2-dimension forcedness override as the live smuggle.


6.1 R3 — Bundle admissibility / uniqueness (HIGHEST LEVERAGE)

(a) Precise statement. The bundle/weight on K₆ returning χ = −3 is selected inside a declared admissibility class, not proven the unique admissible one. There is no internal theorem saying "the admissible weight is unique." This is the only path that could lift the given-E conditioning (R1): if the admissible bundle is forced by carrier + ℤ₆ centre + hypercharge ledger (no 3-aligned tie-break, no LEP input), then "3" is forced by geometry+admissibility alone, moving the gate toward FORCED.

(b) Why it's hard / traps to avoid. A bundle-uniqueness theorem requires ruling out all admissible weights — an undischarged universal negative. This mirrors the program's already-DEMOTED A5-actor / SHAPE Lemma 3 attempt ("is E_matter the minimal actor realization?"), demoted to OPEN because its success criterion was exactly such a universal negative, left UNMET (competitor matrix all "Unknown"). Trap: any "minimality" claim that secretly uses "it gives three" as the selector is the κ³/π failure mode and must be killed, not banked — it is the SHAPE-suite live smuggle in disguise. Trap 2: do not conflate the bounded, plug-able part (the finite weight-enumeration + index map) with the unprovable absolute-uniqueness phrasing; only the former is a real hole, the latter is a dissolved unicorn (§7).

(c) Exactly what closes it. A bounded, target-blind computation: 1. Enumerate the K₆ weights compatible with the ℤ₆ centre (the Tong-congruence lattice q ≡ 3z₂ − 2z₃ mod 6). 2. Apply the BWB index to each; build the map weight ↦ χ. 3. Test whether minimality (lowest weight) singles out a unique weight giving |χ| = 3 — without invoking "three" and without the LEP input. 4. Sensitivity: does any neighboring admissible weight also give 3 (degenerate), or give 2/4 (excluded only by LEP)? If 2/4 are excluded only by LEP, R3 stays open and R5 is confirmed.

Success criterion (DERIVED-CLOSED, unlikely): minimal admissible weight proven unique and its index forced to 3 with no 3-aligned tie-break and no LEP input. Realistic ceiling (AXIOM-CLOSED): name AXIOM-MIN-WEIGHT-LIFT — "the admissible bundle on K₆ is the minimal-weight twisted-spin ((Spin × G_SM)/ℤ₆) lift compatible with the ℤ₆ centre and the hypercharge ledger; the family count is its Borel–Weil–Bott index" — and check it yields 3. This is writable with no generation number in sight, so it passes the κ³/π falsification test as a statement. Expected outcome (sharper-OPEN): minimality does not single out a unique weight, or excludes 2/4 only via LEP ⇒ "3" stays rigid-given-a-selected-bundle, selection conceded. A REFUTING result is a valid close: a strictly-lower-cost admissible weight giving χ ≠ −3 ⇒ the minimal-weight axiom is false and the bundle was 3-selected — a structure-first datum departing from the target (the good kind of negative).

(d) Machinery & inputs. Borel–Weil–Bott on SU(3)/T² (closed-form, GP.3); the ℤ₆ Tong-congruence lattice (REDUCTION_SPINTWIST_C §2); the hypercharge ledger (A2.2). Start from: DOSSIER_SG3 §4.1; the m₀/Freudenthal machinery (SHAPE_M0_DELTASPINC §1.1). Frozen reference value: χ = −3 (0fd19c9ae0c1).

(e) Leverage. If R3 closes, R2's tie-break dependence falls, R1's given-E conditioning lifts, and R5 becomes boundary-rigid. It is the busiest node's single highest-value move.


6.2 R2 — Bundle-selected vs bare-carrier-forced (the SHAPE suite's named LIVE SMUGGLE)

(a) Precise statement. "3" is read off a chosen weight. On the cheaper carrier CP² = SU(3)/U(2) (4D) the family count is a continuous bundle modulus — "three by dial." K₆ (6D) is preferred because its count is a rigid integer, but that preference is purchased by paying +2 dimensions under an anti-fitting tie-break ("adjustable counts as fail") that is aligned with the 3-generation target, overriding the declared lower-dimension Occam rule (REVIEW_SHAPE_SUITE defect 4 — the suite's named live smuggling surface).

(b) Why it's hard / traps to avoid. The firewall question is whether the anti-fitting tie-break is a genuine prior pass/fail (written before, and independently of, the 3-generation target) or was inserted to make K₆ beat CP² (i.e. to reach 3). The firewall, datum by datum (DOSSIER_SG3 §4.3):

Selection datum Record Target-blind? Verdict
Carrier K₆ over CP² abelian-isotropy uniqueness (CP² over-produces gauge at Gate 2) yes at Gate 2 (gauge over-production is data-independent) PASS — K₆ is the unique clean carrier for gauge
The +2-dimension payment "for forcedness" at Gate 4 §4.6: "selector pays two extra dimensions for K₆'s forced count" SUSPECT — overrides lower-dimension Occam with anti-fitting FLAGGED — the live smuggle
Anti-fitting firewall ("adjustable counts as fail") a declared prior pass/fail predicate (count must be a deformation-proof integer) borderline — a type predicate, not a value predicate PASS-as-type / SUSPECT-as-tiebreak
The bundle/weight returning −3 selected in the admissibility class NO at the weight level FAIL — the weight is selected (this is R3)

Trap: do not bank the elimination of CP² as a clean win — it leans on the target-aligned tie-break. Trap 2: the carrier-over-CP² choice is legitimately forced for gauge (SG-2); do not let that legitimate result launder the family-count tie-break, which is separate.

(c) Exactly what closes it. Re-run the selector with the 3-generation target MASKED: 1. Compute the family-count map on CP² = SU(3)/U(2) explicitly; confirm it is a continuous bundle modulus (not an index). 2. Verify the type-axiom AXIOM-COUNT-MUST-BE-INDEX — "an admissible family count must be a deformation-proof topological integer, not a continuous bundle modulus" — rejects CP² for being a dial, value held abstract (a CP² tuned to four fails it just as a CP² tuned to three does). 3. Survey the sub-6D SU(3) carrier list for any rigid-integer-count survivor; confirm K₆ is the minimal one, value-blind. 4. Falsification test: does the +2-dimension payment still go through on "rigid-integer-count" alone, with the 3-target masked?

Success (AXIOM-CLOSED, likely if the masked run passes): the +2-dimension payment is justified by type (rigid-integer-count), not target (the value 3); the smuggle is defused. DERIVED-CLOSED (possible, bounded): K₆ proven the unique minimal-dim rigid-integer-count SU(3) carrier, value-blind. sharper-OPEN: the masked run shows the payment only goes through because it yields 3 ⇒ smuggle confirmed (the honest negative the suite warns about). REFUTING result: a sub-6D carrier has a rigid-integer count ≠ 3 the selector should have preferred ⇒ the carrier choice was 3-selected.

(d) Machinery & inputs. The selector / category-relativity layer (Appendix B1/B2, B2.0.3.4); the CP² = SU(3)/U(2) bundle-count computation; the sub-6D SU(3) homogeneous-space list. Start from: DOSSIER_SG3 §4.3; REVIEW_SHAPE_SUITE defect 4.

(e) Leverage. Closing R2 at the type level removes the gate's reliance on a target-shaped criterion; it is the credibility anchor for the strong magnitude claim. Partially shared with the SHAPE shape-minimality dossier.


6.3 R4 — "spin-ℂ" wording REFUTED for pure SM (cleanest correctness win)

(a) Precise statement. The manuscript (§6.4, Appendix E.1, CR4.5, GP.3) calls the object the "spin-ℂ Borel–Weil–Bott index," but pure SM content provably forbids a spin-ℂ structure (DGL arXiv:1910.11277 Sec. 7). The value 3 is intact; the structure label is wrong. The genuine forced object is the twisted (Spin × G_SM)/ℤ₆ (n = 2).

(b) Why it's hard / traps to avoid. Not hard — the refutation is already computed on-disk (REDUCTION_SPINTWIST_C; SHAPE_M0_DELTASPINC §2). Trap: do not present this as a value downgrade — |χ| = 3 is a count, not a structure label, and survives. Trap 2: do not assert "twisted-spin index = 3" as a re-derived value yet — the from-scratch twisted-Dirac re-index is BLOCKED (see below); the corrected wording is reachable now, the re-derived value is not.

(c) Exactly what closes it. Two tiers: - DISCLOSED-CORRECTED (reachable now, owner wording patch): replace "spin-ℂ index" with "twisted-spin ((Spin × G_SM)/ℤ₆) index" in §6.4 / Appendix E.1 / CR4.5 / GP.3, citing DGL 1910.11277 + Tong 1705.01853 (+ Hsieh–Tachikawa–Yonekura; García-Etxebarria–Montero 1808.00009). Named axiom: AXIOM-TWISTED-SPIN-GIVEN-E — "the global spin structure on the active branch is the twisted (Spin × G_SM)/ℤ₆ forced by E's charges mod 2; the family count is the index of the associated twisted-Dirac operator on K₆, |Index| = 3; spin-ℂ is OBSTRUCTED for pure SM." This is FORCED-GIVEN-E (it references only the SM charges, the centre, and Dai–Freed single-valuedness; carries no generation number). - DERIVED-CLOSED of the value-under-corrected-structure (BLOCKED_INPUTS): mount the absent twisted-Dirac spectrum file k6_dirac_spectrum.csv and confirm the twisted-Dirac index on K₆ = 3, matching the BWB-on-line-bundle value. A REFUTING result (twisted-Dirac index ≠ 3) would be a genuine downgrade — but the structure-and-value consistency is expected.

(d) Machinery & inputs. The spin-ℂ obstruction scan + ℤ₆ kernel + (−1)^F identification (REDUCTION_SPINTWIST_C §2; scripts cited there). The twisted-Dirac index computation requires k6_dirac_spectrum.csv (ABSENT — must be generated, not fabricated). Start from: DOSSIER_SG3 §4.4.

(e) Leverage. Pure correctness/honesty gain; no value change. It also clarifies R6 and R7, which both reference the twisted structure.


6.4 R6 — The G04_chirality certificate is a consistency lint, not a from-scratch re-derivation

(a) Precise statement. certificates/G04_chirality/ lints the parity table (survivors = SM set, mirrors projected, counts consistent) against the declared index values −3 and (+3, 0) — under those declared values (CR4.11). It does not independently regenerate the index from the bundle data target-blind. The §6.4 card's "Claimed certificate pass" is therefore a lint pass, not an independent proof.

(b) Why it's hard / traps to avoid. Trap: do not print the certificate as an independent proof of the index — that would overclaim until the from-scratch re-index runs. The re-index is BLOCKED on the same missing k6_dirac_spectrum.csv as R4. The BWB-on-line-bundle symbolic leg and the APS parity-table leg are re-derivable now; the full twisted-Dirac re-index is not.

(c) Exactly what closes it. A fail-closed reproducibility task: 1. Mount the K₆ bundle data + the twisted-Dirac / BWB computation. 2. Re-derive the index from the bundle data target-blind (do not read the declared −3). 3. Confirm it equals the frozen −3 and that the APS boundary count re-derives (+3, 0) from the parity ledger ac4d2df3e708. 4. Re-hash R1.3/R1.4 and confirm the meta-hash recomputes to a5b1e6f9d951.

Success: VERIFIED — "Claimed certificate pass" becomes machine-real beyond a lint. REFUTING result (valid close): the index fails to regenerate ⇒ Gate-4 downgrades per the §5.3.7 falsifier (count ≠ 3 / count moduli-dependent / mirror survives) — and the downgrade cascades (R8).

(d) Machinery & inputs. The G04_chirality certificate harness; the parity ledger ac4d2df3e708; the bundle data 0fd19c9ae0c1; the (absent) k6_dirac_spectrum.csv. Start from: DOSSIER_SG3 §4.6, W9.

(e) Leverage. Converts an asserted index into a machine-checked one with no new physics — high value-per-effort, but gated on the same missing twisted-Dirac artifact as R4. Closing R4's CSV unblocks both R6 and R4's DERIVED-CLOSED tier.


6.5 R7 — The chirality SIGN convention bit is unpinned (and its default is on-disk WRONG-sign)

(a) Precise statement. The magnitude 3 is robust and convention-independent. The sign −3 (chirality orientation — "left-handed, not right-handed") rides an unpinned Pin⁺/Pin⁻ convention bit, the same bit that leaves BG-10's e^{±iπ/4} unfixed (SHAPE_M0_DELTASPINC §4; HANDOFF_SPECIALIST_3). The family-count magnitude is independent of it; the chirality label is convention-stated, not pinned by the frozen record.

The honest, structure-first datum that must not be glossed: the geometry's default setting of this bit comes out the WRONG sign in the sister gate. CLOSURE_CAMPAIGN_RESULT_2026-06-24 (line 33) records that the default χ = −3 (= 5 mod 8) gives arg = −135° = e^{−3iπ/4} — the wrong sign for the BG-10 leptogenesis target (which needs σ_ν = +1 → e^{+iπ/4}). "The axiom DEPARTS from the target rather than wearing it." So "left-handed, no surviving mirror" is convention-stated, and the default convention currently points the wrong way downstream — a concrete, falsifiable, on-disk finding, not a bland "convention bit." (Do not launder this into a generic "unpinned bit" — that is the exact failure the re-grade review caught.)

(b) Why it's hard / traps to avoid. It is a single discrete bit, but it is shared across three threads (SG-3 chirality, BG-10 phase, SHAPE q-sign) — three threads landing on one wall (convergence, evidence the wall is real and singular, not evidence it has been crossed). Trap: do not assert the chirality label "left-handed" as an established output; per the frozen record only the magnitude is settled. Trap 2: do not hide the wrong-sign default — surface it as the real datum.

(c) Exactly what closes it. Hand the spin-bordism specialist (same handoff as BG-10 Rule A / HANDOFF_SPECIALIST_3) the question: is the orientation Pin bit on S¹_Y/ℤ₂ fixed by Dai–Freed single-valuedness of the chiral-fermion measure, or is it a free convention? Named axiom: AXIOM-CHIRALITY-ORIENTATION — "the orientation/Pin bit fixing the sign of the index as left-handed is a single discrete convention choice on the S¹_Y/ℤ₂ reflection; the family-count magnitude is independent of it." AXIOM-CLOSED (bit named/isolated) is the realistic endpoint. DERIVED-CLOSED if the bit is shown Dai–Freed-forced. Structure-first outcome: if the twisted-structure default is the correct (left-handed) sign, R7 strengthens; if the default is the wrong sign (as the on-disk datum currently shows for BG-10), that is a structure-first datum worth recording — possibly a real obstruction for the downstream leptogenesis leg, not a free pass.

(d) Machinery & inputs. Dai–Freed / spin-bordism single-valuedness of the chiral-fermion measure (García-Etxebarria–Montero arXiv:1808.00009); the twisted (Spin × G_SM)/ℤ₆ structure (R4); the mod-8 ledger (CLOSURE_CAMPAIGN_RESULT_2026-06-24 line 33). Start from: DOSSIER_SG3 §4.7; SHAPE_M0_DELTASPINC §4.

(e) Leverage. The count is unaffected, so this is a chirality-label residual, not a count residual. But it is shared with BG-10 (η_B / leptogenesis sign), so resolving the bit downstream-correctly closes a wall common to three gates.


6.6 R5 — The index-changing deformation is excluded by data, not topology

(a) Precise statement. The index −3 is deformation-proof inside the declared admissibility class. The neighboring values (2, 4) that an index-changing bundle deformation would produce are excluded by the LEP N_ν bound (2.984 ± 0.008), i.e. by data. So "no dial" is topological in-category + empirical at the class boundary — two distinct robustnesses that must not be conflated.

(b) Why it's hard / traps to avoid. No closure path "solves" this; the task is to state the two robustnesses distinctly. Trap: do not present "no fourth family by topology" without the qualifier — the in-category rigidity is topological, but the exclusion of the 2/4 deformation leans on LEP.

(c) Exactly what closes it. A consistency sweep (DISCLOSED-CONSISTENT, already substantially in place via Objection 4): keep the topological-vs-empirical split explicit wherever "topologically forced / no fourth family" appears. Named axiom: AXIOM-INDEX-RIGID-IN-CATEGORY — "the family count is invariant under every continuous deformation within the declared admissibility class; index-changing deformations across the class boundary are excluded by the LEP N_ν bound, an external datum." This becomes boundary-rigid (i.e. fully topological) only if R3 closes — R5 is the dependent of R3.

(d) Machinery & inputs. Appendix C2/C4/E; the Objection 4 answer; the LEP N_ν = 2.984 ± 0.008 datum. Start from: DOSSIER_SG3 §4.5.

(e) Leverage. Keeps the "no dial" claim correctly scoped; closing R3 would absorb it.


6.7 R8 (dependency record) — the blast radius of "3"

Not a closure target, but recorded so a reviewer sees SG-3 is the dependency graph's busiest node. SG-8's generation module dim 𝒢_gen = 3 is inherited from this index (A1.13). The SG-7 threshold δ-vector is load-bearing on the count −3 (Appendix E / G.3.2: "if the count were −2 or −4 the column sums would not reproduce (+4.8424, −3.1112, −1.7313)"). A revision of "3" cascades to Gates 5/7/9/10 (CR4.7; CR4.11 downgrade rule). Action: keep the cascade wired and the given-E qualifier travelling with "3" everywhere it is consumed (SG-8 already carries it; SG-7 is diagnostic-only and so cannot certify "3" downstream).


6.8 Attack-plan roll-up

Residual Technique Named axiom (κ³/π-clean) Specialist target / owner artifact Realistic endpoint
R3 bundle uniqueness selector + no-go AXIOM-MIN-WEIGHT-LIFT weight-enumeration + index map; bundle-uniqueness theorem sharper-OPEN → AXIOM-CLOSED if minimality yields 3 value-blind
R1 given-E axiom-floor AXIOM-CONTENT-GIVEN (closeable only via R3) AXIOM-CLOSED (where it sits); gate stays DERIVED-GIVEN-E
R2 bundle-selected (live smuggle) firewall + selector AXIOM-COUNT-MUST-BE-INDEX masked-target selector re-run; minimal-dim rigid-count carrier proof OPEN → AXIOM-CLOSED if masked run passes
R4 spin-ℂ wording computed no-go AXIOM-TWISTED-SPIN-GIVEN-E owner wording patch; twisted-Dirac re-index (CSV absent) DISCLOSED-CORRECTED now; DERIVED-CLOSED BLOCKED_INPUTS
R6 certificate lint machine-lane (none) re-derive index target-blind from bundle data (CSV absent) BLOCKED_INPUTS → VERIFIED
R7 chirality Pin bit axiom-floor AXIOM-CHIRALITY-ORIENTATION Dai–Freed bit-forcing (HANDOFF_SPECIALIST_3) AXIOM-CLOSED (count unaffected; default currently wrong-sign)
R5 data-excluded deformation scope-the-claim AXIOM-INDEX-RIGID-IN-CATEGORY consistency sweep DISCLOSED-CONSISTENT (dependent on R3)
R8 inheritance dependency record (none) keep cascade wired DISCLOSED

REDUCE-vs-RELOCATE verdict. The plan does not turn one hard problem into three harder ones. Each path either (a) names a single target-blind axiom paying a debt in plain sight (R1, R2-type-axiom, R5, R7), (b) attempts a bounded falsifiable computation that could remove conditioning (R3 weight-enumeration; R2 masked-target selector), (c) is a pure correctness fix already on-disk (R4), or (d) is a mechanical owner/CSV task (R6) or a dependency record (R8). The two genuine count-moving attempts (R3, R2) are smaller than the original gate (one enumeration / one selector run each) and both carry the explicit κ³/π falsification test. No DERIVED-CLOSED is promised. The honest expected outcome of a full campaign: R4 DISCLOSED-CORRECTED (cleanest win), R1 AXIOM-CLOSED (already there), R5 DISCLOSED-CONSISTENT, R7 + R2-type-axiom AXIOM-CLOSED, R3 sharper-OPEN (universal negative unmet), R6 BLOCKED_INPUTS → VERIFIED if the CSV is mounted, R8 DISCLOSED. The given-E qualifier is not removable short of R3 closing, and R3's own ceiling is AXIOM-CLOSED.


7. Honest ceiling & scope

7.1 What is genuinely shown

The strong, genuine content is real and reproduces: a rigid, deformation-proof integer |χ(K₆,E)| = 3 (closed-form Borel–Weil–Bott), plus a clean no-go for the mirror half (APS (n_L, n_R) = (+3, 0) on the S¹_Y/ℤ₂ fold, with the bare-circle control (+3, +3) proving the fold load-bearing). The rep-theory engine under it (m₀(p,q), the ℤ₆ Tong congruence) regenerates target-blind from the actual SM charges. On the program's grading this ties SG-2 as the strongest gate, because an integer index is more rigid than a coupling-power ratio. The magnitude claim is a confident, falsifiable bet: re-derive the index from the bundle data alone and it must come back −3; a strictly-simpler admissible bundle on K₆ returning a different integer would break it. So far nothing moves the magnitude.

7.2 Dissolved unicorns — shared ceilings, never open weakness, never claimed as proven

7.3 Bright-line claims the frozen docs DENY — never printed as proven

7.4 The anchors paid

In the terminate-on sense (every residual must reduce to a measured invariant or a single named, target-blind posit — never to "it gives three"), this gate's target anchor is the observed spectrum E itself: the family count "3" — and the rigid integer χ(K₆,E) = −3 carrying it — is a facet of E (χ is read with E as input), reinforced at the admissibility-class boundary by the external measured invariant LEP/SLD N_ν = 2.984 ± 0.008 (which excludes the index-changing deformation to 2/4). spectrum-E is measured-but-irreducible: the index reads it, it cannot derive it (given-E ≠ derivation of E). The would-be deeper target — "3 across all geometries / bare-carrier-forced" — has no independent witness and stays OPEN (R1/R2/R3); the in-category "no-dial" rigidity is SCHEME-ANCHORED to the declared category (AXIOM-INDEX-RIGID-IN-CATEGORY); the orientation/chirality bit awaits a new named invariant (Dai–Freed, AXIOM-CHIRALITY-ORIENTATION). The count's magnitude is genuine and reproduces; its forcedness across geometries is not anchored.

The gate does not depend on ℏ, M_Pl, α_i(M_Z), y_t, or |V_us| for the count itself (those enter downstream gates that consume "3"). It depends primarily on spectrum-E, with LEP N_ν at the boundary.

7.5 The single load-bearing sentence

The geometry genuinely makes the family count a rigid deformation-proof integer of the correct kind (3, no mirror) — but it does so by reading a bundle whose data is E itself; it certifies this geometry+bundle yields three given the particles we observe, not that three is forced across geometries, and not that the bundle is unique. given-E is the load-bearing qualifier of the whole gate.

STATUS-UPGRADES:0. Frozen branch dcc66f1b2685 / a5b1e6f9d951 READ-ONLY. given-E ≠ derivation of E; selection ≠ derivation; AXIOM-CLOSED ≠ proven; dissolved ≠ solved. Nothing applied, nothing deployed.


References & source map

Website source-of-truth (common material — link, don't duplicate): - Paper I, GUT.html — https://physics.magflowmeters.com/articles/GUT.html — §6.4 (Gate-4 card), §5.3 (narrative), §3.4 (topological-count machinery), §4.6/§4.9 (selector dimension-payment / anti-fitting firewall), Strongest-Objections block (Objection 4), Appendix E (E.1/E.2/E.3/E.6), CR4 (CR4.1–CR4.13), GP.3 (counting theorems), B1/B2 (selector / category-relativity), C2 (K₆ dossier), C4 (S¹_Y/ℤ₂ dossier), A2.2 (bundle ledger), A1.8 (parity table), A1.13 (generation module), R0 (freeze records). - Paper IV, TOE.html — https://physics.magflowmeters.com/articles/TOE.html (downstream cross-references only). - Gaps & Walls Register — https://physics.magflowmeters.com/articles/GAPS_AND_WALLS_REGISTER.html (disposition vocabulary).

Corpus inputs (read for this dossier; all under …/Fable_Version/rendered/): - TOE/PER_GATE_DOSSIERS/DOSSIER_SG3_THREE_GENERATIONS_CLOSURE_ATTACK.md — the closure-attack dossier this expands (§0–§5, R1–R8, witness ledger, attack plan). - TOE/PER_GATE_DOSSIERS/SG3_COMPLETION_HANDOFF/01_DOSSIER.md — same content, completion-handoff copy. - TOE/REDUCTION_SPINTWIST_C_2026-06-23.md — spin-ℂ obstruction (DGL); ℤ₆ Tong kernel; (−1)^F; twisted (Spin × G_SM)/ℤ₆; T3 REFUTED; FORCED-GIVEN-E. - TOE/SHAPE_M0_DELTASPINC_QSIGN_2026-06-23.md — m₀(p,q) = min(p,q)+1 table; ×2 spinor-weight; Δ_spin-c fail-closed; Pin⁺/Pin⁻ convention bit unpinned. - TOE/GATE_REGRADE_BESTCASE_AND_HOLES_2026-06-29.md — SG-3 regrade (held DERIVED-GIVEN-E); review flags; the on-disk WRONG-sign default datum (line 209; CLOSURE_CAMPAIGN_RESULT line 33: default χ=−3 = 5 mod 8 ⇒ arg = −135°). - TOE/SPECIALIST_HOLE_QUEUE_2026-06-29.md — the six SG-3 holes as work-packages.

External literature (cited at the corpus's own level): - Davighi, Gripaios, Lohitsiri, "Global anomalies in the Standard Model(s) and Beyond," JHEP 07 (2020) 232, arXiv:1910.11277, Sec. 7 — no U(1) ⊂ G_SM gives all-odd Weyl charges ⇒ SM has no spin-ℂ. - Tong, "Line Operators in the Standard Model," JHEP 07 (2017) 104, arXiv:1705.01853 — ℤ₆ = maximal trivially-acting centre; q ≡ 3z₂ − 2z₃ mod 6. - Hsieh, Tachikawa, Yonekura, "Anomaly of the Standard Model" — twisted Spin × G_SM /ℤ_q; (−1)^F in the gauge centre. - García-Etxebarria & Montero, "Dai–Freed anomalies in particle physics," JHEP 08 (2019) 003, arXiv:1808.00009 — Dai–Freed / spin-bordism fermion-measure. - Allanach et al., arXiv:2111.04148 + anomaly-free atlas JHEP 02 (2019) 082 — infinitely many anomaly-free chiral U(1) extensions ⇒ anomaly-freedom fixes neither content nor generation number.

Dossier built 2026-06-29. Our geometry (13D K₆ branch) only. Common material referenced to the published website source-of-truth, not duplicated. STATUS-UPGRADES:0; nothing applied, nothing deployed.