Verification and inclusion
On-page book · Complete original sources
# Verification and inclusion ledger
The book is a developed critical edition of the two supplied manuscripts, not a certification of every original assertion. Both complete originals are preserved byte-identically in `book/source-editions` and on the local page. Their hashes, all chapter headings and original line positions are recorded in `source-provenance.json`. The coverage review documents what was read in this run versus earlier audits. Historical false claims are retained as sources, with corrections in the critical text; they cannot consistently be formalized as valid theorems.
## Current kernel-checked layer
The working critical book has 40 chapters (348 pages in the Linux working edition), with fourteen unique canonical on-page interactive tools, full-source code views and complete historical source volumes. Larger examples include genus4/5/9/16, finite depth-capped hyperbolic lattice windows and a nine-qubit QCNN.
The pinned Lean 4.24/mathlib suite has **360 printed theorem dependency reports** including covering-obstruction, worked persistence-resolution, nodal DG and actual finite RG modules. The full suite passed with warnings treated as errors: 357 reports use standard foundational axioms only and 3 are axiom-free. There are no admissions or custom project axioms. This is a report count, not a count of all book theorems. The historical standalone seven-result suite is unchanged and was not rerun for this milestone.
| Material | Actual formal/computational interface | Remaining boundary |
|---|---|---|
| Nodal/projective resolutions | Existing PowerSeries/module resolution layer; BranchExtension actual infinite polynomial algebra and nonsplit module maps; exact path/matrix diagnostics | General dessin category interfaces and all module classifications not fully formalized |
| Semi-invariants/AR | BookInterfaces Plücker and actual nodal matrices; nine-module/twelve-map/four-mesh exact checker | Almost-split universal factorization, whole component classification, Ringel quilt and derived equivalences remain prose/primary theorems |
| Clock/shift and channels | BookInterfaces explicit clock/shift, distinct-site matrices, arbitrary finite complex Kraus positivity and trace preservation | Infinite operator domains, C* completions, physical implementations and QFT reconstruction not fully formalized |
| Arithmetic character/Gram | ShortRowGram actual finite-field masks, complete orthogonality, finite energy expansion, Eisenstein norm and mod7 evaluation; ShortRowPairBarrier universal finite complex Gram duality and weighted pair-square inequalities | Planar Poisson, full CRT/ideal-character interface and low-gcd short-row estimate remain separate; seven finite Hecke Euler-restoration identities do not formalize infinite continuation/contours |
| Theta arithmetic/descent | Existing Jacobi/projective/tame power-series and matrix layers | Full order functor, arithmetic descent and global positive determinant comparison remain open |
| Higher genus/hyperbolic geometry | HyperbolicCoordinates formal disk/Lorentz and 4D norm identities; uniform prose passport proof; exact permutation/ramification checks through genus16; bounded reflection diagnostics | No formalized general uniformization or certified compact hyperbolic quotient; displayed patch is numerical and finite |
| Learned renormalization | Actual PennyLane three-architecture and larger nine-qubit training; NumPy comparison, analytic target/gradient checks; actual Strawberry Fields training | No optimizer convergence theorem, hardware run, general QFT RG theorem or quantum advantage claim |
| ToricGT hybrid | Actual pinned upstream Weyl module, trained PennyLane surrogate, six exact rational 15-minor PSD certificates | No upstream language-model campaign reproduced; no inferred arithmetic L-function operator |
| Persistence extension | Actual MvPolynomial worked ungraded free resolution: injectivity, kernel/image equalities, ideal quotient, projectivity and zero residue Hom boundaries; 19 new printed dependencies. General prose staircase/CW/Schur/flag/cone and exact finite laboratory remain independently reviewed | General staircases, graded minimality, CW/Schur/category interfaces, surface-order coefficient tensor cone and channel norms are not yet fully Lean formalized; quantum distances are not interval certificates; Sage source supplied unexecuted |
| Numerical architecture inspector | 22 saved cases and 836 shape-labelled DAG nodes plus a computed clock; actual backend/reference checks, all-case browser tests, 3D/flat dependency views, exact coordinate values, gate/covariance/witness/expert/certificate panels and one canonical book embed | Optional upstream probes, full model campaigns, upstream campaigns and general mathematical formalization remain separate; new GPU cases are scoped below; numerical arrays are not kernel-checked theorems or new training |
| Covering/GLS/geometric extension | Actual theta finite-quotient cover, 36 exact Hom pairs, local endomorphism radical and higher-genus non-lifting proofs; three fresh Lean matrix statements with standard dependencies | General covering/module/AR/derived/Fukaya interfaces and uniform higher-genus obstruction are not fully Lean formalized; no GLS identification or arithmetic positivity inferred |
| DG foundations | Actual exterior-line algebra, differential, parity Leibniz, contraction, ideal augmentation, semilinear clock and ambient polynomial nodal sequence:20 kernel-checked components; reviewed full prose node/completion/Ext/star/filtered-module constructions and one canonical finite-window tool | Full categorical Z-graded DGA, completed topology, concrete q-substitution, Ext identification, A-infinity transfer, HH/GNS and derived equivalence are not fully formalized; finite boxes are k-subcomplexes only |
| Sage | Exact scripts supplied | Installed launcher points to missing app; Sage execution not passed |
| NVIDIA | Executed cuStateVec7/11/23-qubit surface circuits and cuTensorNet contraction on the Linux RTX4090; isolated approved conda environment and explicit lock | Largest case has norm/inverse-echo checks; no independent large-state reference, physical hardware execution or speedup claim |
The web book places selected Lean and executable code beside chapter content. These panels label scope rather than imply complete chapter formalization. The original manuscripts also contain experimental and proposed engineering statements: they need measurements or model validation, not a Lean proof that unperformed experiments happened.
## Book-wide formalization backlog
1. Connect the existing actual path/order definitions to each general dessin resolution and prove its exactness uniformly.
2. Formalize the exact-category AR factorization properties and separate quotient, lattice, stable and derived interfaces.
3. Develop determinant-functor maps with their perfection, linearity and symmetry hypotheses.
4. Formalize the analytic lattice Poisson and low-gcd kernel estimates; the latter is still an unresolved mathematical problem.
5. Formalize operator domains, regularized determinant convergence and the precise spectral zero-location bridge once independently constructed.
6. Formalize physical finite-channel statements further, and treat continuum, hardware, measurement and learning claims according to their actual evidential type.
A complete GRH proof and complete Lean formalization of all material have **not** been established. No fitted spectrum, trained network, finite numerical check or restatement of an open bridge is accepted as completion.
## Coprime pair-square arithmetic trial
The current book is148pages,32chapters; focused articles28/57pages. Four new
ShortRowPairBarrier declarations prove actual finite-complex frame duality,
row-diagonal and weighted/unimodular lower bounds, with printed standard
axiom dependencies. They do not formalize the full ideal-character passage,
lattice count or fixed-field prime-ideal theorem. The resulting prose
application rules out a proposed absolute pair-square route at the3/4
short-row endpoint; it does not settle the signed original-family mean square.
The character is sextic in the pinned original; its square is cubic. Earlier
high-gcd exposition has this label corrected. The approved CPU environment
runs deterministic exact integer sextic diagnostics, not training or a GRH
certificate. No sources, environments, pins or checkpoints changed.
## Actual RG and quantum-feature TokenGT correction
Current book159pages/34chapters with nine unique tools; focused articles30/62
pages. The new actual-RG chapter precedes auxiliary compression/covariance
benchmarks. FiniteRG.lean checks9 full finite scalar Gibbs marginal/action,
partition, retained-observable, nesting, sandwich and transfer-trace
identities. All9printed types use only standard foundations. Quantum
spectral-log/partial-trace/operator and full learned/QEC/categorical programs
are not all kernel checked. Full pinned suite295passes,293standard-only and2
axiom-free. The standalone historical7-result suite is unchanged, not rerun.
New exact/full-action and floating-point checks cover all generated constants,
Walsh/Pauli terms, nested partition/observable comparisons, Γ=0 classical
limits, quantum tensor ordering, trained-heldout/OOD action/constant/partition/
observable errors and matched classical baselines. The12RG circuit angles
are nominal;3finalRZ parameters are structurally inactive for itsZ readout.
A five-qubit generation/dephasing/sampling example reports learned-distribution
shot error separately from teacher bias. Sage source is supplied unexecuted.
Every original architecture has actual passport/quiver/order/forbidden-pair
context and selected tensor/gate/token supports. Four matched TokenGT variants (three quantum-feature, one classical), a concept critic and continuous GFlowNet are explicitly bounded
finite hybrids. Actual F2Rips homology, ideal bit-flip QEC logical-channel
memory evaluation and exactCSS [[10,4,1]] ranks/distances are checked by
separate code and independent review. These computations don't certify
semantic reasoning, global GFN distribution accuracy, physical topological
phase, coherent quantum attention, HMS/materials results or GRH.
Book proofs and evidence: research/actual-rg.md,
planning/reviews/actual-rg/REVIEW.md, examples/rg/README.md,
examples/quantum_tokengt/README.md and outputs/checks/2026-11-03-actual-rg/.
Both complete historical manuscript copies and source-clone pins are preserved.
## Linear-relation and Fitting source expansion
Current book165pages/35chapters,17 vector diagrams,ten canonical tools;
focused articles32/65pages. LinearRelations.lean checks11 statements using
actual Submodule relations, graph composition/inverse, explicit intertwiners,
both filtration directions, arbitrary finite-dimensional Fitting splitting
at dimV, kernel stabilization and unit-power annihilation. Full-suite306
reports pass (304 standard-only,two axiom-free). No custom axioms/admissions.
The equivalence of the specified full A_m branch category with finite
endomorphisms, Jordan/Hom classification, topology of continuous completed
modules, abstract splitting-functor classification, Cantor coding and cyclic
transport determinant are proved in prose rather than packaged in Lean.
Finite rational checks cover8 modules/10 full-quiver Hom systems; an independent
oracle checks24 transports/125 Hom cases. The interactive coefficient graph
is a projection of a finite module, not a uniformizing Riemann surface or
an arithmetic operator. The source's graph, stable kernel, rigidity, lambda=1,
category and generic infinite-word assertions are explicitly corrected.
## Original-family signed ideal split
Current book168pages/35chapters/17diagrams/ten unique labs; audit papers20/49
and focused papers33/69pages. MobiusSplit.lean adds8 printed statements in
an actual natural-number ArithmeticFunction Dirichlet ring, plus the generic
commutative-ring inverse identity. Full suite314passes,312 standard-only,
two axiom-free. This does not formalize an ideal convolution ring or identify
ideal Möbius with Möbius of the integer norm; those are different objects.
Prose proves the full-mask ideal type-I/type-II regrouping and a fixed-U=1
obstruction to an all-cutoff separate-piece bound, using the named classical
fixed-field prime-ideal theorem. It does not refute the signed original MS
or moving cutoffs. The exact checker covers307 good ideals including inert
primes and three-prime squarefree products,1535coeff/regroup cases,19680zero
encounters and900conductor exponent ledgers. A norm637/U7 sign control stops
extrapolation of the U1 positivity argument. Its rational finite weight is
not a certificate of the smooth-weight/supremum analytic input. Independent
review includes a separate convolution/regrouping/weighted-row oracle.
## Actual infinite polynomial algebra and module extension
`BranchExtension.lean` now checks 21 actual generated-algebra and module
statements. Polynomial evaluation injects k[t] into the algebra, proving
non-finite-dimensionality. Evaluation at zero and dual epsilon gives true
module structures on k^5 and (k[epsilon]/epsilon^2)^5; algebra-linear
injection/projection have exact image/kernel and no module section.
All coefficient fields are permitted, including characteristic2. The fresh
full suite has335 reports:333 standard-foundation only andtwo axiom-free.
The normal-form path-quotient/pullback identification remains proved prose;
Ext objects, ambient AR universality, whole-book formalization and GRH are
not inferred. Independent fresh Lean/whole-polynomial and Q/F2/F3/F5 checks
pass; source and theorem-type review is recorded in
`planning/reviews/branch-extension/REVIEW.md`.
## Reference-informed inspector and ConvexTok
Current edition173pages/35chapters,18source-aligned vector diagrams andeleven
unique laboratories; articles20/49 and33/71pages. The expanded architecture
chapter pins Bycroft's reference source and distinguishes successful live
observations from source-confirmed picking and walkthrough mechanisms. The
original inspector now uses tensor-dimensional planes, semantic lanes,
guided operation focus and exact coordinate picking. Its34saved cases pass
55,299coordinate checks and37,701 recomputed scalar/complex formula checks.
Source/training/trace hashes remain unchanged. Diagnostic-only max-plus or
softmax branches are labelled and dashed; unsupported cell derivations are
explicitly unavailable. Final root browser checks cover theme/mobile,
all collections, loading races, checksums, camera and unique book embedding.
ConvexTok v1 is pinned and fully reviewed; its small LP/rounding formulation
is used in five original/reproduced finite corpora. Sixteen rational primal/
dual pairs certify LP optima;2,312exact-size vocabulary subsets and a separate
2,486at-most-K enumeration certify integer comparisons. Examples include
LP10<IP11, LP31/2<IP16 and restricted genus2/4 path-corpus optima19/31.
Python and48JavaScript replays plus64browsercontrol states pass; deliberately
corrupt certificates fail. This is exact small-instance optimization, not
formal Lean optimization, upstream NLP training or a quantum speedup. No
Fukaya equivalence, arithmetic spectral comparison or GRH follows from
compression. The full book and both intended completion gates remain open.
## Full-mask Hecke type-I scalar expansion
Current book177pages/35chapters/18diagrams/11unique labs; focused articles35/75
pages (audit20/49 unchanged). Seven new complex local/finite Euler restoration
statements and the fresh342-report suite pass,340standard-only/two axiom-free.
Independent focused compilation and literal theorem-type review accepted.
The exact good-ideal checker covers1488cases/93rows/128squarefree columns over
7split/inert primes,78016nonzero terms and29120originalzero encounters.
An independent no-production-import oracle checks288local and3072support cases.
The full primitive Euler ratio, residue, scalar Mellin estimate, actual-beta
aggregation and explicit losing energy envelope have reviewed prose proofs.
Classical abelian reciprocity and Hecke continuation/functional equation,
polynomial strip growth, Stirling, Mellin inversion and ideal/lattice counts
are external dependencies; infinite Hecke objects and contours are not Lean
formalized by the seven finite identities. The same full deleted set is kept
when squaring reduces primitive conductors, including good valuations3and6.
Profile seminorm uniformity does not interchange a scale/profile supremum
with the row mean square; shrinking scale with fixedU requires the short
Vaughan term again. TypeII/signed low-gcd averaging, positive determinant
comparison and GRH remain open. Independent review:
`planning/reviews/hecke-type-i/REVIEW.md`; exact source pins:
`planning/HECKE_TYPE_I_SOURCES.json`.
## Prior milestone: ambient normal-image faithfulness
At the ambient-only milestone the book had178pages/35chapters/18diagrams/11unique laboratories;
focused articles35/76pages (audit20/49 unchanged). Eight new declarations
in BranchExtension.NormalPath prove slot extensionality, positive path
length, the length equation, recovery from start and length, selector
biorthogonality, arbitrary finite-support coefficient recovery, actual
LinearIndependent and injectivity of the Finsupp expansion. These use
actual unbounded polynomial matrix pairs over every field. That350-report suite had347standard-foundation only andthree axiom-free.
An independent reviewer freshly compiled the final source with warnings
as errors and checked literal theorem types and both TeX proofs. The
production exact checker verifies475normal keys through degree1000003,
225625selector pairs,1030arrow steps,15625products and200finite expansions
over Q/F2/F3/F5. A separate stdlib oracle checks525keys/275625pairs.
These finite diagnostics complement, rather than replace, the universal
Lean proof. Source and finite exact evidence load inside the existing
canonical relation lab; no duplicate iframe or replacement of the
infinite algebra has been introduced.
At that milestone generated-subalgebra membership and spanning remained
open; the next subsection records their discharge. A separately defined
normal-path quotient algebra and categorical/AR universal properties remain
further formalization obligations. No claim of a full
book formalization or arithmetic positive determinant/GRH theorem is made.
Review: planning/reviews/normal-path-basis/REVIEW.md.
## Actual generated-algebra normal coordinates
Current book180pages/35chapters/18diagrams/11unique laboratories; focused
articles36/78pages, audit20/49unchanged. Ten further Lean reports construct
all permitted images as actual shortest arrow products times arbitrary
cycle powers, prove the normal span equals the generated subalgebra, and
prove actual A-valued finite expansion bijective with explicit inverse
selector coordinates. `normalEquiv` is a genuine linear equivalence of
the infinite generated algebra; degrees and finite supports are unbounded.
The full360-report suite uses357standard-only reports andthree axiom-free.
Production exact checks construct50shortest arrow words and375preimages
through degree1000003, with5625product round trips and200expansions over
Q/F2/F3/F5 (187product examples nonzero);three forbidden ambient constant
shapes fail reconstruction. The independent no-production-import oracle
checks320constructive preimages,105625products,400generated expressions
and25invalid coordinate pairs. General membership/spanning/bijection is
supported by Lean, not extrapolated from finite samples.
A separately constructed mixed-zero path quotient and its universal
multiplicative comparison, completed order/lattice and categorical/AR
interfaces are still further formal work. No claimed full book
formalization, spectral positivity or GRH follows. Independent final
review and focused fresh compile are recorded in
planning/reviews/normal-path-algebra/REVIEW.md.
## Learned generative physics extension (2026-11-04)
Twelve trained hybrid diffusion/flow models across three seeds and one coherent
four-qubit state flow ran on the Linux RTX4090. Exact partial trace of the
seven-spin K2,5 Gibbs teacher defines the physical target. The four auxiliary
feature qubits are not its physical spins or a faithful order representation.
Matched controls show no entanglement advantage. Five additional inspector
records contain250 nodes;3752 scalar witnesses independently replay with
maximum error below2.8e-7. Known-teacher reverse-mixture control is separate
from the DDPM Gaussian reverse-kernel approximation.
The new chapter has complete prose proofs of the specified random-feature
RBF/readout limit, mixture score/velocity identities, actual genus-two
travel-order right-module resolutions and Hom/Ext/Tor computations, and
dual-number Frobenius gluing. Exact checks cover coefficient learners over
F2/F3/F5, finite cover trees, pants axioms and CHSH; source-pinned reviews
separate graph covers, ribbon surfaces, compact covers and coefficient fields.
These additions introduce no new Lean declarations: the previously checked
360-report suite is preserved, and their general homological/probabilistic
formalization remains open. Three digital accelerator reference models are
tested specifications, not synthesized chips. CFT/string/anabelian/HMS and
chemical/biochemical benchmark extensions retain explicit further obligations.
Approved GPU environment creation and failed probes are recorded; existing
Mac/Linux environments were preserved. The larger23-qubit circuit is untrained.
Its447 gates and8,388,608 amplitudes are a finite matter simulation, not an
arithmetic spectral construction. Both project completion gates remain false.
## Complementary dataset verification (2026-11-04)
Default records: Aster16,384/6,144 parents/99 components;
Mora22,951/517 parents/31 components. Static all-row spectra, MMFF energy and
force recomputation, geometry CIP, parent/component grouping and typed
condition/target projections pass. Maximum spectral discrepancy1.87e-14;
MMFF stored/recomputed differences zero. Mora cross-split Morgan maximum
.6470588235, nearest applicable geometry descriptor .1003620684Å; the declared
limits are .65 and .10Å. Giant peptide components are preserved.
Source anchors:576 Aster,266 Mora (74 PubChem-derived MMFF,192 PDB effective).
All train; independent anchors-vs-default-heldout chemistry and geometry audit
has zero threshold crossings. Maximum Morgan.320754717; no equal-element
geometry pair applies. Source atom order, formal charges, isotope/chiral/bond
stereo metadata and actual PDB HEADER/replacement provenance are preserved.
Force diagnostics are evaluation-only; selected PubChem anchors have zero
assigned E/Z bonds. This does not certify geometric E/Z for future corpora.
174 Aster and241 Mora electronic trajectories pass522/723 independent SciPy
expm comparisons. Twelve Mora NVE trajectories use .05fs steps for20fs;
recomputed forces/energies agree, maximum energy drift .00253903kcal/mol.
Read-only native checks reject Mora computed labels at its current observed
supervision gate and verify Aster condition-only tokenization. No model
enrollment, optimizer run or source-repository change occurred.
Sixteen actual web examples,14 full Givens density replays (max2.78e-15),
wavelet zero-sum and structure exports pass; desktop/mobile control tests
have no uncaught exceptions and390px viewport has390px scroll width.
Independent reviews and receipts: `planning/synthetic-training/review/` and
`outputs/checks/2026-11-04-synthetic-training/`. No new Lean coverage is claimed.
## Molecular, orbital and periodic generative extension
Parser/mesh checks cover actual1L2Y/1HRW/1AZU, covalent links, first-model
selection, source HELIX/SHEET and explicit peptide-link/TER negative controls.
Nine full-source atom/stick/cartoon views and mobile controls pass.
Five finite Schrödinger experiments have independent expm and Krylov/current
checks (max4.74e-14). Hydrogen radial norms, phase vortex/current and Lz
checks pass; the renderer distinguishes isosurface threshold from probability.
Eight2400-update CPU models retain actual weights/history. Literal Pauli
features match independent Kronecker calculations6.66e-16; parameter FD error
3.53e-10; wrapped score FD error2.55e-8; independent intrinsic poses match
RDKit4.44e-15Å. All8 browser models,5 propagation cases and4 WebGL orbital
modes pass. Diffusion reference is the correct positive-noise q_.03, with
exact-score64/128-step controls; no zero-noise recovery or advantage claim.
The torus conditional-field proof is independently reviewed, including smooth
positive density and uniform Lipschitz bounds. No new Lean declarations.
Evidence: orbital-verification.json, torsion-verification-v2.json,
orbital-chemistry-browser.json and planning/synthetic-training/review/.
## Analytic/game/gluing and spatial cloud checks (2026-11-04)
Current book212pages/37chapters,13unique canonical laboratories; focused
articles42/108pages. Both original source volumes remain byte-identical.
H2plus6-Gaussian Coulomb and Trp-cage20-orbital coherent fields preserve
full-space particle count analytically; independent integral/exponential
checks and8browser volume/isosurface/delta views pass. Quantized display
charge errors .00158/.00141 are explicit. These are different physical scopes.
Twelve actual q3/4/5 holomorphic readouts have four-part conditional uniform
budgets and specified infinite-feature/source-condition proofs. Floating
high-precision demonstrations are not interval certificates. Twelve trained
finite game runs preserve one stalled tilted classical solution.72 initial/
final/noise replay cases agree with stored scores1.34e-15, normalization
1.78e-15 and no-signalling2.23e-16. CHSH SOS, tilted attainment, magic-square
parity/classical enumeration and six dessin face/Euler cases have independent
review. Vector-state Gram positivity is not tracial microstate realization.
HoTT cell gluing is a conceptual reference, not a new Lean4 axiom.
New desktop/mobile browser checks cover12game runs,12analytic models,6dessin
cases and explicit source links with no JS exceptions. PDF/HTML are built in
staging; failed overfull/TOC attempts are preserved and did not replace live
outputs. New work adds no Lean declarations; prior360reports are not recast
as complete formalization. Evidence: outputs/checks/2026-11-04-synthetic-training/
{cloud-physics.json,density-cloud-browser.json,games-models.json,learning-browser.json}
and planning/synthetic-training/review/. GRH remains unresolved.
P11 continuation supplies exact rational growth, alias and Taylor constants
for the q3/4/5 completed Dirichlet examples. The proof uses pi>3 and a rational
exponential envelope; downloadable v4 certificates record center1/2, radius12/5,
source hash and strict Fraction/integer inputs. M_r<4/10/6 is exact, not a
floating estimate. Sample/optimizer rounding and analytic Lean remain open.
The book remains212pages, articles42/108pages. No new Lean declaration.
## Remote validated-oracle and masked-moment supplement (4 November 2026)
The immutable final-v5 certificate encloses all 192 true circle samples,
87 DFT/GD coefficients and three exact rational normalized polynomials.
Strict uniform errors are below 5e-15 on |s-1/2|<=7/10 for the three specified
conductors. Individual budget components are rounded upward before their
exact sum. Independent review covers 843 primitive cases, 1160 exact
recurrence identities, 24 negative controls, 1536 sample-component checks
and 144 compact-boundary diagnostics. Diagnostic high-precision agreement
is not substituted for analytic interval inclusion. The GD value is a
closed-form iterate; arbitrary requested accuracy and all-conductor rates
are not implemented. Producer and certificate hashes remain unchanged.
The fixed-twist AFM numerator moment, separately fixed-unitary extension
and exact full-mask remainder have independent adversarial prose/TeX review
with pinned primary inputs. Finite bookkeeping checks do not prove the
classical analytic theorems. D^(2delta)U^(2-4delta), principal residue, short
term and signed covariance remain explicit. No new Lean proof is inferred.
Existing pdflatex compiled both technical-note editions twice into v5 staging,
then clean logs/PDF headers/source hashes and representative proof-page
renders were inspected before promotion: 5/14 pages, no overfull boxes or
undefined references. Prior builds/renders and replaced technical-note aliases
remain preserved. Receipt: outputs/checks/2026-11-04-validated-oracle/technical-notes-promotion.json.
Final independent acceptance: review/final-acceptance.json and
arithmetic-adversarial/afm-tex-review.json in the same check namespace.
Three source pins and clean source worktrees, both original manuscript hashes,
old PDF/native-MathML integrity and unique13labs pass the existing integrity
checker (report redirected to this new namespace without replacing its old
receipt). Existing full book/HTML/articles were not rebuilt or promoted here.
Linux currently lacks make4ht/latexmk/tikz-cd and pinned Lean4.24; no new
dependency or toolchain installation was made. Prior360 Lean reports retain
their earlier scope and were not rerun as a Linux proof check. Full integration
and analytic/categorical formalization remain unfinished.
The previous public rational-constant asset deployment is now verified live,
byte-identical and tested across all three conductors with zero browser
exceptions; this is distinct from the new unpublished interval certificates.
Evidence: outputs/checks/2026-11-04-validated-oracle/publication/.
## Weighted-cutoff and integrated edition milestone (5 November 2026)
The reviewed original-family masked type-I remainder now has the U-independent
bound ||E||² <= C_(nu,S,epsilon) ||W||_infinity² (HD)^epsilon
[H + D H^(1/3) + (HD)^(2/3)], for U²<=D, fixed finite-order nu/S and
one common smooth profile. Recombination retains every original zero mask,
primitive drop, principal residue and literal element multiplicity. The proof
uses the pinned classical BGL powered large sieve and has independent
adversarial prose/TeX review; it is not a new Lean proof. At H=D^(3/4),
a relative D^(1/2) loss remains. It does not prove the original signed short-row
estimate, the needed II-I covariance or the arithmetic positive-determinant bridge.
The exact cutoff series has an explicit U^(-kappa+epsilon) convergence rate
on Re(s)>=1+kappa. Locally uniform convergence of the actual series on
Re(s)>1/2 would force zero-freeness for a fixed nonprincipal character;
that stronger convergence is an unproved hypothesis, not a GRH proof.
See research/weighted-cutoff-moment.md and papers/weighted-cutoff-*.tex /
uniform-cutoff-moment-*.tex. Exact recombination diagnostics preserve raw
sextic zeros and reject the naive unconstrained Mobius square.
Accepted staged artifacts are now promoted on Linux: book233pages/40chapters,
focused articles49/129pages, technical notes8/20pages; native HTML contains
6216MathML expressions and13unique laboratories. All5356oldMathML subtrees,
384oldtables, both source volumes and all13oldiframes are unchanged. Existing
vector assets were source-hash checked; the two article-only grids preserve
all nodes, arrows, hooks, signs and labels. The bounded remote builder uses
installed pdflatex/Poppler, Pandoc3.10 and librsvg/cairo, without installing tools.
Fresh reproduction and exact-source transformation audits pass.
Firefox layout checks at320/390/768/1440 preserve full18px equations in
focusable horizontal scroll containers, with no page overflow. This corrects
an inherited MathML overflow issue. Stable dark-mode contrast is14.54:1;
representative PDF/grid and desktop/mobile/night renders were inspected.
Three malformed old TeX4ht munder nodes remain explicitly inherited; the860new
expressions have zero such defects. Article logs retain existing atop/over
and seven underfull-box warnings per edition, with no overfull boxes or
unresolved references. No new Lean declarations; prior360reports retain
historical scopes and were not rerun with an unpinned Linux toolchain.
Evidence: outputs/checks/2026-11-05-weighted-moment/promotion/promotion.json;
outputs/checks/2026-11-04-weighted-moment/{review,native-browser-v9};
outputs/article-integration-20261104/audit-v1/source-preservation-audit.json;
planning/book-integration-20261104/MOBILE_QA.md and
planning/article-integration-20261104/HANDOFF.md. Earlier failed builds and
replaced aliases are preserved. Both completion gates remain false.
### Independent signed-frontier continuation (5 November 2026)
The next exact reduction is accepted at derivationSHA7dc97b2e61a8d3bc286356bf9eae28fb986d29978582a97f9bfbfd7e329ab9c1.
Principal subtraction vanishes on the balanced a1*a2² nonprincipal block.
The raw Gram retains symmetric-difference oscillation and the union zero mask.
On the full active annulus A_U=C_U-mu, exposing the precise signed covariance
lower bound BT_sigma; neither a saving nor an energy lower bound is proved.
Independent224coefficient,2endpoint,75776raw-product,1296complex-covariance,
18Gram and54signed-target controls pass. Radicand valuations1/2mod6 are
distinct from tame conductor exponent1; this wording correction is documented.
Source/review: planning/weighted-moment-20261105/{frontier,review}.
Both gates remain false. Do not repeat centering or mask exclusion as a new
analytic solution; attack the signed covariance with all required uniformities.
## Signed covariance and two-cutoff frontier (5 November 2026)
New shared proofs are papers/signed-bilinear-{core,details}.tex in chapter40,
with executable examples/arithmetic_cutoff/signed_covariance.py and
research/signed-cutoff-covariance.md. In the window2D/7<=U²<=D, every actual
active coefficient is mu(m)(r_U(m)-1), where r_U is an even ordered-factor
count. Thus it is odd/nonzero. Elementary lattice/squarefree counting gives
an explicit positive lower constant for its normalized coefficient norm,
for each fixed nonzero smooth profile. This is not a row-energy lower bound.
A finite example in the ORIGINAL sextic family, with D=400,U=17,nu=1,
all72literal B7 rows and the complete four columns of norms439/475, has
Q(C,mu)=-27/25 despite nonnegative pointwise coefficient overlap and a
positive-definite real Gram. A genuine common nonnegative smooth profile
has exact integer-norm values10/1. A second guaranteed nonprincipal246-row
subset atH89 preserves raw zeros and gives-69/50; it is not the full
nonprincipal family. Independent direct F439/F19/F25 exponentiation
reproduced1272raw characters, full complex Gram matrices and all12amplitude
cases. These are exact finite sign counterexamples, not asymptotic BT_sigma
or GRH counterexamples. No lower bound for arithmetic row energy is inferred.
The high-window bilinear attempt retainsDX after legal factor Cauchy/BGL;
the older regrouped estimate is stronger. The exact two-variable Perron
factorization, normal correction domain, half-integer sharp cutoff and
finite-height error are proved. Safe reciprocal-L contours reproduce the
oldU^(2-4delta)energy cost, with M4(W)^2 and fixed margins retained. Moving
the contours requires actual reciprocal/zero-residue control; numerator
moments do not justify dropping those terms. The coordinate slice is a
meromorphic identity, including the stated numerator-zero caveat.
Independent proof and final-byte TeX acceptance is under
planning/signed-bilinear-20261105/{review-covariance,review-perron}; exact
reports are outputs/checks/2026-11-05-signed-bilinear/. The new book is
237pages/40chapters/6412nativeMathML, with13unique laboratories. Focused
articles51/133pages and paired notes9/24pages are synchronized. Three-pass
builds, source audits, XML/numbered-reference checks and320/390/768/1440px
browser checks pass. Paragraph headings now match the unnumbered PDF style.
Old source volumes, diagrams, data and environments are preserved. Earlier
overfull/undefined-proposition staging failures are retained; valid outputs
alone are promoted. Inherited amsmath/underfull and3old munder warnings stay
explicit. No new dependency, training/GPU job, Lean declaration or GRH proof.
Next: prove a QUANTITATIVE lower covariance budget or coupled arithmetic
fourth moment that permits the observed negative covariance, uniformly in
the original masks, conductors, cutoffs and common profile; cover neighboring
near-saturating blocks. Alternatively control the actual two-variable
reciprocal integrals and all crossed residues. Do not retry automatic
covariance positivity, coefficient shrinkage, marginal independence or a
pole-free contour shift as if these missing inputs had been established.
The best global envelope here still losesD^(1/2) atH=D^(3/4); short/principal
assembly, analytic/category Lean and the positive determinant bridge remain
open. Both completion gates remainfalse; the existing heartbeat stays active.
## Requested-error training and lower covariance budget — 5 November2026
Independent exact readout acceptance: outputs/checks/2026-11-05-adaptive-readout/review/final-acceptance.json.
It binds final-v3 producer/proofs/certificates and checks192 DFT rectangles,
192degree budgets,666updates/18,802states,68negative controls and7tests.
Separate166Fourier/288boundary comparisons are diagnostics, not proof.
Refusals at/below the fixed certificate floor and strict-slack termination
are tested; actual optimization is the stated rational midpoint objective.
Independent arithmetic review is under planning/covariance-budget-20261105/review/
and its exact-v1 receipt. The high-window O(DlogD) coefficient moment and
negative covariance budget are proved in prose; the original-family coupled
row saving and GRH remain open. No new Lean declaration was run or asserted.
Candidate source-preserving builds pass at243pages/6601MathML, focused53/138
pages, notes11/30pages. Old6412math subtrees are ordered/unchanged;19source-
matched vector diagrams,13canonical embedded labs and both originals stay
preserved. Focused browser controls/refusal/source links and four viewport
widths pass, with representative PDF and stable-night views inspected.
Inherited amsmath/underfull/3old malformed munder diagnostics remain scoped;
there are no new overfull, unresolved-reference or malformed MathML failures.
## Adaptive theta/family checks — 5 November 2026
Governing scientific receipt: outputs/checks/2026-11-05-adaptive-theta/review/final-acceptance-v2.json
(and sampler-final-acceptance.json). Exact primitives/tails/guards/source
captures,192 samples at eta = 1e-32 and full1e-24 family certificates pass independent
review; sampler 7,241 checks / family 8,666 checks, plus 17 focused unit tests.
All 590 updates / 593 states / 29,057 complex coordinates reconstruct exactly from
lossless rational records. High-precision comparisons remain diagnostics.
The variable Taylor-order convergence addendum is separately checked and
source-bound; no formal Lean theorem is claimed.
Arithmetic acceptance: outputs/checks/2026-11-05-unequal-product/review/acceptance-v1.json.
It accepts only the conductor/mask correction and novelty addendum, not a
new small-conductor/global power saving. Full primitive/raw zero scopes and
known high-gcd overlap are explicit.
Source-preserving staged builds: book 251 pages / 6,890 MathML, focused 56/146 pages,
notes 14/38 pages,40chapters/13labs / 19 source-matched vector diagrams. All 6,601
old math subtrees remain ordered and unchanged. Initial MathML parse/layout
warning is diagnosed by preserving final-DOM and controlled hidden-layout
evidence; the final v5 HTML passed strict browser checks before promotion.
Inherited 3 malformed TeX4ht munder nodes remain isolated; no theorem or zero-
location result is inferred from rendering/numeric checks.
Final promotion: `outputs/checks/2026-11-05-adaptive-theta/promotion/promotion.json`.
The v5 browser record confirms no new page errors across four viewport widths.
The parse guard restores formulas even if the external layout helper is blocked;
its lifecycle fallback has independent checks. All 6,601 previous MathML subtrees
are preserved. Three inherited malformed munder nodes remain disclosed.
## Quadratic-pivot and joint-dual checks — 5 November 2026
Independent canonical proof bytes: outputs/checks/2026-11-05-joint-dual/review-quadratic/production-acceptance-v3.json and joint-gauss-production-acceptance-v2.json. GL v2 PDF SHA matches the previous5cc5f... pin; its theorem is an external analytic input. Root3167 exponent/slack assertions, author and independent991 exact cyclotomic/CRT/orientation assertions and inspector1037 numerical cases pass. These do not prove a signed incomplete estimate. Scoped renderer paragraph-label check is separately accepted; lower numbered headers remain strict and qp:near is unnumbered as in the PDF.
Staged PDF pages255/57/150/15/41 and native7065MathML pass source-preserving builds. All6890old math subtrees remain ordered/unchanged;13canonical labs and19SVGs retained. Failed staging outputs remain preserved. Three inherited malformed munder nodes and old amsmath/underfull warnings remain scoped; no new Lean declaration is claimed. Production promotion and final browser receipts must be read separately before calling the candidate delivered.
Final native v4 browser QA passed with zero new page errors. Promoted only checked v3 PDF / v4 native candidates; root inspected PDF pages243/246 and inspector desktop/mobile-night. All6890old math subtrees preserved. Receipt: outputs/checks/2026-11-05-joint-dual/promotion/promotion.json.
## Ordinary conductor refinement and finite arithmetic learning — 5 November 2026
This accepted frontier supersedes earlier descriptions of5/4 as the best
available global unsigned exponent. A classical ordinary all-primitive-character
comparison gives Lambda_A << (HD)^epsilon A6[D+(A1 A2 A3 A4 A5)^2],
with the arbitrary overlapping r6 retained as its actual coefficient mask.
Combining with the powered-family bound and exact capacity yields
H + D H^(1/6) + (DH)^(2/3) for the original residue-subtracted energy,
with fixed nu/S and one common possibly complex W. At H=D^(3/4), power7/6
improves5/4 by1/12 but still exceeds the target3/4 by5/12.
The whole old A1=A2=A,A3=B,AB=H^(1/3) face now costsD; it is no longer
the worst upper-bound face. New upper-envelope witnesses are
(log_D A1,log_D A2)=(5/12,1/6) and pure log_D A1=3/4.
They are not actual-energy lower bounds. The proof includes all-modulus
primitive Gauss completion, nonunit vanishing, reduced Eisenstein fractions,
a lattice large sieve, raw conductor/zero transfer and principal assembly.
It is a refined application of classical sieves, not a new classical sieve.
Independent exact-production proof receipt: outputs/checks/2026-11-05-arithmetic-learning/review-mixed/acceptance-v1.json.
A separate finite Fourier-feature learner targets all210 actual original
coupled correlation coordinates,95(h,q) blocks including65 h1 blocks.
D6000/U60/H625 and one stated common smooth four-bump profile activate all15
squarefree ideal columns; principal nu,2262 literal rows and a separately
labelled2148 guaranteed nonprincipal subset retain4530 raw column zeros.
Four tau1e-12 runs execute216 metric-gradient updates and200 budget-limited
ordinary updates, saving every exact rational state. Both signed observables
and the full-vector tolerance are certified; weighted spectral loss and
metric/ordinary gradient conventions are distinct. Exact labels are supplied
before training, so no speedup, unseen-conductor generalization or uniform
analytic saving is asserted. Immutable v2 manifest binds source/input/output.
Independent acceptance-v2 in outputs/checks/2026-11-05-arithmetic-learning/review-learning/
reconstructs31668 symbols/450 Gram targets/186552 histogram cells and every
saved update/certificate. The failed logging-only v1 remains preserved.
Next substantive arithmetic research must estimate the actual signed joint
conductor/dual-frequency weights before per-conductor absolute Cauchy,
including h1 and the new remaining upper-envelope neighborhoods, with all
coherent fibers, raw common-prime zeros, ray/squarefree/capacity restrictions,
fixed profiles and conductor uniformities. A finite learned Gamma certificate
is not this varying-family saving. Short/principal target assembly, independent
positive arithmetic determinant/zero-location comparisons and scope-specific
Dirichlet/Hecke/Dedekind/elliptic/Artin obligations remain open.
No new analytic/category Lean declaration, PSD or GRH proof is supplied.
Pinned Linux4.24 compiler approval remains pending; do not install/substitute
or repeat the completed inventory without an explicit reply/changed state.
Keep both completion gates false and the existing heartbeat ACTIVE.
Evidence and reproduction: research/mixed-conductor-and-learning.md;
shared papers/mixed-conductor-{core,details}.tex and
arithmetic-learning-{core,details}.tex; examples/arithmetic_learning/README.md.
Do not rerun accepted author/reviewer grids or frozen216/200 traces as new
research. Preserve all earlier sources/runs/releases and original HEAD/index.
New shared fragments extend chapters38/40 and every paired edition. Staged
book261pages/40chapters/7352MathML, focused59/157 and notes17/48 pass final
source-preserving builds. All7065 previous MathML subtrees remain unchanged
and ordered;13 unique canonical iframes and19 source-matched SVGs remain.
Nativev6 / PDFv5 pass four-width/night browser QA: all7352 formulas visible,
zero new page errors or malformed nodes. Three inherited malformed munder
nodes and old amsmath/underfull diagnostics remain disclosed. Failed shell,
undefined-reference and overfull candidates are preserved. The final proof
bytes and full source bindings govern, not those failed candidates.
Final PDFv5/nativev6 and paired article/note artifacts are promoted after
source-bound build/browser acceptance. Previous production bytes are preserved
under outputs/checks/2026-11-05-arithmetic-learning/promotion/previous/;
promotion.json is the authoritative18-file manifest. Independent integration
checks compare all7065 prior MathML subtrees against7352, rather than only the
older5356 baseline. Root independently inspected physical PDF pages217/248
(printed209/240), desktop-night replay and mobile layout. Final integrity
checks pass in root/project-verification.json; publication remains a separate
verified transition. No new Lean or GRH proof is asserted.
Final diff corrections: landing-page source now states261/7352 and public
status; four article/note abstracts now distinguish the stronger unsigned
comparison from the open signed estimate. These exposition-only fixes rebuilt
concise-v3/extended-v5 and notes-concise-v3/notes-extended-v4 with unchanged
59/157/17/48 page counts. Governing current complete artifact promotion is
outputs/checks/2026-11-05-arithmetic-learning/final-diff-fixes/promotion/promotion-v2.json;
the first promotion and every previous byte remain preserved. BookPDFv5,
nativev6, all4 mathematical fragments and frozen learner bytes are unchanged.
Root integrity passed again in root/project-verification-v2.json.
## Moving-profile and coherent-mean boundary — 5 November 2026
For principal nu=1, fixed S above2,3 (or a fixed larger S), and U²=D,
the actual norm-radial A_U and literal epsilon=1+omega phase admit ONE
common complex smooth profile W_D with supnorm1. It varies with D, has
derivatives O_k(D^k), and vanishes at the annulus endpoints. On actual
nonprincipal rows epsilon*beta_v^6 the primitive conductor is432, with
power conductors432/27/16/27/432; every original S/v collision remains
deleted. The ideal v is arbitrary, including repeated/overlapping primes.
An elementary positive-density coprime-triple proof gives full original
E-energy >=(c_S+o(1))D H^(1/6), H=D^theta,0<theta<1. This uses outer
row-energy positivity and a SELECTED literal class's h1 A-A contribution;
it does not lower-bound the signed h1 sum over unselected rows.
At theta3/4 lower power9/8 is below the current upper7/6. No new unsigned
upper saving, fixed-W counterexample, positive determinant or GRH follows.
The DH^(1/6) term is necessary in the varying-profile supnorm class for
theta<2/3. A target uniform through M_k(W)^alpha for every epsilon would
need k*alpha>=1−5theta/6; this does not refute existing derivative costs.
The actual orientation projection exactly returns the coupled Gamma norm;
rank supplies no conductor saving. Coherent restriction has compression
3^(-omega(q))*11* and leakage lambda(1−lambda), so it cannot commute with
projection for free. An exact original D4000/U63/q-five-prime fixture has
four active orientations, Gamma=3 and variance0, but selected h1AA=3/1000.
Omitting the literal unit phase incorrectly gives−3/4000. This rules out a
variance-only centering rule without its mean term on nonprincipal rows.
The A6=1 mixed/pure upper-envelope sectors remain open.
Sources: papers/profile-uniformity-{core,details}.tex;
research/profile-uniformity-and-coherent-means.md; examples/profile_uniformity/.
Design review and exact joint note: planning/signed-profile-20261105/.
Independent implementation acceptance-v1 reconstructs all30482 norm groups,
1219280 raw-mask entries,40 distinct arbitrary ideals through norm64, and
14720 direct literal symbols across the six frozen scales. The governing
v2 manifest is a90d1c356c276b36430b58e342ea7aee6dfe6cb33c0d7db0a4ffcb5ab17bc460.
The fixed-X64 good-S family has19 rows; the separate capacity-matched small
caps2..4 contain only the good-S unit. Finite ratios do not prove the
asymptotic theorem. The new standalone profile-uniformity.html inspector
is a hash-bound exact-data view, not browser training. Preserve failed v1
manifests/builds, all earlier sources/outputs/releases and the fixed-profile
frontier. No new Lean declaration or compiler/dependency change occurred.
Next: bound actual signed h,q,K weights BEFORE per-conductor absolute Cauchy,
retaining coherent means, fixed common profiles, full union/common-prime
masks and conductor/ray/squarefree/capacity uniformities, especially h=1
and the A6=1 remaining neighborhoods. Short/principal target assembly,
independent positive determinant/zero-location and scope-specific GRH
obligations remain open. Pinned Linux4.24 approval remains pending; do not
install/substitute/reinventory without explicit reply or changed state.
Both completion gates remain false; keep this existing heartbeat ACTIVE.
Final source-bound production promotion: outputs/checks/2026-11-05-signed-profile/promotion-final-v2/promotion-v1.json. Only checked bookPDFv3/nativev3 and focused concise-v1/extended-v2 plus notes-concise-v1/notes-extended-v2 promoted. Current book266pages/40chapters/7605MathML, all7352 previous subtrees byte-identical/in-order,13canonical iframes and28SVG references/assets preserved. Focused61/162pages and notes19/53pages pass. Independent production-proof acceptance-v1 and build-acceptance-v1 govern; UI frozen-receipt plus final browser-20261105T172508737117 and root native-qa-v2 pass. Root inspected new proof pages and desktop/mobile/night inspector screenshots. Three inherited munder nodes and old amsmath/underfull warnings remain; no new malformed MathML, unresolved reference, overfull box or Lean declaration. The failed mathscr first build, earlier wording candidates and two schema-mismatch promotion guards are preserved; the guards stopped before any promotion. The successful script's terminal summary prints the historical directory name; the actual record above is authoritative and binds15changed artifacts. Original source HEAD/index and all earlier outputs remain preserved. Publication is a separate verified transition.
## Signed coprime slice and exact character twist — 5 November 2026
The actual signed h=1 slice on the A6=1 two-pivot row blocks now has a
separate unconditional baseline. Gcd–Mobius inversion retains the actual
X(da), common complex W, nu, full common-prime raw zeros and every original
row/capacity/ray mask. Positive subsequence norms may be enlarged only
inside the exact divisor decomposition. With R=A1*A2, M=max(A1,A2),
|S1(X,Y)| << (HD)^epsilon ||W||infinity^2 min{
R+DR/M+D^(2/3)R/M^(1/3), D+R^2}, X,Y in {A_U,C_U,mu}.
The L=D/Nd in [1/2,1) endpoint is counted directly, not passed to a sieve
outside its domain. At H=D^(3/4), the remaining mixed and pure upper-bound
neighborhoods still cost D^(7/6). No new power saving follows: the sufficient
A-A signed divisor budget sum mu(d)Gd << D H still lacks 5/12 powers.
It is sufficient for this slice, not asserted necessary for all later signed
assembly. Principal/residue assembly and all other row sectors remain separate.
The actual Jacobi column gauge is mu(m)j(m)=Psi(m) on squarefree columns.
Psi is the normalized Eisenstein generator character of exact conductor
(2(1+2omega)), norm12, infinity type(1,0). Its unitary normalization is
infinite order, so the earlier finite-order moment input does not automatically
apply. The exact TWO-factor completion has central quadratic Gauss phase
and one orientation-dependent Jacobi phase; column gauging and changing the
additive generator preserve the matrix norm and do not estimate the kernel.
The central quadratic phase law is inherited, not a new theorem of this run.
A gauged Euler product has reciprocal L(s,Xi); no pole-free contour shift or
unproved reciprocal moment is used. Actual q-dependent Fourier/capacity
weights remain outside that auxiliary fixed-twist series.
A NEW fixed common smooth plateau W, independent of D, at D60000/U244,
H3833/A1=37/A2=7 gives exactly20 original nonprincipal rows after masks.
For TWO selected columns and their complete two-orientation q fiber,
Gamma=-3 +/- i*sqrt(3), selected h1AA=-1/10000. Centering gives zero pairing
but nonzero centered squared norm6 (average variance3); the omitted mean
has squared contribution36 against naive variance budget12. The complete
selected two-column Gram is positive with eigenvalues17 +/- sqrt(21);
its coprime off-diagonal slice is indefinite. This finite remaining-sector
control is not an asymptotic lower bound, whole-family covariance or GRH
counterexample. The inspector exposes all20 rows, actual zeros, phase gauge,
and all18 signed pairings. No browser training or new data inference occurs.
Sources: papers/signed-weights-{core,details}.tex and
research/signed-coprime-weights.md; independent phase/budget/Mellin/diagnostic/
canonical reviews under outputs/checks/2026-11-05-signed-weights/.
The new core/details are shared verbatim with chapter40 and all four focused
wrappers. All prior moving-profile, learning, fixed-family adaptive theta,
source distinctions and scope limitations remain valid. No new Lean theorem,
conductor-uniform theta construction, determinant bridge or zero-location
result is asserted. Pinned Linux4.24 approval remains unanswered; do not
install/substitute/reinventory or ask again absent an explicit reply/change.
Next: estimate the actual mean-retaining signed joint h,q,K weights or the
signed gcd divisor sum before absolute Cauchy, especially h1 and the mixed/
pure A6=1 neighborhoods. Preserve full union/common-prime zeros, coherent
fibers, conductor, literal-ray, squarefree/capacity and common-profile
uniformity. Do not rebrand this D^(7/6) baseline, the exact character gauge,
finite centering control or reciprocal Euler identity as the missing signed
saving. Short/principal assembly, independent positive determinant/zero
comparison and separate Dirichlet/Hecke/Dedekind/elliptic/Artin obligations
remain open. Both completion gates remain false; keep the existing heartbeat.
Final signed-weights promotion: outputs/checks/2026-11-05-signed-weights/promotion-v1/promotion-v1.json binds15promoted artifacts from bookPDFv1/nativev1 and all four focusedv1builds. Delivered272pages/40chapters/7891MathML, all7605previous literal mathsubtrees byte-identical/in-order;385oldtables,13canonical iframes and28SVGrefs/assets retained. Focused62/168pages; notes20/59. Independent review-integration/acceptance-v1 and rootnative-qa-v1 pass; all20newrows/18pairings and preserved oldcontrols pass desktop/mobile/night/keyboard. Root inspected new proof PDF and matrix screenshots. No new malformed node, overfull box, unresolved reference or Lean declaration. Three inherited munder nodes and unchanged oldamsmath/hyperref/underfull warnings remain disclosed. The native record's2535math count is its new-chapter segment, not the complete7891formula book. Original sourceHEAD/index, oldoutputs and all prior releases are preserved; publication is separate.
## Mean-retaining signed divisor lift and complete-column reference — 5 November 2026
The exact original coprime slice now has an arbitrary-coefficient divisor
incidence formulation: D Q1(x,y)=sum_d mu(d) Z_d(x,y), with
F_d(x;u)=sum_(d|m) x_m chi_m(u) and
Z_d=sum_u w_u F_d(x;u) conjugate(F_d(y;u)). The original coefficients
include nu and the ONE common complex profile. Keep full divisors d=m,
actual row zeros, common-prime deletions, literal/ray and capacity masks.
Residual a,b may overlap. There is no extra mu(q) in this identity.
The positive norm majorant B(m)=sum_(d|m) lambda_d gives an exact finite
coefficient-error certificate. Computed Frobenius lambda_d values are
weighted nonzero matrix-entry counts, not an analytic signed saving;
classical uniform lambda_d inputs are asserted only for original unweighted
rows. The absolute divisor triangle still discards the needed cancellation.
For fixed residuals a,b and row u, the ACTUAL signed d series has local
factor 1-q^(-s)(1-q^(-v1)-q^(-v2))(1-q^(-w1)-q^(-w2)). Its exact factorization
has NINE Dedekind zeta factors (four numerator, five denominator) times a
normally convergent correction, with explicit half-integer cutoff,
finite-height errors and successive Perron limits. An independent finite
cutoff expansion keeps all EIGHT nonzero short-divisor states and gives
Z_U(s)=N_U(s)/zeta_K^T(s), with N_U holomorphic for Re(s)>0. Its elementary
bound retains U^(4 max(1-sigma,0)+epsilon), hence the old U^(2-4eta) cost
on sigma=1/2+eta, 0<eta<1/2. Shared-profile Mellin inversion embeds the
actual finite coherent outer sum over a,b,u. Neither this identity nor the
normal correction bounds its reciprocal zeta factors or justifies discarding
crossed residues. An explicit overlapping-residual control detects the
nonzero short-state correction. No uncancelled pole of the complete coherent
sum, new power saving, determinant comparison or zero-location result follows.
A NEW exact full-column evaluator uses all annular squarefree IDEALS, not
rational-squarefree norms. The fixed study profile is chosen ONCE with
width delta=1/240000, support [1,2], value one at all strict integer nodes
for D<=60000 and zero at endpoints. It is distinct from the older broad
plateau; its large derivative cost O_k(delta^(-k)) is explicit and the CLI
refuses D>60000. Different declared row blocks at D=144,576,60000 have
41/164/16953 columns, 53/212/21750 divisor buckets and 5/11/20 actual
nonprincipal rows. These three finite cases are not an asymptotic growth fit.
The largest immutable run retains 339060 raw column-row values (47908 zeros),
73420 incidences and 1305000 complex channel states. All nine signed cross
components are retained. In (A_U,C_U,mu) order the exact matrix is
[[-30714,-11287,19427],[-11287,2644,13931],[19427,13931,-5496]]/30000.
The A-A even/odd divisor masses 543416 and 604844 give -5119/5000 after
normalization by D. Among its columns,1361 are squarefree ideals with a
nonsquarefree rational norm. Independent ideal/finite-field arithmetic
reconstructs every large raw value, F/Z state and gcd-poset class; the naive
5748084180 pair-row visits are an operation-count comparison, NOT an executed
runtime benchmark. Small cases also have independent direct pair checks.
All original v1 scientific sources/tables stay frozen. A separately reviewed
v2 helper corrects ONLY the N=U=1 edge; every accepted column has N>U and
its data/provenance remains v1. No accepted tables were rerun for that fix.
The distinct signed-divisor.html inspector hash-verifies lossless compressed
payloads for every case and exposes actual columns, ideal incidences,
row/channel shapes and values, signed divisor contributions and all nine
pairings. It replays exact data; no browser training occurs. The unique
source-bound workflow PDF/TeX/SVG shows the actual dimensions and masses.
PDF and native MathML share the canonical papers/signed-divisor core/details
with all four focused wrappers. The new narrow table scroll container fixes
measured320px overflow while preserving full table/math bytes and font sizes.
Governing records: planning/signed-divisor-20261105/root/DIVISOR_INCIDENCE_LIFT-v2.md,
analytic-attempt/SIGNED_D_SERIES-v3.md, papers/signed-divisor-{core,details}.tex;
outputs/checks/2026-11-05-signed-divisor/review-design/ and independent large
review under analytic/large-full-column-review-v1/. The exact full-column
run manifests, fixed sources, all failed candidates and old evidence remain
preserved. The new finite profile is not a bounded-seminorm asymptotic theorem,
a whole-family counterexample, learned generalization, quantum/upstream
training, full Lean formalization or GRH proof.
Next research: use the actual short-state expansion or a proof-preserving
U-smooth/rough split to estimate the MEAN-RETAINING signed d sum/coherent
h,q,K weights before absolute Cauchy. Any rough-prime simplification must
first prove its support, endpoint and common-prime constraints; it has NOT
been established here. The current unsigned envelope remains
H+D H^(1/6)+(DH)^(2/3), with power7/6 at H=D^(3/4), still5/12 above target.
Keep coherent orientation means, all full-union/raw-zero masks and
conductor/profile/ray/squarefree/capacity uniformities. Short/principal
assembly, independent positive arithmetic determinant/zero-location
comparisons and separate Dirichlet/Hecke/Dedekind/elliptic/generalArtin
obligations remain open. Pinned Linux4.24 compiler approval is unanswered;
no installation, substitution, repeated inventory or repeated question.
Actual analytic/category Lean and both completion gates remain unfinished.
Keep the ONE existing heartbeat ACTIVE; no guaranteed proof schedule.
Final signed-divisor promotion: outputs/checks/2026-11-05-signed-divisor/promotion-v1/promotion-v1.json binds the accepted bookPDFv1/nativev3 and four article/notev1 builds. Current delivered edition279pages/40chapters/8176MathML, preserving ALL7891 prior literal math subtrees in order,385oldtables/13canonical iframes/28oldSVG refs. One new source-matched workflow SVG/PDF/TeX gives29totalSVGrefs. Focused64/175pages; notes23/66pages. Independent source/proof/data/UI/build preservation reviews and finalnative-qa-v3 pass. The failed narrow-table320px candidate and measured local-scroll probe are preserved; finaltable/math bytes and fonts remain unchanged. Root visually inspected workflow and new PDF proof pages. Three inherited malformed munder nodes and old amsmath/hyperref/underfull diagnostics remain disclosed; no new malformednode,overfullbox,unresolvedreference or Lean declaration. This is not a new signed exponent, determinant or zero-location bridge. Publication is separately gated. Protected sourceHEAD/index and all previous worktrees/evidence remain unchanged.
## Explicit rough-prime mask budget — 5 November 2026
A NEW quantitative A-A result replaces the rough-prime deletion by a defined
auxiliary kernel with an explicit correction. Work over the Eisenstein
ideals, with finite fixed S containing primes above2,3, original unitary nu,
ONE common complex W supported in[1,2], full original good squarefree annular
columns, and actual finite NONZERO literal rows1<=Nu<=H with weights w>=0.
Keep all internal row squarefree/coprime/unit/ray/capacity/sixth-factor masks.
The required window is U>=1,U²<=D,2D<=7U². It is not an all-cutoff theorem
or a theorem for arbitrary selected-column subsets/extra column filters.
Write d=f*e for the common IDEAL divisor, with f U-smooth/e U-rough; f is
NOT the original column-pair gcd h. Actual A_U(ez)=A_U(z), with NO mu(e)
inside A_U. Nonunit residual pair max(Na,Nb)>=7 permits at most one rough
prime; the unit pair a=b1 permits two DISTINCT rough prime IDEALS and
forces f1. The exact nonnegative C2 correction includes conjugate primes
of equal rational norm. It is not a lower bound for the whole signed slice.
Let K0 be the smooth-f signed completed norm, KT the original rough-prime
completed norm with full(p,abuS)=1, and KS its explicitly AUXILIARY S-only
rough-prime extension. All outer mu(f) signs, smooth masks and original
raw chi_a(u) zeros remain. With Rw=sum w, B=||W||infinity and kappa_U(u)
the number of distinct good rough prime ideals dividing u,
D*S1AA = K0-KT+C2,
Delta=(KS-KT)/D,
|Delta| <= B²[229376 sum_u w_u kappa_U(u)+2048 Rw],
0<=C2/D<=1792 B²Rw,
S1AA=(K0-KS)/D+E,
|E|<=B²[229376 sum_u w_u kappa_U(u)+3840 Rw].
If H<=D^(3/4), kappa_U(u)<=1 at EVERY scale, so
|E|<=233216 B²Rw<=2^18 B²Rw. For actual unweighted two-pivot rows with
bounded literal multiplicities, Rw<<A1*A2<=H; the correction is O(B²H),
without a divisor-power or profile-derivative factor. This bounds ONLY
the accounted correction, not the remaining signed main (K0-KS)/D.
The existing7/6 global envelope and missing5/12 saving are unchanged.
Principal/residue assembly and other coefficient channels remain separate.
The proof retains the exact disjoint row/residual prime union. Residual
prime deletion forces f1,a=p (and/or b=p) on support and gives exactly
2Re(T_1(P;u) conjugate(z_p(u)))-|z_p(u)|²; the last intersection term is
indispensable. The row term retains all raw zeros before any error triangle.
A support identity gives A_U(g)=mu(g)(2*1_(Ng<=U)-1) when good SF Ng<=7U,
including the unit. Active rough T terms have N(fa)<7U and therefore
|A_U(fa)|=1 exactly. This is not unrestricted multiplicativity, a proxy or
a shrinkage claim for the whole original annulus. Elementary ideal counting
and sum_f(Nf)^(-2)<=64 give the explicit constants; no prime theorem is
needed for the mask result. Both scaled comparisons are WEAK at B=0/Rw=0;
prior strict-product source/acceptance remains historical with a correction
addendum, never silently rewritten.
A separately reviewed coherent Stieltjes substitution retains Gf(t) as the
actual weighted row norm and mu(f) OUTSIDE, with exact lower boundary
-Gf(U)(pi_K^S(U)-Li(U)). Only the fixed-S prime measure is approximated;
growing deleted masks are already accounted by the correction. The analytic
Li/PIT subsection requires U>=7. Its operator variation budget retains
M1(W)^2 and a TRUE weighted norm input; the classical specialization is
ONLY the accepted unweighted/bounded-natural-multiplicity A6=1 two-pivot
family. At H=D^(3/4), U~sqrt(D), its inherited positive-operator majorant
has worst power5/6 times delta_S(U), not a new saving. The pinned fixed-field
Thorner--Zaman2019 Theorem1.1 supplies a subpower prime error after a fixed
threshold, retaining/absorbing any fixed exceptional zero. That upper budget
does not certify the3/4 target; actual signed errors may cancel and K0-J is
still unestimated. No actual-error lower bound, growing-conductor PIT,
pole-free contour shift, Artin holomorphy, positive determinant or GRH follows.
The DOI/publisher URL/cache hashes are in PUBLIC-CITATION-v1.json; full
publisher PDF/text/HTML caches remain local and excluded from publication.
Fresh deterministic controls use D200/U10/H53 with52columns5rows,
D512/U16/H107 with139columns7rows and D900/U30/H384 with250columns2rows.
The last is a declared noncritical capacity, NOT floor(D^(3/4)); its row
rough correction is exactly zero. A new fixed smooth W_delta,delta1/4096,
has exact flat strict norm nodes only on the stated integerD<=1024 scope
and explicit large derivative constants. It is not the previous profile.
All raw original/auxiliary masks, inert and conjugate primes, unit residual
states and non-squarefree auxiliary p² terms are retained. Raw
(K0,KS,KT,Erow,Eres,C2,D*S1AA) are respectively
(339,235,209,20,6,21,151),(366,781,711,48,22,89,-256),
(436,410,366,0,44,28,98). Normalized Delta values13/100,35/256,11/225
and actual-clean errors47/200,159/512,2/25 each divide by D ONCE.
These finite signs do not make Erow/Eres universally nonnegative.
Independent arithmetic reconstructs complete ideal universes/SF subsets,
3561 literal phase values (including auxiliary residual universes),3353
COMPLEX Fd transfers,199smooth/3154rough grid states INCLUDING exact-zero
placeholders,5119terms/39auxiliary nonSF terms and273767 ordered original
column-pair-row checks. All2510 active-profile reduced coefficients satisfy
the exact coupled short identity. A captured norm247=13*19 example rejects
its unrestricted use outside Ng<=7U. No old table/trajectory/evaluator was
rerun; fresh arithmetic/phase controls do not prove the analytic constants.
The compact exact summary web/data/rough-mask-controls.json SHA7596718f is
a projection of these immutable new inputs, with no browser training.
Governing canonical proposals f6de9534/a7d91548 match shared
papers/rough-mask-{core,details}.tex and book fragments. Independent gate
outputs/checks/2026-11-05-rough-mask/review/canonical-acceptance-v1.json
SHA89fa5333 binds analyticv5, zero/domain addendum, independent data/projection,
root PIT review and primary citation. Source/builds and all earlier failed
reviewer schema assumptions are preserved. Both originals, source clones,
all prior environments/outputs/cache/history and unrelated progress remain
untouched. No new Lean declaration, dependency, authentication or GPU work.
Pinned Linux4.24 compiler approval remains unanswered; do not install,
substitute, reinventory or repeat the question.
Next: estimate the ACTUAL clean mean-retaining signed main (K0-KS)/D, or the
coherent K0-J and signed prime-error cancellation, using the exact coupled
cutoff/Mobius coefficients and original row masks/conductors. Do not recast
this explicit correction bound, elementary unit-sector count or inherited
5/6 operator budget as the missing5/12 saving. Short/principal assembly,
independent positive arithmetic determinant/zero-location bridges and the
separate Dirichlet/Hecke/Dedekind/elliptic/generalArtin obligations remain
open. Actual analytic/category Lean and both completion gates remain false.
Keep the ONE existing heartbeat ACTIVE; no guaranteed proof schedule.
Final rough-mask promotion: outputs/checks/2026-11-05-rough-mask/promotion-v1/promotion-v1.json binds bookPDF/nativev1 and four focused/notev1 builds. Current edition288pages/40chapters/8552MathML; ALL8176previous literal math subtrees remain ordered/byte-identical, all386previous tables retained in388,13canonical iframes and29SVGrefs/assets unchanged. Focused67/183pages; notes25/74pages. Source-bound detailed proof, independent new ideal/phase/data checks, nativeQA at all widths and rootPDFvisual review pass; no new malformednode/overfull/unresolvedref/Lean declaration. Three inherited munder nodes and old amsmath/hyperref/underfull diagnostics remain disclosed. Exact zero/domain/full-column qualifiers supersede historical weaker/strict-scaled drafts without rewriting them. The explicit mask error has target size; the signed clean main, global5/12 saving, principal/determinant/zero-location bridges and full Lean remain open. Original sourceHEAD/index, all earlier releases/environments/cache/history and Macpreview remain preserved. Publication is a separate concrete gate.
## Physical-radicand sieve and coherent main — 5 November 2026
This substantive frontier supersedes older current7/6/global5/12 and
pure5/6-positive-certificate descriptions; their mathematical history stays.
NEW external input: Alexandre de Faveri, Optimal large sieve for fixed order
characters, arXiv2610.04045v1,2October2026, Theorem1.1. This is a pinned
submitted preprint theorem being invoked, NOT a new proof of that sieve,
a priority claim or complete Lean verification. PrimaryPDF SHA256:
97720b3cacccbedb9c2014520ad4d128ae23afd0a781bdeff27dab4e86b7c726.
Full primary reading caches remain LOCAL; public metadata pins the version.
The exact application uses K=Q(omega), containing mu6; mu12 is not needed
for Theorem1.1. Unique ideal generators alpha=1mod3 remove unit ambiguity
in the source's allowed fixed Section2.3 construction. BOTH indices are
sixth-power-free ideals. Physical radicand norm, valuations1..5, lower
character orders and all raw common-prime/union/ray/SF/capacity masks remain.
There is NO radical relabeling or arbitrary-family reciprocity premise.
Freeze arbitrary possibly nonsquarefree/overlapping r6 with its ACTUAL
coefficient collision mask. For original raw and residue-subtracted sums,
Phi(H,D)=H+D H^(1/6)+H^(5/6)D^(1/3)+H^(1/3)D^(5/6),
times B²(HD)^epsilon for fixed nu/S and ONE common complex W, B=supnorm.
The r6 sums use powers-6,-5,-2; the Dterm counts H^(1/6) ideals. All global
S-factor exponents are positive; principal residues/cutofflogs are retained.
For1<=H<=D every term is <=D H^(1/6). Critical9/8 improves previous7/6 by
1/24, still3/8 above target3/4. For fixed0<theta<1 and H=D^theta, the accepted moving-W_D obstruction matches
this unsigned SUPNORM-CLASS power ONLY when W may vary with D. It does not
prove fixed-W sharpness or refute suitable explicit-seminorm bounds/GRH.
For original UNWEIGHTED A6=1 two-pivot h1 rows, the exact gcd-Mobius
subsequence norms retain A_U(dn), every raw zero and direct subunit counts:
|S1AA| << B²(HD)^epsilon [H+D+H^(5/6)D^(1/3)+H^(1/3)D^(5/6)].
Critical13/12 improves former7/6 by1/12, still1/3 above target. The Dterm's
S-factor exponentzero uses a finite polylog count, NOT a convergent sum.
Under inherited high-window/capacity hypotheses,
|(K0-Kall)/D|<=|S1AA|+236800 B² actualrowmass.
This retains the combined mean; neither separate main is estimated at
target scale. Arbitrary positive weights require a TRUE weighted norm.
Previously reviewed research-only reductions are now integrated with full
proofs. KS=Kall+P2 has ORDERED(p,q),p=qallowed and onlyq's rowmask;
P2/D<=3584 B²rowmass differs from unordered DISTINCT C2/both masks.
Old233216 remains for the old main;236800 belongs to the new main.
Endpoint-exact two-jump Mertens summation keeps both closed-prefix terms,
fullT including common primes, and196608 M1²rowmass D/(UZ)m_(14DH)(Z).
Zcuts residual lengthX, NOT smoothf. Hypothetical scalar powers are not
proved; boundedX has unitf and marginal cancellation cannot bound the full
coherent outer sum. Euler deletion includes ALL supported prime powers;
only fixed-field subpower transfer follows, with fixed exceptional-zero
qualifications. The old sufficient smooth-f tail bound remains valid as a
coarse inherited localization, not current optimality or signed saving.
The joint Stieltjes operator keeps -E_S(U)Q(U), orientation means/equalnorm
fibres/rawzeros/fullsignedmatrix. Rough-prime positive compression is NOT
invariant and supplies no full energy lower bound. Radial Mellin Parseval
keeps equalnorm correlations. The current short positive input is the MINIMUM
with older valid bounds on1<=L<=D, directunitcount on1/2<=L<1. Pure budget
improves5/6 to19/24; mixed7/9 remains smaller. These are sufficient positive
MAJORANTS, not actual-error lower bounds. Keep the minimum INSIDE the
variation integral and the pointwise lowerendpoint.
CONDITIONAL diagnostic only: an all-scale/all-epsilon squarefunction
d2(T)<<T^(-alpha+epsilon) would imply zeta_K zero-free STRICTLY right of
max(1-alpha,1/2) by Mellin discrepancy continuation and primepowers.
Scalar inputs sufficient for pure/mixed majorants would be U^(-1/12),
U^(-1/18), yielding conditional strips11/12,17/18. NONE of these inputs or
strips is proved or asserted necessary for the actual signed target.
Canonical shared sources: papers/coherent-main-{core,details}.tex and
papers/physical-sieve-{core,details}.tex, exact book counterparts. Two
independent application reviews and one joint source-bound canonical review
keep the deep theorem external. The EXISTING quadratic-pivot inspector
shows currentglobal9/8/A6=1slice13/12, historical curves, suppressed unknown
constants/epsilon losses. The entire former cube face remains better
controlled at power1 by the old ordinary comparison. Displayed numbers are
bound expressions, not measured norms/energies/training/GRH.
Next: improve ACTUAL mean-retaining combinedK0-Kall or coherentK0-J PLUS
signedprimeerror/endpoint BEFORE per-f/per-conductor independent absolute
Cauchy, retaining exact generalA_U (NOTroughshortlemma onK0), fixedcommonW/
explicitseminorms, fullunion/commonprime/ray/SF/capacity masks, means/equalnorm
fibres and neighboring mixed/pure blocks. No target cancellation follows
from this new positive norm application. Short/principal target assembly,
independent positive determinant/zero-location bridges and separate
Dirichlet/Hecke/Dedekind/elliptic/generalArtin obligations remainOPEN. No
Artinholomorphy/GRH/desiredpositivity/zeros premise or new analytic/category
Lean. Linux4.24 permission stillUNANSWERED: no install/download/substitute/
reinventory/repeatedquestion. BothgatesFALSE; ONE heartbeatACTIVE. Preserve
sourceHEAD/index, all releaseworktrees/indices/originals/data/outputs/caches/
environments/history/Macworkspace/old8765preview/privateactive-modeldata.
Final physical-sieve promotion: outputs/checks/2026-11-05-physical-sieve/promotion-v1/promotion-v1.json binds valid bookPDF/nativev3, focused concisev1/extendedv2 and notes concisev1/extendedv2, plus inspected inspectorv3. Edition303pages/40chapters/9172MathML, focused71/200pages, notes29/89. ALL8552previous literal math subtrees remain ordered/byte-identical, all388oldtables remain in389,13canonical iframes/29SVGrefs/assets unchanged. Current external-preprint-based unsigned envelope has critical9/8 and original unweighted A6=1two-pivot signed comparison13/12; target3/4 is still open. Globalgap3/8/scopedsignedgap1/3 supersede former5/12. Newshortpositivecertificate19/24pure and min7/9mixed have only CONDITIONAL scalarinput/strictstrip diagnostics; no input or newzero-free strip is proved. Fixed0<theta<1 moving-profile supnorm sharpness is not fixed-W sharpness. Every original mask/commonprofile/mean/endpoint and arbitraryweightednorm qualification remains.
All6source-bound PDF/native builds, independent mathematical/application/integration reviews, focused exponent/model controls and UI/native desktop/mobile/night/keyboard/source/PDF/visual checks pass. No new malformedMathML/overfull/unresolvedreference/Lean declaration. Three inherited munder nodes and old amsmath/hyperref/underfull diagnostics remain; source-matched cachedPDF searchability/publicSVG limits unchanged. Failed hash-line/native-spacing/earlycapture candidates and all old outputs remain preserved. Both completiongatesFALSE; target signedassembly, independent determinant/zero-location and scope-specific GRH/analytic-categoryLean remainOPEN. Linux4.24approval stillunanswered; no newinstall/substitution/question. OriginalsourceHEAD/index, all releaseworktrees/indices/environments/caches/history/Macpreview/privateactive-data protected. Publication is separately gated; keepONEheartbeatACTIVE.
## Research-only intermediate-kernel evidence, 6 November 2026
No new book/PDF/HTML/asset build or promotion occurred: the verified303-page/9172MathML physical-sieve edition remains current. New source-bound prose proof gates and152 saved synthetic consistency assertion entries are documented in planning/intermediate-kernel-20261106/HANDOFF-v1.md (SHAbb9463e074d822f68cc652ad20d255052536fd7c651febe6fb9a6d9bd2133e74) and outputs/checks/2026-11-06-intermediate-kernel/RESEARCH_ACCEPTED-v1.json (SHA5470159d31d46bf5ce42fc52fcae4dc4013dc662b054f115f27fcc937ff62d63). The project root-normalization gap closes only for the stated actual primitive unitary Hecke conventions, with two independent reviews. The ALL-component fixed-u0 two-pivot finite-prefix estimate includes primitive S zeros, different source-class bilinear Cauchy, all inside-prefix dual modes and actual AU/common W. It proves no infinite-complement/whole signed/global gain, determinant/zero-location or analytic/categoryLean/GRH completion. Numerical controls at65digits are not interval enclosures or original arithmetic-family data. No old checks/builds/browser audits rerun; current canonical paper sources and12 HEAD/indexstates remain unchanged. Linux4.24 approval remains unanswered with no install/substitution/repeatedquestion.
## Research-only coherent-band proof evidence, 6 November 2026
No new book/PDF/HTML/asset build or promotion occurred. Current303pages/9172MathML and whole signed13/12/global9/8 remain unchanged. planning/coherent-band-20261106/HANDOFF-v1.md SHA90a6be03a2518def4243e63c00e587da6837bad04c86dd6fe9f71597cfaa5a82 and outputs/checks/2026-11-06-coherent-band/RESEARCH_ACCEPTED-v1.json SHA34b4f9a51315cafd237a63b13ef4ac5d72b9ceef500d83be573d39dace991ae7 bind independent analytic gates for finite-cutoff Schur transfer, the NORMALIZED same-C/rho1 identity-d-class auxiliary cubic-band bound, full elementary equal-norm off-diagonal sector control and literal-ray recombination with its precise failed cancellation step. No physicalPSD/energy lower bound, all-d/highdual target, determinant/zero-location or Lean/GRH completion follows.
1511 new saved assertion entries55digits are consistency checks on actual-good-ideal auxiliary matrices/Perron integrals; they are NOT interval enclosures or complete original physical-family data. Separate exact O/(6) local controls verify36 Fourier inversions/Parseval108/unit projections, not original q characters or asymptotic estimates. No old controls/build/browser audit rerun, no install/env/Lean substitution. All12 protected HEAD/index states and canonical paper sources remain preserved. Both gates false; Linux4.24 approval still unanswered.
## 6 November 2026 — coherent-band local milestone
FinalbookPDFv2/nativev2:319pages/40chapters/9810MathML. All9172old literalmath subtrees ordered/byte-identical;389oldtables,13oldcanonicaliframes and29oldSVGrefs preserved, totals14iframes/30SVGrefs. Focused76/215pages; notes34/105pages. All5source-boundPDF builds, mathematical/source/artifact reviews, PDFaux-number/native-source closure and readablefigurev4 visualgate pass. Combined-browser-v2:58labchecks/54layouts,1218savedpositions/3654renderedstagecells,25nativegroups,193globalWS,fetch/newpageerrors/pending0. Finalfigure-only native-browser-v3:25groups,183WS,fetch/newpageerrors/pending0,3idlewaits/1actualdeparture+initial; exact750msquiet/30sidle errorpolicy retained. Three inheritedmunder nodes and oldTeXwarnings remain, no new malformednode/overfull/unresolvedref/Lean declaration. Knownwarnclassification exactlymatchespredecessor and exact3before/afterDOM is checked. Every failed version remains. Governing18artifact promotion and sourcebound gates: outputs/checks/2026-11-06-coherent-book/. No new public release, fullformalization, training or GRH conclusion.
## Reviewed coherent-book release edition — 6 November 2026
The reviewed release edition contains a 319-page book with 40 chapters and 9,810 native MathML formulas, focused articles of 76/215 pages and analytic notes of 34/105 pages. All 9,172 previous literal math subtrees, 389 tables, 13 earlier canonical inspectors and 29 SVG references are preserved; totals are 14 inspectors and 30 SVG references. The readable original workflow and complete matrix inspector retain all 1,218 saved auxiliary positions, actual ideal generators/factors, complex numerical values and raw zeros. Saved55-digit values are numerical consistency, not interval certificates, trained models or complete original-family data.
The shared coherent-band sources give primitive root normalization, exact finite Poisson/BV-prefix comparisons, actual-cutoff/profile Schur transfer, the restricted identity-divisor cubic comparison, full primal equal-norm localization and the literal-ray mean. Separate UB01 research is Main-only evidence outside the book and browser data. The original unweighted A6=1 two-pivot signed h1 bound13/12 and global unsigned9/8 remain unchanged. Full signed unequal-norm cancellation, short/principal assembly, independent positive arithmetic determinant/zero-location bridges, analytic/category Lean and GRH remain open.
This section describes verified CONTENT prepared for substantive publication; it does not certify that deployment has completed. A new release requires its exact frozen selection, source/credential/visibility/integrity gates, strict project-prefix browser checks and postpublication hosted hashes/trees. The final PUBLIC_RELEASE_VERIFIED receipt and post-release handoff will govern actual deployed commits. Earlier303-page/local-only statements above are dated historical preparation snapshots. Both completion gates remain false and the one existing heartbeat remains active. Original sources, all old HEAD/index states/worktrees, outputs/caches/environments, Mac workspace/old8765 and private active data remain protected. No Linux4.24 install/substitution/repeated question is authorized.
## VERIFIED LIVE coherent-book release and research-only continuation — 6 November 2026
Main87f89878f470ab98b84055fe72f28766d89202c6 and Pagesf6773684b6ea4d5ee067409c62c88c80d1529080 are VERIFIED LIVE. The authoritative receipt is /home/iska/Documents/LaTeX-projects/GRH-publication-20261106-coherent-book/outputs/publication/20261106-coherent-book/release-preparation-v1/PUBLIC_RELEASE_VERIFIED.json (SHA d85801388a02748e4a1607f288aa3bb974bcd8cb674b4e010ca52cf8e5e959ed). The delivered edition has319pages/40chapters/9810MathML,389tables/14canonical labs/30SVGrefs; all9172old literal math subtrees,389old tables,13old labs and29old refs remain. Five PDFs are319/76/215/34/105pages. All25changed hosted hashes,854selected Git-tree entries/modes/sizes/parents/refs,1247links and854payload roundtrips pass. Strict stage and LIVE220groups each retain ALL191predecessors; the inspector passes58checks/54layouts/1218strings/3654cells. LIVE591timing entries=579protocol entries+12navigation markers,1initial+11departures,13idle checkpoints, zero fetch/new-page errors/timeouts/pending, unchanged750ms-quiet/30s-idle guards. Five live PNGs are byte-identical to their staged renders. Three inherited munder nodes remain; no new malformed node or Lean declaration.
The original preflight directory-index resolution failure and hosted-verifier API trailing-slash404 are preserved. Their narrow reviewed producer fixes changed NO product bytes or browser filter; no republishing occurred. Source HEADdba37ff/index0ef1af remain, origin/main anonymously fetched87f89878, all13HEAD/index states and12release worktrees preserved, including the new225-file indexaca074d8. All edits/builds/checks/durable artifacts ran on Linux; existing Macgh was used only for in-memory publication. Macworkspace/old8765, originals/caches/environments/histories/private active data and all other project sites remain protected.
Research continued after publication OUTSIDE this frozen release, current book, Pages and Main selection. The accepted actual finite primitive sixth-part mean retains every AU/commonW/phase/root/r/k/ell/e/h/rawzero/capacity factor and full q deletion. A self-contained classical fixed-ray disk count t^(1/3), applied cosetwise before complex weighting, sharpens ONLY the same mean's normalized error to B^2(DXA2)^epsilon RXV^(-17/3)(1+sqrt(A1X/D^2)); top-band error is RX^(1/18). Pure critical37/48 still exceeds target3/4; mixed29/48 is only an error certificate. The retained principal coherent mean is unestimated. Governing research-only aggregate outputs/checks/2026-11-06-coupled-sixth/RESEARCH_ACCEPTED-v2.json SHA3be411e2f432bf47a9dd1a8cb8e6eed92d6d1e3f9f666abe8210e9e5ed50ff71 binds independent reviews. No lattice-discrepancy priority, whole signed saving, learning, interval, quantum, Lean or GRH result is claimed.
NEXT: estimate the ACTUAL principal coherent mean jointly with nontrivial sixth parts and r/k/root/column orientations plus every other class/rho/S mode, before conductorwise/positive-d triangles. Retain full means/equalnorm fibres/commonprime/fullunion/actualAU/fixedW/explicitseminorms and all conductor/ray/SF/capacity uniformities. The present lattice-count route is DONE, not the next rerun. Whole two-pivot13/12 and global9/8 are unchanged; short/principal assembly, independent positive arithmetic determinant/zero-location bridges and the separate Dirichlet/Hecke/Dedekind/elliptic/generalArtin scopes remain OPEN. No Artin holomorphy/GRH/desired positivity/zeros is assumed. Analytic/category Lean remains unfinished. Linux4.24 permission is still unanswered: no install/download/substitution/reinventory/repeated question. Both completion gates remain FALSE, P4/P7/P9 in_progress, and ONE existing heartbeat remains ACTIVE. Integrate research-only notes with a future substantive reviewed source/book milestone; no status-only republish or proof schedule.
## Principal-mean research-only checkpoint — 6 November 2026
NO NEW PUBLIC RELEASE or canonical book/UI/PDF/data change. Main87f89878f470ab98b84055fe72f28766d89202c6 / Pagesf6773684b6ea4d5ee067409c62c88c80d1529080 remain the verified319page/9810MathML edition. New independently reviewed notes and failed-step corrections are OUTSIDE that frozen release/book/Main selection. Governing aggregate outputs/checks/2026-11-06-principal-mean/RESEARCH_ACCEPTED-v2.json SHAbbe4a216c1d513ba988479920f222305f3d15c4dbff2317c2f0432a20981eaa1.
The actual retained mean has an exact finite weighted-Mobius a-band compensation, with a/b overlaps, original root/k/column/ray/capacity and all S values retained. In the top band X/V^6=64<7^6, only a=1 occurs and a finitew<=64 root/column moment remains. The cumulative density multiplier kappa_T/[(6z-1)zeta_K^T(6z)] has removable value1 atz1/6, not0. A finite s-band factor is entire; an isolated analytically continued zero and unweighted finite-ray orthogonality do not delete the actual root/conductor-weighted mean.
The exact incomplete1/6 Bessel moment gives the actual coherent difference F_V-E_V with ORIGINAL finiteK/fullT/eta² and every endpoint. The signed-k endpoint error from completeHQ5/3 is at most24 C0^(-1/2)M0 A^(1/6)(b-V), with A=A1*Nh*Nr/(Nl*Ne*F), DISTINCT from A1. Under explicit nonzeroG/largeK/low-argument conditions, endpoint correction has a component lower bound; this is NEVER a lower bound for M, a row form or energy. LaneA's power-free deletion T=S union supp(mn) differs from LaneB's k deletion T=S union supp(mn*v2). Conductor drops restore no deleted prime. Omitting endpoints/analytically completing the finite kernel is the failed cancellation step.
The proposed mean norm corollary is valid, but its claimed10/9 actual-sector advance was RETRACTED before acceptance/publication: accepted UB01 fixed-s comparison already implies the SAME actual-band bound (R/D)V F6(X/V6,D), through mixedX~D^(3/2) atV=.5D^(1/6). Thus10/9 is only a weaker separate-mean/error certificate, not a new range. The density g(d)^2 dampening is compensated by residualg(a)g(b); no signed-d saving. Unequal virtuald support Nd<2D/13 is a constant cutoff only. A bounded-residual virtualprime-d block already has absolute budget R V Z/D (Z=X/V6), puretop exponent-3/16; fixed-family PIT there is not the missing low-d estimate. Its separate phase/PIT diagnostic remains author-only/unaccepted. Originalpair gcdh, virtuald, sixth-free a and originalk are distinct.
Two new stdlib Fraction checks guard the formal budget and its inherited dominance; they are not original-family numerical data, intervals, proof of cancellation, or old control reruns. Failedv1 comparisons/notes and every before byte are preserved. No new exponent, whole signed estimate, determinant/zero-location, learning, quantum, Lean or GRH result. Whole originaltwo-pivot13/12/global9/8 remain unchanged.
NEXT GENUINE task: estimate the actual finitew<=64 root/column/k moment, or the actual F_V-E_V jointly, with all source/rho/S modes, originalAU/onefixedcommonW/explicitseminorms/fullunion/rawzeros/equalnorm/orientationmeans/conductor/ray/SF/capacity data BEFORE positive-d/conductor triangles. Low virtuald has growing residual/conductor data; bounded-residual fixed-conductor PIT cannot substitute. Do not rerun the finished lattice, density-damping, freeEuler/fractional completion or inherited norm-recovery routes. Integrate these research-only notes only with a future substantive reviewed source/book milestone; no status-only republish.
The short/principal assembly and independent positive arithmetic determinant/zero-location bridges remain open separately for Dirichlet/Hecke/Dedekind/elliptic/generalArtin; no Artinholomorphy/GRH/desiredpositivity/zeros assumed. Analytic/categoryLean unfinished; Linux4.24 approval remains unanswered, no install/download/substitution/reinventory/repeated question. Both completion gates FALSE; P7 in_progress; P4/P9 unchanged, ONE existing heartbeat ACTIVE. SourceolderHEAD/index, all13states/12releaseworktrees/frozen225index, originalsources/datasets/outputs/caches/environments/histories/Macworkspace/old8765/privateactive data remain protected. All work LinuxSSH; no auth/publication/externalMCP/Claude configuration change.
## Root-moment arithmetic-tail research checkpoint — 6 November2026
Research-only, OUTSIDE the frozen319-page/9810MathML public book. Main87f89878/Pagesf6773684 and all canonical PDFs/nativeMathML/UI/data remain unchanged. Governing new aggregate outputs/checks/2026-11-06-root-moment/RESEARCH_ACCEPTED-v1.json SHAd95f711210cb1f078076861ab23d34158efda5c12f8ee9aac57551f78cfa6fd9; full derivations/gates and current frontier are in planning/root-moment-20261106/HANDOFF-v1.md.
New independently reviewed arithmetic: for the ORIGINAL unweighted PURE two-pivot top principal mean, the actual Gauss-weighted large-short-divisor difference satisfies |G_D[A_U,A_U]-G_D[A_Y,A_Y]| << B M(W)D^epsilon(Y^-1/2+D^-1/6),1<=Y<=U<=sqrtD. The external pinned physical cubic theorem atn3 is an explicit INPUT. Exact mu*A_U ternary grouping retains residualmu(c); cubic Gauss CRT exposes the ACTUAL a-by-z cubic residue matrix and T_U(z)+T_Y(z), independent of a. All class/rho/primitiveS/root/rawzero/originalcoprime/optionalunequal/capacity masks survive, including varying other-column deletions. Fixed complex commonW costs explicit Mellin seminorm M(W); no supnorm-only varying-profile assertion, coefficient norm shrinkage, generic conductor substitution or new deep sieve is claimed.
The same signed bound keeps the ORIGINAL finiteQ2/K/fullT. At A1~H=D3/4,X=D3/8,Y=min(U,D1/8), actual retained M/D reduces to finite L0[A_Y,A_Y] up to target-sized error: Besselpower-1/16 and large-short-divisor powers3/4,31/48. The residual has a,b<=Y and c>=D/Y²>=D3/4. That SMALL-short-divisor Gauss moment remains UNESTIMATED. The separate P_top-M_top upper certificate37/48 remains above3/4; it is not an actual lower bound. Whole signed13/12/global unsigned9/8, short/principal/full assembly and determinant/zero-location remain OPEN.
Matched-endpoint finite models have a real-axis factorial remainder with no exponential I-majorant; the infinite hierarchy is inherited. The failed smooth-inner/absolute-short-divisor and automatic-Schur routes are recorded precisely, not claimed as target savings. Historical drafts/corrections and3 NEW exact Fraction consistency guards are preserved; guards are not interval proofs or original-family data. No old finite table, numerical/browser/build/auth/publication audit was rerun.
NEXT: estimate the ACTUAL finite-k long residual-Mobius cubic-Gauss small-short-divisor bilinear moment, or jointly its discrete-s discrepancy, BEFORE losing the original coefficient/mean/root/k/w/other-column/fullunion/source/rho/S/ray/SF/conductor/profile/capacity coupling. This is not fixed-conductor smooth PIT or completed-frequency orthogonality. Keep all earlier accepted safeguards; do not rerun the nowfinished model/norm/large-tail proof as the missing full target. Integrate these notes only with a future substantive reviewed book/source milestone.
No analytic/categoryLean or GRH conclusion. Dirichlet/Hecke/Dedekind/elliptic/generalArtin bridge obligations remain separate; no Artinholomorphy/desiredpositivity/zero assumption. Linux4.24approval remains unanswered: no install/download/substitute/reinventory/repeated question. Both completion gates FALSE, P7 in_progress, P4/P9 unchanged, ONE existing heartbeat ACTIVE. SourceHEADdba/index0ef,13HEAD/indexstates/12releaseworktrees/frozen225index, every original/data/output/cache/environment/history/Macworkspace/old8765/privateactive datum remain protected; no canonical/publication/dependency/auth/config/externalMCP change.
## 6 November 2026 — reviewed local Gauss-tail book milestone
The Linux working book is333pages/40chapters/10299nativeMathML; all9810previous literal math subtrees are ordered and byte-identical. All389tables/14canonical iframes/30oldSVG references and85non-HTML book assets remain unchanged. One original short-divisor partition/cubic-matrix diagram raises SVG references to31. Focused articles are81/228pages; analytic notes39/118pages. Sources papers/gauss-tail-{core,details}.tex and matching generated book copies retain85shared labels plus the book-only figure label, with full proofs and exact PDF-aux numbering.
The prior accepted actual pure Gauss-weighted large-short-divisor tail is now fully cited/explained in the book: the two disjoint cutoff boxes give TU+TY before the physical cubic matrix norm. The only deep input is the previously pinned external deFaveri2610.04045v1 Theorem1.1/Example2.2 n3; no new sieve or reproof is claimed. The original finiteK, all primitive roots/S values/source/ray/class/rawzero/full-union/common-complexW data and explicit Mellin seminorm remain.
NEW independently accepted direct reduction retains the ACTUAL discrete sixth-part s and EVERY original rho: P_X[AU,AU] reduces to its original-finiteK small-AY j0 model. It does not introduce the separate principal-density/counting-mean split. At A1~H=D3/4,X=D3/8,Y=min(U,D1/8), errors have powers-1/16,3/4,31/48 (the coefficient tail is exactly0 ifY=U). There is NO k/s or w/s coprimality mask; overlapping frequencies and all w<64/nonunit S modes remain. The old separate37/48 counting-error certificate is neither improved nor set tozero; its effects stay in the exact-discrete small model. That model is still unestimated at target size. The full signed13/12 and global unsigned9/8 are unchanged.
Separate RESEARCH-ONLY work OUTSIDE these book fragments uses the exact J5 finite ideal HB identity and the same physical cubic norm on balanced residual blocks. The linear TypeII column costs M(W)D^epsilon D^(1/3)Y^(1/3); the stated pure principal leading component has critical power11/16. This is a scoped signed-component estimate, not an energy lower bound or a full target theorem. Exhaustive one-long-unrestricted-factor classification explicitly assumesD>=2^64. BOTH-TypeI, including long prime residuals, remains open. See planning/gauss-book-20261106/lane-a/BALANCED_RESIDUAL_HEATH_BROWN_CUBIC-v2.md and independently reviewed research aggregate outputs/checks/2026-11-06-gauss-book/RESEARCH_ACCEPTED-v1.json.
Focused checks: all5source-bound final PDFs pass; current book and detailed editions are buildsv3, unchanged concise editions are source-bound v2 reuse. Final nativev3 adds489MathML with zero new malformed nodes/rawTeX/conversion warnings. Independent mathematical/source/artifact reviews and representative five-PDF/native visuals pass. The SAME focused browser harness passes40checks at desktop1440/mobile320, both palettes and keyboard horizontal scrolling; all86labels/63equations/65references and source/PDF/figure hashes pass. It records105global protocol events+1initial marker, zero actual parent departures, fetch/HTTP/newpageerrors/pending/timeouts0, with750msquiet/30sidle guards. This is focused local QA, not a rerun of the220group public release audit. Three inherited munder warnings and oldamsmath/hyperref/underfull diagnostics remain, with no new overfull/unresolved reference or Lean declaration. Source-matched cached PDF searchability and public-SVG reproduction limits remain disclosed.
Fresh synthetic formal algebra checks cover2304 coefficient equalities/847 target coefficients and2184 dyadic scale boxes, with exact integer powers, distinct equal-norm formal ideals and seed2026110609. A deliberately incorrect J4/same-fifth-root truncation fails at norm544825 (true+1,wrong-23,remainder+24). These support the new HB algebra/grouping; they are not actual character/Gauss/family data, interval arithmetic or Lean. No old scientific tables/traces/control suites were rerun. Every failed source guard, overfull/Pandoc candidate and night palette diagnostic is preserved. The night filter was readable but altered the new source palette; only the new diagram now keeps that palette on a white canvas, with old theme/assets unchanged.
This is LOCAL promotion only. No new public release: Main87f89878f470ab98b84055fe72f28766d89202c6/Pagesf6773684b6ea4d5ee067409c62c88c80d1529080 and the hosted319page/9810MathML edition remain current. Existing Macworkspace/old8765 preview and all13HEAD/indexstates/12releaseworktrees are preserved. No dependency/toolchain installation, authentication, token access/transfer, external MCP activation, Claude/global config or Git/index/publication action occurred.
NEXT implementation: prepare a separately reviewed substantive release from this milestone with all5PDF exports and the exact new source-only research closure; do not publish a status-only update. NEXT arithmetic: bound the actual residual long TypeI Gauss moment or the full exact-discrete small-AY kernel before losing r/k/w/s/root/column coupling. Retain fixedW/explicit seminorms, exactAU, allclass/rho/nonunitS/full-union/commonprime/rawzero/ray/SF/conductor/capacity hypotheses, and return to neighboring pure/mixed blocks and signed full assembly separately. Do not rebrand the finished balanced block, cutoff tail, scalar-count, density, free-Euler or norm routes as the full target.
P4/P7/priorP9 and both completion gates remain unfinished; one existing heartbeat stays ACTIVE. Independent positive arithmetic determinant/zero-location comparisons remain separate for Dirichlet/Hecke/Dedekind/elliptic/generalArtin; no Artinholomorphy/GRH/desiredpositivity/zero assumption. Actual analytic/categoryLean is unfinished. Linux4.24approval remains unanswered: no install/download/substitute/reinventory/repeated question.
## Reviewed Gauss-tail release edition — 6 November 2026
The reviewed release content is333pages/40chapters/10299nativeMathML, with all9810prior literal formulas,389tables,14canonical labs and30priorSVG references preserved; the new original partition/cubic-matrix diagram gives31SVGrefs. Shared papers/gauss-tail-{core,details}.tex and the five approved PDF exports (333/81/228/39/118pages) are reused exactly from the local milestone, not rebuilt as a new experiment. Read the [Gauss-tail chapter](https://amelie-iska.github.io/GRH-research/book/main.html#gtl:core), [book PDF](https://amelie-iska.github.io/GRH-research/book.pdf), [concise article](https://amelie-iska.github.io/GRH-research/concise.pdf), [extended article](https://amelie-iska.github.io/GRH-research/extended.pdf), [concise analytic notes](https://amelie-iska.github.io/GRH-research/validated-oracle-concise.pdf) and [extended analytic notes](https://amelie-iska.github.io/GRH-research/validated-oracle-extended.pdf).
The actual Gauss-weighted large-short-divisor tail and the new all-rho exact-discrete finite-K top-prefix reduction retain original raw zeros, source/ray/S/conductor/capacity data, fixed common complexW and explicit Mellin seminorm. The remaining small-cutoff model, long TypeI terms and full signed estimate remain open; whole signed13/12/global unsigned9/8, determinant/zero-location, comprehensive Lean and GRH are unchanged. The separately reviewed HB balanced residual11/16 component is SOURCE-ONLY outside this book and is not a full model bound.
This describes reviewed release content; stage/live deployment is certified separately by the new publication receipt. Main87f89878/Pagesf6773684 and the public319page/9810MathML edition remain the last VERIFIED LIVE state until that receipt exists. Historical local/pending/older-edition notes are preserved snapshots. The one existing heartbeat remains active, both completion gates are false, all source/private/environment histories stay protected, and unanswered Linux4.24approval licenses no installation or repeated question.
## 6 November 2026 — verified public Gauss-tail edition and continued arithmetic
The final cumulative release is VERIFIED LIVE: Main1f346d1a825aa6eff1479197aef45d293a878902 / Pagesde2614bf439e1c6ab8af09f3242af5a163f16f79. The authoritative receipt is /home/iska/Documents/LaTeX-projects/GRH-publication-20261106-gauss-favicon/outputs/publication/20261106-gauss-favicon/release-preparation-v1/PUBLIC_RELEASE_VERIFIED.json, SHA2569c912f6e419734fa0fa5f315717204a7e44324bb8dd880a207fbfa13f866a965. This supersedes the preceding pending/last-319-page wording; those passages are retained historical snapshots.
The public book is333pages/40chapters/10299MathML. All9810prior literal mathematical subtrees remain ordered and byte-identical; all389tables/14labs/30oldSVGreferences survive, with one new relevant partition/cubic-matrix figure (31totalrefs). The five source-bound PDF exports are book333, concise81, extended228 and analytic notes39/118pages. Read the [Gauss-tail proof](https://amelie-iska.github.io/GRH-research/book/main.html#gtl:core), [book PDF](https://amelie-iska.github.io/GRH-research/book.pdf), [concise article](https://amelie-iska.github.io/GRH-research/concise.pdf), [extended article](https://amelie-iska.github.io/GRH-research/extended.pdf), [concise notes](https://amelie-iska.github.io/GRH-research/validated-oracle-concise.pdf) and [extended notes](https://amelie-iska.github.io/GRH-research/validated-oracle-extended.pdf). The original Mac loopback preview remains untouched.
The first substantive publication was Mainb6e1571219d92432454ffd1afac2c812a9f28b9e / Pages6b3565d4e76c46a07cad6b32ef4b21274e800493:249selected Main paths/884API payload checks and16changed hosted files. Strict LIVE then exposed13 account-root/favicon404 responses. The failure was preserved, not filtered. A real404-origin control reproduced it; a single651byte self-contained SVG-icon insertion in the shared public runtime removed the implicit request. A separate additive repair selected35Main paths and changed ONLY public-base.js on Pages; ALL634other final public files, including every PDF/source/math/data/figure, are byte-identical to the first333-page bundle. One staged repair attempt correctly stopped at an inherited hardcoded runtime hash; its replacement binds the SAME exact-SHA assertion to the newly frozen bundle. The staged origin now returns404 rather than204 for an undeclared favicon. Neither error filters nor HTTP/WS/idle assertions were relaxed.
Final635-file bundle inventory29d5d5eaafbb30ed2f3220887b44bc5da2bb598dad9c1eea8a154718f6dc451b,670repair payload roundtrips/refusal controls,670-file credential-pattern scan/zero matches, exact source pins, allunselected public leaves/modes/actual one-parent commits/refs and all17cumulative hosted hashes pass. The honest Pages chain is f6773684→6b3565d4→de2614bf, not an invented single17-file commit. The failed240second deployment check and all33pending rounds are preserved; actual runtime deployed12:27UTC, and final hosted17 checks passed at13:33UTC.
Strengthened STAGE and actual LIVE each pass220inherited groups+40Gauss-tail groups+58laboratory checks/54layouts, preserving1218saved string positions/3654rendered cells. Governing LIVE run outputs/checks/2026-11-06-gauss-release/qa/live-browser-20261106T133342216621Z: fullreceiptcf9ca5892b183830d9e3a4524e52be4692834f9b478e2ca0fddd102be0f96a0a, focused7688a0cb8ff132b582d00ec53fc99362ef1ff25158010938581d4a18fe0d030f. Its645global events include632protocol events+13navigation markers=1initial+12actualdepartures;15successful idle waits and minimum758ms quiet exceed unchanged750ms/30second guards. HTTP/page/fetch errors, timeouts, departure/final pending requests and favicon requests are zero. Owned tabs/contexts/sessions and stage listeners closed; pre-existing contexts and9335/8765 services preserved. Final independent deployed audit c4bd788a and scientific-scope addendum f4793cb6 pass.
Arithmetic work CONTINUED after publication, outside this frozen book and both release indexes. Governing outputs/checks/2026-11-06-gauss-release/research-aggregate-v1/RESEARCH_ACCEPTED-v1.json SHA77671eeeeb33fc060db945c7e5062e2a3423e8fdcf7e458422afff60691dae2d binds14current notes/reviews/controls:
- NEW actual pure finite-k cubic-band estimate retains the coupled two columns before a positive common-divisor triangle: A1*V times[J^-2+(DJ)^-1+D^-1/3 J^-4/3], with exact Bessel remainder Lambda/J and original finite cutoff. At critical J0=D^(1/32), the large-k sector is target-sized; this is a scoped external cubic-sieve application, not a new sieve or full signed saving.
- The exact TypeI deletion/frequency transfer shifts frequency to d*t and scale to T/Nd, preserving their product and all raw zeros. Its sharp-interval H(alpha,beta) criterion with alpha+beta<=7/8 is explicitly hypothetical, sufficient only and unavailable; additive common-lift completion failed to provide it.
- Governing localization-v2 c0befeb4 (v1prose corrected, no quantitative rerun) leaves actual Nk<=D^(1/32) BOTH-TypeI. The exact identity is mu(m)A_Y(m)=sum_(abc=m,Na,Nb<=Y)mu(c), SF/coprime with unit cases and NO extra-minusmu. T>=cD^(1/2), Nt<=CD^(17/32), T*Nt=D*Nk<=D^(33/32), full deletion NQ<=CD^(51/32), and opened frequency<2D^(33/32). Frequency gcd=k holds only on original support before auxiliary openings. These are a reduction/windows, not a bound for the remaining kernel.
- The new paired-long attempt99fe6c5a matches SAME pinned deFaveri2610.04045v1 Proposition8.1 exactly (squarefree q=k, unrestricted paired long outputs). Opposite Gauss CRT crosses cancel the proposed free long-long cubic interaction; physical cube-free height is LM*min(L,M), not radical LM. The legitimate paired certificate retains LM/J^(3/2), weaker than accepted J^-2 offunit. The actual unit/very-short-k Gauss-vector mean remains open. g0 mayzero and is never divided out.
Fresh825Fraction capacity/1024abstract support checks and13independent rational expressions are FORMAL only, not actual character/energy data or Lean. No finished scientific tables/training/traces were rerun. Full third-party primary caches and private pin locators remain outside public closure.
NEXT genuine arithmetic: estimate the actual mean-retaining unit/very-short-k BOTH-TypeI Gauss vector jointly with every rho/source/e/w/s/root/commonprime/full-union/rawzero/conductor/ray/SF/capacity constraint and ONE fixed common complexW/explicit seminorm, or its original signed discrete discrepancy BEFORE positive divisor/conductor/column triangles. Generic paired norms, the finished large-k/large-short/TypeII/scalar/density/free-Euler routes and finite controls are finished, not the next experiment. Whole signed13/12/global unsigned9/8, short/principal target assembly, independent positive determinant/zero-location bridges and separate Dirichlet/Hecke/Dedekind/elliptic/generalArtin scopes remain OPEN; no assumed Artin holomorphy, desired positivity/zeros or GRH.
All15HEAD/index states/14release worktrees are protected, including sourceHEADdba37ffaa78c82ade1385f7ee56c907ab0c67d3a/index0ef1af515b143ef68fed1d9a42aef7c00d5d1d7c3ef6832d2926d48e0925214c, first333frozen249indexce6a9b71 and repairfrozen35index42e7a893. Existing Macgh was used ONLY for authorized publishing in memory; all source/science/build/browser/checks ran SSH-Linux. No credential read/copy/transfer, login, dependency/toolchain installation, external MCP activation or Claude/global configuration change occurred. Preserve every original/environment/cache/history/private model and old preview. Actual analytic/categoryLean remains unfinished; Linux4.24approval is still unanswered, so no install/download/substitution/reinventory/repeated question. P4/P7/priorP9 and both completion gates remain unfinished; the ONE existing heartbeat stays ACTIVE. These append-only local release/research receipts will enter a future substantive source/book milestone; do not create a status-only republish.
## Research-only unit/short-k Gauss checkpoint — 6 November 2026
No new release or canonical book/PDF/laboratory change. Main1f346d1a825aa6eff1479197aef45d293a878902 / Pagesde2614bf439e1c6ab8af09f3242af5a163f16f79 and the verified333-page/10299MathML edition remain current. Latest release authority is planning/gauss-release-20261106/POST_RELEASE_HANDOFF-v1.md and its bound PUBLIC_RELEASE_VERIFIED receipt. This note is outside that frozen edition; integrate only with a future substantive reviewed milestone.
New independently reviewed research aggregate: outputs/checks/2026-11-06-unit-gauss/RESEARCH_ACCEPTED-v1.json (SHAf878074ae8df66e9032875b93a281e4143699015cfa43a76d8968c759a1c2870); detailed handoff planning/unit-gauss-20261106/HANDOFF-v1.md.
- ACTUAL pure short-k BOTH-TypeI auxiliary common-LONG-divisor tail: M(W)^2 A1 V (D X A1)^epsilon I(cLambda)[D^(-1/2)+R0^(-2)+D^(-1/6)/R0], normalized once. AtR0=D^(1/32), critical powers5/16,3/4,59/96 fit the target. This opens every common divisor d, not just the greatest: residual a,b may overlap; no residual coprimality is added. d is not originalh, finitek or oldsmoothf. Exact finiteK/Bessel/commonW/raw/full-union/source/rho/Gamma/class/ray/SF/conductor/capacity data remain. Required finite-S joint-selector companion51f1c728 expands HxH exactly with fixed mass<=|H|sup|f|; its d phase is chi(d)psi(d), not a generally assumed thetaS square. Separate genuine rho(d)^2 and Gamma factors remain.
- The attempted full estimate FAILED at unit/small d, whose inherited normalized13/16 certificate still exceeds3/4 by1/16. No actual coherent mean is discarded. Optional equal-norm auxiliary subtraction is counted globally, not removed inside separated norms. The independent interval-Schur lemma costs2(1+ceil(log2 n)) for arbitrary row intervals; full equal-norm fibres remain.
- A separate source-BARE fixed nonzero-angular sharp-interval corollary gives11/12 unnormalized,5/12 normalized, from the pinned external September2026 repository preprint. It is NOT the actual finite-twisted family: exact Gactual(c)=chi_(3,c)(delta)Gsource(c), the full finite kappa and Q deletion remain unmatched. Deletion keeps productD*Nk but moves level; generous smooth margin .01-.49*kappa is negative at1/32. Cuspidal and prime-only noncubic results cannot fill these interfaces.
Fresh formal checks:625 common-Gauss/gauge Laurent identities,256 Mobius reconstructions,65 exact prefix/exponent controls; interval controls2144prefix/5524interval/14235row-set/256endpoint;25 source/interface checks including9Fractions, plus root's focused Perron exponent receipt. These are synthetic algebra/source/interface evidence, not actual character/energy data, old-table reruns or Lean. All source versions are pinned; full third-party primary bodies and private fetch/cache pin records stay excluded from public closure. Wording/source-status/divisor/unit/S-selector corrections and failed preliminary probes remain preserved.
NEXT genuine task: estimate the ACTUAL joint d<=D^(1/32), unit/very-short-k BOTH-TypeI Gauss mean/discrete discrepancy, retaining every finite/angular/source/rho/e/w/s/primitiveS phase, raw/commonprime/fullunion mask, fixed commonW/explicit seminorm, norm fibre and capacity/conductor/ray/SF uniformity BEFORE positive divisor/conductor/column triangles. The finished long-d/large-k/large-short/TypeII/generic-norm/scalar/density routes and formal controls are not new experiments. Whole signed13/12/global unsigned9/8, short/principal target assembly, independent positive determinant/zero-location and separate Dirichlet/Hecke/Dedekind/elliptic/generalArtin obligations remain OPEN; no assumed GRH/Artin holomorphy/desired positivity or zeros.
P7claim32a1ca5c-761b-4f5a-ab03-bf5fcd9928c5 remains in_progress; P4/P9 and both completion gates remain unfinished. All15HEAD/index states/14release worktrees, source0ef/index, frozen249ce6/frozen35patch42e, every original/output/cache/environment/history/private model and old Mac8765 preview are preserved. All scientific edits/checks ran via SSH-Linux; no publication, new dependency/infrastructure/GPU/auth/token/config/Claude/externalMCP/Lean action. Linux4.24approval remains unanswered: no install/download/substitution/reinventory/repeated question. Keep the ONE existing heartbeat ACTIVE; no status-only commit/republish or guaranteed proof schedule.
Administrative P7 checkpoint: daa7bb6d-231a-48c2-a286-c9c0eb14c47c; evidence and derivations above govern scientific acceptance.
## Research-only whole-prefix / fixed-twist checkpoint — 6 November 2026
No new public release or canonical scientific book/PDF/lab build. Main1f346d1a825aa6eff1479197aef45d293a878902 / Pagesde2614bf439e1c6ab8af09f3242af5a163f16f79 and the reviewed333-page/10299MathML edition remain current. Read planning/twisted-gauss-20261106/HANDOFF-v1.md and outputs/checks/2026-11-06-twisted-gauss/RESEARCH_ACCEPTED-v1.json SHA590866c2da718fb92e96b86e7e9dfbd4135b4dee1be5e6fb1161fe5f88c4f322 before older next-task text. These notes are outside the frozen book; integrate only with a future substantive reviewed source/book milestone.
NEW actual unweighted PURE BOTH-TypeI localization: group the ORIGINAL shorta,b together with ALL short HB factors into prefixp BEFORE a physicalp-v cubic norm. The signed own-prefix coefficient is divisor-bounded; no separateY count is paid. The large-whole-prefix contribution is <=B M(W)(DXA1)^epsilon A1 V[D^(-1/6)+Z^(-1/2)], retaining all actual root/source/rho/Gamma/primitiveS/raw/fullunion/commonprime/ray/SF/capacity constraints and one fixed complexW. AtZ=D^(1/8), powers31/48 and3/4 fit target, leaving BOTH Np,Nr<=D^(1/8). This is new relative to old balanced residual-c TypeII and the old a>Y tail, not a rerun or relabeling.
On this newly localized support, Nk,Nd<=D^(1/32), t=p*d*k and exact full deletion h|rad_good(r*s*w), with (h,t)=1 forced by raw cubic zeros. Thus Nh<<D^(3/16), N(th)<<D^(3/8), residual length>>D^(21/32), and (T/Nh)N(th)=D Nk. There is a uniform9/64 power margin below sqrt(D Nk). The old worst-prefix moved-level obstruction is REMOVED in THIS smaller sector. Fix opposite residualb BEFORE p-averaging: (p,b)=1 stays in a LEVEL coefficient, not a false(a,b)=1 mask. Required finiteS product expansion and all legal k/w/s overlaps remain. Conditional source-averaged H_tw would give35/48, but the actual full finite-twist H_tw is still unavailable.
NEW admitted SOURCE theorem: for primitive fixed arbitrary finite-order residuekappa moduloF COPRIME3, fixed nonzero angularell, and movingSF r coprime3F, the exact cube completion is L^(Fr)(3s-1/2,kappa^3Theta3ell), with all raw deleted primes kept. Same-CRT finite matrices give refined dual absolute normR^(1/2+epsilon); generic transform counting gives coarseR^(3/2). Completed-numerator convexity is proved before division byL ONLY where its argument has realpart>1. Sharp Perron/partial summation gives V1(W)(RT)^epsilon min(sqrtT,T^(1/4)R^(1/8)), or the coarse exponent3/8. V1=sup|W|+integral|Wprime|; fixedF/kappa/ell dependence is explicit, not all-conductor uniformity. Their localized ACTUAL powers39/64 or41/64 are CONDITIONAL on full actual source/ramified matching, not new unconditional whole/model bounds.
Exact normalization now reads vartheta(c)=Theta1(c)barChi3,c(2), and the actual inverse-different correction combines into sourcefrequency12t. This is NOT a goodSF level. A general3-ramified finite character has an exact coefficient projector using shifts h*lambda^(2-f), trace denominatorlambda^(f+1), l1mass<=3^(f/2), selecting v_lambda(mu)=-3. Its Weyl/cusp closure and quantitative dual absolute norm remain OPEN; fixed dimensionality or a zero-producing naivepsi(lambda4mu) insertion is not a proof. Originalnu is fixed FINITE ORDER, not assumed orderdividing6, so higher3-primary conductor is allowed. Fixedq0 overlap reductions do not dividekappa(t); conductor-drop h terms retain literal Nh^-1/2 orNh^-s and rescaled profiles, under the required normalization companion.
Fresh checks are bounded evidence:105 meaningful exact finite-matrix states in a144-dimensional cyclotomic integer basis;64new affine geometry vertices;319height/2871horizontal/259fallback Fractioncases;root14scope/budget checks+8vertices. These are synthetic algebra/parameter checks, not actual-family energy data, training, Lean or replays of accepted old fixtures. All major derivations have independent gates. Only pinned unconditional DR2109.07463v3 Section5 statements are used for the FE route; no GRH-conditional prime theorem. Full third-party bodies/private cache and failed Patterson-fetch receipts stay excluded from public closure; source metadata stays pinned.
NEXT genuine research: prove the exact ramified3/cusp Weyl matrix for the new projected theta vector, all actual unit/source/primitivee/w states and raw conductor-drop zeros, with an adequate moving-level dual absolute bound. Even coarseR^(3/2+epsilon) already suffices through conditional41/64; refinedR^1/2 is not necessary for this target. Alternatively estimate the original jointly masked mean before positive triangles. Do not repeat finished prefix/large-d/large-k/TypeII/scalar/norm/control routes as new work. Whole signed13/12/global unsigned9/8, short/principal/full-frequency assembly, independent positive determinant/zero-location and separate Dirichlet/Hecke/Dedekind/elliptic/generalArtin obligations remain OPEN; no assumed GRH/Artinholomorphy/desiredpositivity or zeros.
P7claim9d375667-ff87-48b0-8f5c-1e25d969e9fc / checkpoint1210e91f-8dc8-4cbf-b087-dd27f230d640 remain in_progress; P4/P9 and both completion gates remain unfinished. Preserve all15HEAD/indexstates/14releaseworktrees/source0ef/frozen249ce6/35patch42e/every original/output/cache/environment/history/private model/Macworkspace/old8765. All scientific/project writes and checks SSH-Linux. No canonical build, publication/commit/push, dependency/infrastructure/GPU/auth/token/config/Claude/externalMCP/Lean action. Linux4.24approval unanswered: no install/download/substitution/reinventory/repeated question. Keep the ONE existing heartbeat ACTIVE; no status-only republish or proof schedule guarantee.
## Accepted ramified-Gauss and original full-A_U prefix research —6 November2026
Research-only; public333page/10299MathML edition unchanged. Authoritative handoff planning/ramified-gauss-20261106/HANDOFF-v1.md SHA14afea77ece78dae8cd5e84fd66db3e077c5dbc15ad24700a781ec31ff99246a.
Aggregate outputs/checks/2026-11-06-ramified-gauss/RESEARCH_ACCEPTED-v1.json SHA f3f199171068235f9161c769eda1209d443d46dc3416e95a6b316f1cfc14c503 governs.
The previously missing arbitrary fixed finite-character ramified3 interface is constructed: exact rational-translate SL2/Kubota reduction into KNOWN27 theta cusps; explicit transformed constants; pure nonzero-angular Mellin FE with NC^(1-2s) including Jacobian; actual lambda^3mu primary selector; raw cube completion with no restored Euler prime; known-cusp coefficient norms and conservative R^(3/2+epsilon) moving dual cost. The inherited completed-convexity/sharp-interval argument now gives actual fixed-family H(3/4,3/8), including original higher3-primary nu, full raw zeros, both angular signs and unit/subunit fallback. No refined sqrtR claim. Ell0 residues at1/6 and5/6 are retained, not declared absent.
The actual family is a fixed residue family from original finite-order nu, source ray quotient, Gamma, primitiveS and fixed Kummer numerators. Jacobi rho=mu*vartheta is NOT finite: its ONE varying vartheta supplies angular+/-1; genuine common/prefix phases remain outside. No densityg or arbitrary growing-conductor coefficient function is inserted.
NEW root theorem: planning/ramified-gauss-20261106/root/FULL_AU_PURE_PREFIX-v3.md SHA668f71eb4036d93e7907c7bba36ab59c44001fefad9463c2910490a7a4b04d98, WITH REQUIRED coefficient/annulus precision1f52dac7, prefix-mask precisiondfb2b1e8 and formal-control-scope precisionbd816dfd. Independently accepted B gatef2258609 and C gate5e01ac18.
For the ORIGINAL unweighted PURE finite primitive-dual top piece P_X[A_U,A_U] of DIRECT-v2(2.1), D>=2^64,1<=U<=sqrtD,A1~D^(3/4),X=D^(3/8),V=.5D^(1/16),
|P_X| <<_fixed,eps (DXA1)^eps A1V
{B[M(W)+V1(W)]D^(-1/9)+B^2[A1X/D^2+D^(-1)]}.
The normalized leading power is101/144, below3/4 by7/144. The raw numerator D*P_X has one additionalD. Whole signed13/12/global unsigned9/8 remain unchanged.
Proof: open B_U=mu*A_U (NOT original C_U) only on actual Nm>U annulus, retain exact j1/units and all short-factor endpoints. Classify ALL at-most11 original/HB factors together from fixed scales before any a,b triangle. Balanced boxes cost columnD^(1/3). A nonbalanced long factor is genuinely unrestricted coefficient1 because original a,b<=sqrtD and truncated mu factors cannot exceed the2/3 scale at D>=2^64. Whole prefix p<<D^(1/3). Its >Z part costs D^(1/3)+sqrt(D/Z). At p<=Z freeze ORIGINAL opposite n; keep prefix(p,n)/(p,k)/(p,s,w) zeros in the actual prefix coefficient BEFORE opening the long-variable FULLQ=rad_good(pnksw). No auxiliary common d is opened. Including original p^-1/2, h-deletion costs D^1/4 Nk^3/8 Np^-3/8 Nh^-3/8. Sum gives D^1/4 Nk^3/8 Z^5/8. Z=D^2/9 balances columnD^7/18; original finite-k majorant Nk^-13/8+epsilon converges without changing the identity. DIRECT Bessel error is applied directly to A_U,A_U: no A_U-to-A_Y error. All rho/e/w/source/root/primitive phases, means, equal-norm fibres, raw zeros, common W, capacity and original finiteK survive. Legal k/s/w overlaps remain.
### Checks and limits
New exact root geometry:4680 rational cases,648 matrices/24 rational subgroup/27 cosets;3942 wrong-diagonal negative controls. Root finite-geometryv1 serialization failure is preserved; v2 accepted. C:40SL2/120H3/320phase/88Mellin/544primary formal cases. B:432 affine/204primary/12rational envelopes. NEW root793091 formal scale compositions,761471 typed subset,1842TypeI/759629TypeII/28j1, exact Fraction budgets. These are NOT original-family energy data, training, asymptotics or Lean. Typed exponent grid is not exhaustive physical dyad coverage: exact proof inequalities handle constants. No finished old tables/tails/scalar/norm/controls were rerun. V1/v2 proof wording, coefficient conventions, all failed harnesses and required addenda remain preserved.
Only pinned unconditional DR2109.07463v3 Section5/AppendixA and previously admitted physical cubic/HB/CRT/Bessel inputs are used; not the paper's GRH-conditional prime results. Full primary bodies/private cache pins remain excluded from future public closure; citation metadata/versions/hashes are public. Root source closurebc66018a binds44current pins. Scientific claims are prose/external-input theorems, not kernel-checked Lean.
SSH refused during the preceding run; no Mac project fallback occurred. It recovered18:46:51UTC. All15HEAD/index states/14worktrees,18promoted scientific artifact bytes and CPUlock passed preservation5cb46cca. SourceHEADdba37ff/index0ef1af, frozen249ce6/35patch42e and every original/output/cache/environment/history/private model/Macworkspace/old8765 are preserved. No auth/token/dependency/infrastructure/GPU/config/Claude/externalMCP/Lean/build/publish action occurred. Linux4.24approval is unanswered: no install/download/substitution/reinventory/repeated question.
### Checkpoint and next concrete work
P7 claim373b0e7f-3795-4a6a-bb23-fa74388e9004 / helper checkpoint215d3552-f50f-4a24-84c3-1c9c717b5ebc remain in_progress. P4/P9 and both completion gates remain false. The helper is administrative, not a proof checker.
Next substantive implementation milestone: coherently integrate the accepted ramified/full-A_U results into the critical book and focused/note sources, with an original canonical construction visual, staging and independent PDF/MathML/UI preservation checks before any release. Reconcile the existing P4 claim first. Do not perform a status-only republish.
Next genuine arithmetic: map and estimate the ORIGINAL frequency complement outside this declared representation-fixed pure top prefix, or the actual mixed/neighboring two-pivot extension, before losing the joint phases/means/masks. The new result does not bound all pure sectors, the infinite dual-frequency complement, full h1, short/principal assembly, arbitrary weighted rows, independent positive determinant/zero-location bridges, or separate Dirichlet/Hecke/Dedekind/elliptic/generalArtin scopes. No assumed Artin holomorphy/GRH/desired positivity/zero locations. Analytic/category Lean and GRH remain open. Do not replay the now-finished ramified FE or new formal controls as new research.
Keep ONE existing heartbeat ACTIVE; no cleanup/reset/no-op/status commits or proof schedule guarantee. Native automation metadata may be updated on Mac only; all project/scientific work remains SSH-Linux.
## 6 November 2026 — reviewed ramified-cusp/full-coefficient local edition
The Linux working edition is348pages/40chapters/10858nativeMathML. Every one of10299 earlier literal math subtrees remains ordered and byte-identical;389tables,14canonical iframe tools,31earlierSVG references and all91 earlier non-HTML book assets (92 original book files including HTML) are preserved. The new original rational-cusp/whole-factor construction figure raises references to32. Focused concise/extended PDFs are85/243pages; analytic notes43/131pages. All five are freshly built from pinned current sources; no dependency or Lean substitution was used.
Shared papers/ramified-gauss-{core,details}.tex and matching book sections contain the explicit known27 cusp construction, literal Kubota phase, actual raw finite-character family, angular Mellin factor NC^(1-2s), coefficient convention/normalization and coarse R^(3/2) dual norm. The ramified fixed-finite-kappa sharp H(3/4,3/8) estimate follows from the identified unconditional theta/cubic inputs. Zeroth-angular residues at1/6 and5/6 are retained. Jacobi rho is NOT finite order; exactly one angular vartheta remains. DR2109.07463v3 Section5/AppendixA and the pinned deFaveri2610.04045v1 cubic input are external theorems, not new sieve or Lean reproofs. Full primary bodies/private publisher caches remain excluded from public closure; versioned citation metadata and hashes remain pinned.
The ORIGINAL unweighted PURE finite primitive-dual TOP prefix P_X[AU,AU] now has a complete original-coefficient proof. Under D>=2^64,1<=U<=sqrtD,A1~H=D^(3/4),X=D^(3/8),V=X^(1/6)/2, the whole at-most-eleven-factor partition and Z=D^(2/9) give normalized leading upper power101/144, margin7/144 below3/4. The raw numerator D*P_X has one extraD. This is the exact representation-fixed finite prefix, not the whole signed h1 form. Opened B_U=mu*A_U is not the project's original C_U; the ternary identity is used only on Nm>U, with literal unit factors retained. Original n is frozen before the long-column bound; prefix(p,n)/k/s/w rawzeros remain outside the full Q deletion. Original finiteK, source/ray/root/Gamma/primitive-e values, all means/equalnorm fibres, allowed overlaps, capacity and fixed-common-W seminorms remain. No AU-to-AY replacement or auxiliary common-divisor opening occurs.
Independent mathematical/figure reviews and source-bound five-PDF/native checks pass. The original figure was inspected as actual PDF and actual SVG and at its printed book size; representative pages from all five PDFs are legible without clipping. The MathML renderer has no new malformed nodes, raw TeX or conversion warnings, and all75newlabels use PDF-aux numbering. Focused desktop/mobile/night/keyboard/source/PDF browser QA and independent18-artifact promotion review govern the local promotion record, not a replay of the completed public release audit. Three inherited munder warnings and old amsmath/hyperref/underfull diagnostics remain disclosed. No new overfull, unresolved reference or Lean declaration is accepted. Source-matched PDF searchability and the public-SVG reproduction limit remain disclosed. Two bounded administrative guard failures (optional-short-title chapter count; book-vs-paper source-map key) were corrected without rerunning builds or relaxing scientific checks; failed receipts and all candidates are preserved.
A separate NEW research-only result remains OUTSIDE this frozen book package: planning/ramified-book-20261106/lane-c/research/BOUNDED_GOOD_PART_COMPLEMENT-v2.md WITH its required fixed-family-match addendum50a7feb1, aggregate23285bc4 and independent B691a56/root32cae907 gates. It estimates the actual lower-sixth complementary cone and whole finite cone with good sixth-free Nr<64 and ARBITRARY allowed S-supported part, at the same101/144 upper budget. Literal finite-ray order L=lcm(6,exponent G_ray), primitive power raw masks and a NEW fixed numerator family are proved; order alone is never used to bound varying conductor. Equal-norm ideals and legal r/s overlaps remain distinct. Fresh54432 phase/raw-mask and4079 formal cone controls are synthetic, not actual energy/Lean data; completed controls were not rerun for the v2 precision. Growing good sixth-free Nr>=64, the full dual ball and infinite/high-frequency complement remain OPEN.
This is a LOCAL substantive edition, not a new public release. Public Main1f346d1a825aa6eff1479197aef45d293a878902 / Pagesde2614bf439e1c6ab8af09f3242af5a163f16f79 still serve the verified333page/10299MathML edition. All science, staging, rendering, algebra and browser checks ran on iska-linux-vscode via SSH. SourceHEADdba37ff/index0ef1af and all15HEAD/indexstates/14releaseworktrees, original manuscripts, datasets/checkpoints/environments/cache/history/private active models and Macworkspace/old8765preview are protected. No auth/token access/transfer, dependency/toolchain installation, infrastructure/GPU, external MCP or Claude/global configuration change occurred. Linux4.24approval remains unanswered: no install/download/substitute/reinventory/repeated question.
Next implementation: prepare a separately reviewed substantive public release from this local milestone with the five PDF exports, source-only public research closure and strict inherited live-browser/source-pin/integrity/visibility/credential gates; no status-only commit or republish. Next genuine arithmetic: estimate actual growing-good-r finite frequencies or neighboring mixed blocks BEFORE losing joint phases/means/raw masks, then assemble the remaining full frequency/short/principal terms. Whole signed13/12/global unsigned9/8 are unchanged. Independent positive determinant/zero-location and separate Dirichlet/Hecke/Dedekind/elliptic/generalArtin obligations remain open; no Artin holomorphy, GRH, desired positivity or zero locations are assumed. Actual analytic/category Lean remains unfinished.
The governing local record is outputs/checks/2026-11-06-ramified-book/root/MILESTONE_ACCEPTED-v1.json and planning/ramified-book-20261106/HANDOFF-v1.md. P4 claim9902aec2-80f1-48e5-9e0a-1e79a2028dff remains in_progress together with P7/priorP9. Both implementation/research completion gates remain FALSE. Keep this ONE existing heartbeat ACTIVE, with no cleanup/reset/no-op/status commit or guaranteed proof schedule.
## Verified public ramified release and accepted growing/mixed cones — 6 November 2026
This entry supersedes historical pending-publication and next-task descriptions, preserving their evidence. Main aeb7705c8bc5b7662a44916ea5039fb2dae7266d and Pages 163b9eaf49298864a12e8f09cb2ec171a8a605df are VERIFIED LIVE. The authoritative receipt is /home/iska/Documents/LaTeX-projects/GRH-publication-20261106-ramified/outputs/publication/20261106-ramified/release-preparation-v1/PUBLIC_RELEASE_VERIFIED.json, SHA256 c3bb690ee951a58759217960b33e6c4900b36d54fade69b47756bbfb3baad499. Public proof: https://amelie-iska.github.io/GRH-research/book/main.html#rg:core ; PDF: https://amelie-iska.github.io/GRH-research/book.pdf .
The released edition has 348 pages, 40 chapters and 10,858 MathML expressions; focused PDFs have 85/243 pages and notes 43/131. All 10,299 previous literal math subtrees, 389 tables, 14 labs and 31 old SVG references are preserved; one original rational-cusp/eleven-factor figure raises the SVG count to32. The reviewed pure finite TOP-prefix theorem has normalized power101/144 under its stated D>=2^64, fixed finite characters and common-W seminorm hypotheses. It uses original A_U, the whole eleven-factor partition, all finite k/fullQ and actual prefix/opposite-column masks. It does not replace A_U by A_Y or introduce auxiliary common-d.
Publication checked 331 Main files/566 closure pins, 641 hosted files and all972 selected Git-tree entries/modes/sizes/parents/refs;625 old hosted files remain byte-identical. All16 changed hosted hashes passed deployment round3, with two pending rounds preserved. Stage and LIVE each passed220 inherited groups+40Gauss+44ramified and Lab58/54. LIVE has693 global events=679protocol+14navigation markers, one initial and13 actual departures,17idle waits, zero fetch/HTTP/new-page errors/timeouts/pending and minimum quiet774ms at idle-wait exits (required750ms). Frozen QA5ccf4f28, ROOT_READY784922eb, ROOT_REVIEW2ee93de4, preflightb227257a, allowlist9103c529, APIb7040130 and independent releasegate1da7d2a6 bind the exact candidate. All972 payload SHA256/GitSHA1/UTF8/base64 roundtrips and refusal controls pass; the972-file credential scan has zero matches. Failed map/stage/admin candidates and111 hash-reviewed generated/CSV/history/harness whitespace diagnostics remain preserved, without normalization or weakened error filters. Owned contexts/sessions are closed; LIVE used no stage server. Actual public desktop figure/mobile theorem were visually inspected. Three inherited munder nodes, old TeX warnings, cached-PDF searchability and the documented public SVG reproduction/standalone-navigation scope limits remain.
NEW accepted research OUTSIDE this frozen331-file release and current348-page book: growing-modulus v2c922a225 WITH required actual-row87ce4a54 and aggregateb50439b9 proves explicit auxiliary raw period F0*rad(r), not a physical-radical norm replacement. It retains actual Gauss/source normalization, one angular vartheta, arbitrary S powers, finite ray order, raw conductor-drop zeros, all equal-norm fibres, fullQ/prefix masks and common-W costs. Legal Z_r=D^(2/9)*Nrad(r)^(-1/3) gives column D^(7/18)*Nrad(r)^(1/6). The original pure cone Ny<=X,Nr<=Y has leading power101/144+delta at X=D^(3/8),Y=D^delta; it is target-sized through delta7/144. The full finite-ball positive budget155/144 is not the complete kernel estimate. Fresh saved local quotient/valuation/Perron/Fraction controls are formal, not ideal energies or Lean, and are finished rather than a next experiment.
The accepted mixed extension is ACTUAL_MIXED_PRIMITIVE_CONE-v2.md9b0b8b6c WITH REQUIRED6fed549f and7d62d6c9; aggregate6456d737, handoffcbdf8ea4 and independent rootb31c7692/B51b4eb05 govern. It applies ONLY to original nonprincipal unweighted two-pivotA6=1 rows with the other four good factors units, fixed u0/S/nu/ray and one common complex W. It keeps ORIGINAL(k,v2)=1, ell|v2, pre-S cutoff sqrt(b/Nell), exact finite A_AB(v2^2)*conj(rho(v2))^2 and primitive S/e/root values. Only the good-local sextic factor uses gamma=alpha_ell*alpha_v2^2*alpha_r^5; total eta may have higher ray order. Quadratic ell twists and raw zeros survive conductor drop. J_np on the unit/unit branch is fixed before Poisson; an arithmetically empty original fibre need not have zero finite projection. The DIRECT original coefficient has no imported density-g. Unequal-norm subtraction follows the unprojected column bound and never becomes an artificial kappa/Q deletion.
For1<=Y<=X and2A2Y<=D^(2/3), the normalized mixed cone is bounded by parameter^epsilon*A1*A2*X^(1/6)*{B[M(W)+V1(W)]D^(-1/9)A2^(1/6)Y+B^2[A1X/D^2+D^(-1)]Y^(5/6)}. At H=D^(3/4),X=D^(3/8),Y=D^(7/144), actual capacity Nu0*A1*A2^2<=H gives main<=H*A2^(-5/6), Bessel<=D^(-19/864)A2^(-3) and equal-norm<=D^(-127/864)A2^(-1), with explicit profile/fixed-data/epsilon costs. Fixed C_H capacity constants have an explicit large-D threshold and bounded-range fallback. Saved7776 modular phase/raw states (432 conductor-drop states),12 Fraction vertices and exact overlap controls ran once; independent reviewers read code/results without replay. No actual Gauss energy/training/spectral/Lean result is claimed. This cone is finished; do not rerun it as new research or publish these notes as a status-only release.
NEXT genuine arithmetic: estimate the ACTUAL r>Y, Ny>X and remaining finite/infinite-frequency complement, retaining all original means, root/ray/S values, common-prime/full-union masks, equal-norm fibres, original capacity and common-profile uniformities BEFORE per-r/per-conductor absolute Cauchy; or derive additional-good-factor primal exclusions explicitly before Poisson. Whole signed13/12/global unsigned9/8, short/principal assembly, independent positive determinant/zero-location bridges separately for Dirichlet/Hecke/Dedekind/elliptic/generalArtin and actual analytic/category Lean/GRH remain OPEN. No Artin holomorphy, GRH, desired positivity or zero locations are assumed.
All scientific edits/builds/checks ran on iska-linux-vscode via SSH. The existing Mac gh login was used ONLY for publishing streamed reviewed Linux payloads in memory; no login/token read/copy/transfer, dependency/toolchain install, infrastructure/GPU, external MCP or Claude/global configuration change occurred. Preserve sourceHEADdba37ff/index0ef1af, all16HEAD/index states/15release worktrees including the new frozen331indexff207f16, all originals/outputs/caches/environments/history/private active models and Mac workspace/old8765preview. Linux4.24approval is still unanswered: no install/download/substitute/reinventory/repeated question. P4 claimf654cf42-34cf-465d-9b2c-1b2dcb7828d7 and prior P7/P9 remain in_progress. Both completion gates are FALSE; keep ONE existing heartbeat ACTIVE. No cleanup/reset/no-op/status commit or proof schedule.
## Full physical-row mask derivation — UNREVIEWED draft, 7 November 2026 UTC
Public Main aeb7705c/Pages163b9eaf and the verified348-page/10858MathML book remain unchanged. This is research outside the frozen331-file release and current canonical book/articles, not a new accepted theorem or a publication.
P7 claim1f11e152-be0d-4385-a859-c05e50253a83 attempted the actual full six-factor primal exclusions before Poisson. Draft planning/full-row-cone-20261106/root/FULL_ROW_PRIMAL_MASKS_AND_CONE-v1.md SHA965a8b56 derives ell|v2v3v4v5 and(k,v2v3v4v5)=1 from the first-five SF/pairwise-coprime row support. Arbitrary overlapping v6 is a column mask and capacity factor, not a new(v1,v6) or(k,v6) restriction. The literal v1=p,v6=p row has physical valuation7 and is allowed. At auxiliary v1=p², forbidding k=p because p|v6 destroys the required1-1 SF cancellation.
The draft separates the q6 column-prefix mask from exact long-variable deletion. If q6 is represented only in Qlong, its Möbius deletion scalar must retain nonzero d|q6 terms unless another actual character zero applies. Giving that scalar a q6 zero while dropping its inner mask restores forbidden columns. Keeping q6 as a finite raw character is a different valid representation that must retain its inner mask and period cost. The draft uses the pure-deletion representation, original finite ray/root/S phases, higher-order rho distinctions, good-only gamma=ell*z*r^5, original nonprincipal unit predicate before Poisson, all original raw/prefix/fullQ/common-W/capacity masks and the original A_U whole-factor grouping.
A quantitative extension to all six row factors is a CANDIDATE ONLY. At the specified finite cone Ny<=D^(3/8),Nr<=D^(7/144), its proposed leading budget is H*A2^(-5/6)*A3^(-11/6)*A4^(-17/6)*A5^(-23/6)*A6^(-5), with explicit smaller Bessel/equal-norm errors and legal modulus/profile/capacity thresholds. It is NOT added to the accepted results or best global bound.
New exact formal controls0cfeda91 passed once:225 primal-inversion states,45 column-mask states,108864 NEW j3/j4/j5 phase states with9072 raw conductor-drop states,7 capacity vertices and4096 six-coordinate Fraction samples. These are formal local identities, not actual ideal/Gauss energy data, a spectral computation, analytic proof checker or Lean. Producer3eeebd3e and33 source pins3191490b are bound by ATTEMPT-v1.json1470a0d4. No old7776/FE/tail/scalar/norm/training/build/browser/publication audit was replayed.
All three requested independent reviewer turns failed because of a usage limit before substantive outputs; no review gate exists for this extension. Status stays UNREVIEWED, P7 stays in_progress and both completion gates stay FALSE. Next concrete task is independent actual-support/phase/deletion/uniformity review of the saved draft, followed by the actual r>Y/Ny>X/restfinite/infinite-frequency complement. Do not rerun these finished formal controls as new research, promote this draft to the book or publish a status-only update. Whole signed13/12/global9/8, short/principal assembly, positive determinant/zero-location separately Dirichlet/Hecke/Dedekind/elliptic/generalArtin, actual analytic/category Lean and GRH remain open. No GRH/Artinholomorphy/desired positivity/zero premise was introduced.
All new work ran on iska-linux-vscode via SSH, with no public branch, canonical artifact, dependency/environment, auth/credential, external MCP, Claude/global configuration or Macworkspace/old8765 change. Preserve all16HEAD/index states/15release worktrees/sourceindex0ef/frozen331indexff207 and every original/output/cache/history. Linux4.24approval remains unanswered: no install/download/substitute/reinventory/repeated question. Keep ONE existing heartbeat ACTIVE.
## 7 November 2026 UTC — finite dual frequency research draft
Author-checked only; independent review pending. See planning/radical-frequency-20261107/HANDOFF-v1.md and outputs/checks/2026-11-07-radical-frequency/ATTEMPT-v1.json. Using the accepted mixed two-pivot kernel, counting frequency ideals by the radical outside the second pivot gives a candidate target-size bound for a larger finite projection, including squarefull r through D^(7/60) at Ny<=D^(3/8). The draft records the exact cuts, fixed-profile costs and preserved raw masks. The separate entire Ny<=D^(3/8) projection has upper exponent 155/144 at the critical line. The pure block's large squarefree frequencies still exceed the target in this method. Fresh exact ideal-index and exponent controls passed; they are not Gauss-energy or Lean evidence.
P7 checkpoint bb7c27a2-6ac3-4ddd-ab07-a27de2fd413d stays in progress. Next bound the actual joint phases before the positive frequency count. Whole 13/12, global 9/8, determinant/zero-location, Lean and GRH remain open. Both completion gates are false. The verified 348-page/10858-MathML book remains current.
## 7 November 2026 UTC — balanced finite-prefix research draft
Author checked; independent review pending. See planning/joint-r-frequency-20261107/HANDOFF-v1.md and outputs/checks/2026-11-07-joint-r-frequency/ATTEMPT-v1.json. The sixth-free norm/gcd lift is inherited from the accepted book, not a new analytic theorem. The new band localization combined with mixed6456 at Ystar=D^(2/21)A2^(-1/7) gives candidate main H X^(1/6)D^(-1/63)A2^(-41/42), X<=D^(1/2), with explicit common-W/error costs and full original masks. At X=D^(3/8) the pure SAME finite prefix improves13/16 to803/1008 (gain1/63), and the mixed witness31/48 to71/112 (gain1/84). These are finite-prefix comparisons; whole13/12/global9/8 remain unchanged. The pure intermediate-r budget still has gap47/1008. Fresh reduced-pairing/exponent controls pass; independent review, the joint intermediate-r/short-k estimate, full high frequencies, determinant/zero-location, Lean and GRH remain open.
P7 checkpoint 3a2cb4a4-c536-4cb9-aac9-74669bc6a747 stays in progress; P4/P9 and both false gates unchanged. Published348/10858 book and protected states preserved. No publication or book promotion. Next estimate the actual joint phases or signed-d sum before positive triangles, without replaying finished audits.
## 7 November 2026 UTC — joint k/r research draft
Author checked; independent review pending. See planning/joint-kr-20261107/HANDOFF-v1.md and outputs/checks/2026-11-07-joint-kr/ATTEMPT-v1.json. Combining actual k,r in a physical sixth-free frequency retains the erased exponent-two prime mask and puts both original columns inside one norm before separate frequency triangles. The candidate target-size exact finite projection Ny<=D^(3/8) with Nk>=D^(47/1296) preserves the actual pre-S cap, common W, primitive/source/ray phases and all zeros; on g=unit the cutoff is D^(47/1848). Exact finite grouping and exponent controls pass. The unit and shorter k terms remain open, as do the whole13/12/global9/8 comparisons, high frequencies, determinant/zero-location, Lean and GRH. P7 checkpoint 7a06bd85-e518-4417-a20e-1afc30dd486f stays in progress; both gates remain false. No book/release promotion; public348/10858 edition preserved.
## 7 November 2026 UTC — two shifted columns research draft
Author checked; independent review pending. See planning/unit-k-divisor-20261107/HANDOFF-v1.md and outputs/checks/2026-11-07-unit-k-divisor/ATTEMPT-v1.json. The new fixed auxiliary d transport preserves distinct short cutoffs and shifts the physical cubic numerator to d*k before both column estimates. It covers the former intermediate Nr~D^(2/21) range even at k=unit. The candidate whole finite prefix at X=D^(3/8) has pure power113/144, a gain1/36 over the established same-prefix13/16; the pure intermediate Nr~D^(1/6) gap5/144 remains. Exact finite transport, raw phase and exponent controls passed. This does not change the whole13/12/global9/8 bounds, high-frequency or determinant/zero-location obligations, Lean or GRH. P7 checkpoint 418becb6-9ce1-45fd-81a3-7194f14b37d7 stays in progress; both gates false. No book/release promotion; public348/10858 edition preserved.
## 7 November 2026 UTC — joint short-prefix/frequency draft
Author checked; independent review pending. See planning/prefix-r-20261107/HANDOFF-v1.md and outputs/checks/2026-11-07-prefix-r/ATTEMPT-v1.json. The short raw R_p(r) already deletes p/r collisions; on live support the joint physical norm uses literal t=r*p^4. The new finite-prefix candidate pure exponent is865/1104, a gain2/69 over the established same-prefix13/16. The next pure transition R~D^(4/23) still exceeds target by37/1104 in this upper budget. Exact raw transfer/grouped Cauchy/Fraction controls passed. Whole13/12/global9/8, high-frequency and determinant/zero-location obligations, Lean and GRH remain open. P7 checkpoint a6d695b6-2093-4447-b121-88ccd375d659 stays in progress; both gates false. No book/release promotion; verified348/10858 edition preserved.
## 7 November 2026 UTC — primitive-conductor prefix draft
Author checked, independent review pending. See planning/structured-prp-20261107/HANDOFF-v1.md and outputs/checks/2026-11-07-structured-prp/ATTEMPT-v1.json. On live pairs \(t=rp^4\), the true good primitive conductor is \(\operatorname{rad}(r)p\). The ordinary all-primitive sieve in the joint positive norm gives the candidate normalized pure finite-prefix exponent \(37/48\) at \(X=D^{3/8}\), a gain \(1/24\) over \(13/16\). The candidate reaches target size for \(X\le D^{1/4}\). The new pure transition leaves an upper-budget gap \(1/48\) at the critical prefix. Finite CRT support and rational exponent controls passed. P7 checkpoint d334a65d-0688-4535-8d27-ad43ffcffd67 stays in progress and both completion gates are false. The full signed estimate, higher-frequency complement, determinant and zero-location obligations, Lean and GRH remain open.
## 7 November 2026 UTC — quadratic cross terms in the finite prefix
Author checked, independent review pending. See planning/quadratic-cross-20261107/HANDOFF-v1.md and outputs/checks/2026-11-07-quadratic-cross/ATTEMPT-v1.json. Poisson in the dual Gram form of the live literal \(rp^4\) operator produces two quadratic cross matrices after handling all four shared-prime cases. The all-sixth-free Type-I estimate reduces the previous \(R,P\asymp D^{1/4}\) comparison from \(D^1\) to \(D^{5/6}\) at \(E=1\). The candidate pure finite-prefix power is \(257/336\). The new comparison is target-sized through \(X\le D^{2/7}\) and leaves a \(5/336\) gap at \(X=D^{3/8}\). Exact residue, CRT and exponent controls passed. P7 checkpoint 38ee8e3e-53d1-4ece-9b14-3658feb33233 stays in progress. Both completion gates are false and the full signed estimate, determinant and zero-location obligations, Lean and GRH remain open.
## 7 November 2026 UTC — smooth long factor in the finite prefix
Author checked, independent review pending. See planning/smooth-long-prefix-20261107/HANDOFF-v1.md and outputs/checks/2026-11-07-smooth-long-prefix/ATTEMPT-v1.json. Reuniting the long dyads at literal small \(Np\) permits the exact common \(W\) on the safe Mellin line. The surviving original long deletion is supported on \(\operatorname{rad}(s)\). Together with the previous pending quadratic comparison this gives candidate pure finite-prefix power \(109/144\), target size through \(X\le D^{1/3}\), and an upper-budget gap \(1/144\) at \(X=D^{3/8}\). Exact regrouping, raw deletion and exponent checks passed. P7 checkpoint aa5eddf4-278c-4a20-b4cb-f622b8b0be67 stays in progress. The full signed estimate, determinant and zero-location obligations, Lean and GRH remain open and both completion gates are false.
## 7 November 2026 UTC — sixth-part counting error at the balanced prefix
Author checked, independent review pending. See planning/sixth-part-mean-20261107/HANDOFF-v1.md and outputs/checks/2026-11-07-sixth-part-mean/ATTEMPT-v1.json. The exact two sixth-part deletion masks yield lcm(d,bi) residual subsequences. A physical sixth-free r norm before the divisor triangles gives normalized counting error B_W^2(DH)^epsilon (H/A2)(X/R)^(1/18). With the earlier pending low-side inputs, the pure critical remainder after retaining the signed high principal mean has power 271/360. The principal mean still permits the 1/144 upper-budget excess, and the whole-prefix candidate is unchanged. The finite ideal-mask and rational power checks passed. P7 checkpoint e1088b0c-464b-4c5a-8f81-f7a07ed008dd stays in progress. The full signed estimate, determinant and zero-location, Lean and GRH remain open and both gates are false.
## 7 November 2026 UTC — signed central principal density
Author checked, independent review pending. See planning/principal-macro-20261107/HANDOFF-v1.md and outputs/checks/2026-11-07-principal-macro/ATTEMPT-v1.json. In the central dyadic band D^(13/40)<=R<=D^(27/80), a group of intermediate scale in either original column permits the earlier joint norm. One-large boxes coalesce and use short density divisors plus the physical-norm coefficient tail. The removed signed part costs H D^(-1/480) with fixed profile/subpowers, conditional on prior (Q) and (M). Both remaining columns have two or three large factors. Two exact coprime ideal columns confirm a surviving two-large coefficient. The whole-prefix candidate remains 109/144. P7 checkpoint 6fdbce10-ee4f-4cad-81ed-80e04782c6cd stays in progress. The full signed estimate, determinant and zero-location, Lean and GRH remain open and both gates are false.
## 7 November 2026 UTC — exact high-cutoff gap pairing
Author checked, independent review pending. See planning/gap-factor-20261107/HANDOFF-v1.md and outputs/checks/2026-11-07-gap-factor/ATTEMPT-v1.json. In the stated central band and 2D/7<=U^2<=D, sum the signed label positions after clipping the literal small product. The above-cutoff complement forces leftover norm below 7. Three large labels cancel and two yield exactly the ordered above-U factorizations of the original column. After restoring the signed coprime projector the normalized replacement costs H D^(-1/480) with fixed profile and subpowers, conditional on previous (Q)/(M). The independent-of-(Q)/(M) thin-gap estimate removes a narrow interval above U. The comparable far-from-cutoff term remains open. Exact signed allocation and endpoint controls passed. Whole-prefix candidate remains 109/144. P7 checkpoint 4b4b6472-dc97-4c28-bb9e-62dbf4060955 stays in progress and both gates are false.
## 7 November 2026 UTC — prime prefixes in the high-cutoff gap
Author checked, independent review pending. See planning/gap-prime-prefix-20261107/HANDOFF-v1.md and outputs/checks/2026-11-07-gap-prime-prefix/ATTEMPT-v1.json. Exact rank and cutoff separation makes an intermediate prime prefix in either column usable. The smaller prefix range and the combined-cofactor selection use the inherited primitive comparison. The wider range uses prior (Q). These parts and the large squarefull frequency part cost H D^(-1/480) in the actual signed central gap, with fixed profile and subpowers. The remainder has two primes per column with combined cofactor norm below D^(1/24) and squarefull frequency part below D^(1/32). Fresh exact comparisons, ideal cases and exponent checks passed. Whole-prefix candidate remains 109/144. P7 checkpoint f2ad88b4-2e4d-4f13-9d26-c1582b9444aa stays in progress and both gates are false.
## 7 November 2026 UTC — joint cofactor bound in the signed two-prime pairing
Author checked, independent review pending. See planning/gap-core-pair-20261107/HANDOFF-v1.md and outputs/checks/2026-11-07-gap-core-pair/ATTEMPT-v1.json. Shared marked prime ideals cost H D^-11/96 in the separated product, leaving the signed projector on the small cofactors. The inherited primitive norm on both cofactor groups gives H (X/R)^(1/6)/sqrt(C C') with fixed profile and subpowers. The joint product at least D^(1/48) costs H D^-1/480 and the marked inert section costs H D^-17/160. The remaining comparable split-prime pairing has joint cofactor norm below D^(1/48). Exact projector, cutoff and exponent checks passed. Whole-prefix candidate remains 109/144. P7 checkpoint 1025101f-29a4-49c4-a827-0898f951f370 stays in progress and both gates are false.
## 7 November 2026 UTC — joint Gaussian completion in the unit-cofactor split-prime comparison
Author checked, independent review pending. See planning/gap-double-poisson-20261107/HANDOFF-v1.md and outputs/checks/2026-11-07-gap-double-poisson/ATTEMPT-v1.json. Joint completion with powers 2 and 5 retains the dual argument beta alpha^4, the quadratic CRT multiplier, both local Gauss phases and raw zeros. Counting labelled dual ideal fibres gives squared auxiliary budgets D^(67/80) on the true diagonal and D^(39/40) on the dual good conductor range B<=D^(3/8), including Gaussian tails. Larger dual conductors remain open; counting them absolutely would cost D^(3/2) instead of inherited D. No selected signed original part or whole exponent is claimed. Exact CRT, frequency, ideal and scale controls passed. Whole-prefix candidate remains 109/144. P7 checkpoint a1c9b4f8-7bf2-4d1d-8e26-d4c9b2cf5c0e stays in progress and both gates are false.
## 7 November 2026 UTC — conductor components in the joint Gaussian comparison
Author checked, independent review pending. See planning/gap-dual-colours-20261107/HANDOFF-v1.md and outputs/checks/2026-11-07-gap-dual-colours/ATTEMPT-v1.json. Exact common-Q Fourier separation retains the CRT phase, both Gauss factors, the diagonal correction and all raw zeros. The quadratic conductor range B2<=D^(7/24) costs D^(157/160); the cubic range B3<=D^(31/80) costs D^(59/60); sparse first frequencies cost D^(467/480). These are auxiliary off-diagonal sectors only. The full separated sixth-order comparison costs D^(169/160), still worse than inherited D on the original squared column. The jointly large conductor sector remains open; no selected signed original term or whole exponent is claimed. Exact labelled ideal and rational controls and finite Gaussian recombination passed. Whole-prefix candidate remains 109/144. P7 checkpoint 365d1018-6af7-46ce-8f97-84c964ad8d05 stays in progress and both gates are false.
## 7 November 2026 UTC — overlap and Gauss compensation in the joint Gaussian comparison
Author checked, independent review pending. See planning/gap-dual-overlap-20261107/HANDOFF-v1.md and outputs/checks/2026-11-07-gap-dual-overlap/ATTEMPT-v1.json. Literal Gauss duplication cancels the identified varying Jacobi angle, retaining the positive cubic Gauss factors, period-four multiplier and all raw twists. With g the full good gcd of the dual arguments and d the sixth root of the quotient beta/alpha_g, the joint cut (Ng)^(5/6)(Nd)^4>=D^(1/12) costs D^(39/40) for that auxiliary off-diagonal sector. This is not a bound on a selected original signed term. At coprime squarefree principal scales the separated budget D^(1+r/6) still loses to inherited D. Exact phase, raw deletion, and rational checks passed. The signed pairing and whole-prefix candidate 109/144 are unchanged. P7 checkpoint 0a6f1c68-1a19-4c64-a6ac-934ae7aab3d2 stays in progress and both gates are false.
## 7 November 2026 UTC — prime-norm strip and source-dependent short prime error
Author checked, independent review pending. See planning/gap-norm-strip-20261107/HANDOFF-v1.md and outputs/checks/2026-11-07-gap-norm-strip/ATTEMPT-v1.json. Exact inclusive Fourier separation lets the previous auxiliary exclusions be subtracted inside the close-norm strip, giving D^(59/60) for |Nq-Nqprime|<=D^(7/48) with mass and subpowers. Separately, conditional on the pinned strict 7/8 preprint, the fixed-field discrepancy yields D^(35/48+eps) for the original short prime-counting error at H=D^(3/4), including the lower endpoint and stated profile norm. No new zero-free theorem, coherent K0-J estimate, or selected original signed bound is claimed. Finite prime counts, strip reassembly, and scale checks passed. Far norms, whole candidate 109/144 and the signed estimate remain open. P7 checkpoint 1bbd94a1-3b50-4195-9805-0a010d8effba stays in progress and both gates are false.
## 7 November 2026 UTC — coherent large-divisor tail for continuous J
Author checked, independent review pending. See planning/coherent-j-tail-20261107/HANDOFF-v1.md and outputs/checks/2026-11-07-coherent-j-tail/ATTEMPT-v1.json. The full prime-power deletion placements and half-open Abel expansion separate the actual cutoff jumps before the unweighted physical norm. At critical H, the original nonunit continuous J tail is target-sized from F>=D^(1/8) with elementary sigma=1, and from F>=D^(1/11) using the fixed-field Mertens consequence of the pinned external 7/8 theorem. The latter gives D^(179/240+eps) at F>=D^(1/10), with the stated profile norm. No new zero-free theorem, small-F or annular K0 estimate, or bound on K0-J is claimed. Exact ideal, Abel, and scale checks passed. P7 checkpoint d5def2ad-a3f0-44fc-8a1f-300dfd3bb693 stays in progress and both gates are false.
## 7 November 2026 UTC — coherent tail for the full annular divisor
Author checked, independent review pending. See planning/coherent-annular-tail-20261107/HANDOFF-v1.md and outputs/checks/2026-11-07-coherent-annular-tail/ATTEMPT-v1.json. The exact high-window annular coefficient is mu times the ordered above-U factor count minus mu. Its full intermediate common divisor splits into one, two or four cells; largest-cell signed Abel retains the max jump and full prime-power deletion before the inherited physical norm. Conditional on the pinned fixed-field 7/8 input, the critical full A-A tail is target-sized at D^(4/7), versus elementary D^(5/8), and costs D^(71/96+eps) at D^(3/5) with profile norm. This includes rough d, so it is not a smooth K0 or original high-gcd bound. Small d including d=1 and K0-J remain open. Exact coefficient, allocation, Abel and scale checks passed. P7 checkpoint 0649306b-405d-479e-8353-9536fa418887 stays in progress and both gates are false.
## 7 November 2026 UTC — two signed cells in a smooth full-divisor tail
Author checked; independent review pending. See planning/coherent-matching-tail-20261107/HANDOFF-v1.md and outputs/checks/2026-11-07-coherent-matching-tail/ATTEMPT-v1.json. On actual unweighted two-pivot annular A-A, disjoint opposite cells retain two independent strict gap cutoffs and two signed Mertens prefixes before the row norm. Exact mutual-coprimality inversion and joint prime-power radical unions retain all raw masks. At critical H, the smooth full intermediate-divisor tail is target-sized from D^(10/19) conditional on the pinned fixed-field 7/8 Mertens input, with bound D^(237/320+eps) at D^(11/20) and explicit profile norms. Elementary D^(5/8) and conditional sharp D^(4/7) remain distinct. The sharp limit loses weighted log Fourier integrability. No small-d/d=1, smooth K0-J, original high-gcd, whole signed estimate, determinant or GRH result. Exact finite controls passed. P7 checkpoint 31fdd8e0-a123-4043-a9a0-9a28bf266804 and both gates remain open.
## 8 November 2026 UTC — actual mixed coherent bands near U
Author checked; independent review pending. See planning/coherent-mixed-band-20261108/HANDOFF-v1.md and outputs/checks/2026-11-08-coherent-mixed-band/ATTEMPT-v1.json. Keep two signed annular prefixes and shorten both actual pivots separately according to the frozen prime valuation. Conditional on pinned fixed-field 7/8 Mertens, the selected smooth same-divisor K0-J and K0-KS band below U is target-sized from D^(52/105), with power 8621/11520 at D^(199/400). Above U on a fixed smooth (U,cU), c<7, the excluded rough divisors are single primes: pair the original smooth annular band with the f=1 rough term at the same prime norm. The matched power is 215/288 with the pinned Mertens and PIT inputs and the band endpoint/masks as stated. This does not identify the above-U selector with one on f in J. Fresh exact finite controls passed. Small d and d=1, pure, unmatched continuous, whole signed bound, determinant, Lean and GRH remain open. P7 checkpoint 27f95326-c6dc-4b52-a95e-ccc603afe809 and both gates stay open.
## 8 November 2026 UTC — product cutoff in the full signed form
Author checked; independent review pending. See planning/coherent-product-cut-20261108/HANDOFF-v1.md and outputs/checks/2026-11-08-coherent-product-cut/ATTEMPT-v1.json. One exact multiplier on the full intermediate divisor transports to F, FP and PQ in the smooth, rough-prime and unordered distinct semiprime branches. The new geometric middle powered bound retains two signed prefixes after valuation-specific row deletion. The critical mixed D^(52/105) and balanced D^(8/17) cutoffs, conditional on pinned fixed-field Mertens, leave only the small annular sum for the entire original signed form, with no PIT input there. The cubic-dominant wedge also admits an elementary rough-free reduction with the inherited powered sieve. Pure D^(10/19) retains the product-cut short term with F << D^(1/38). Its true endpoint and masks remain. Exact finite controls passed. Small annular d and d=1, whole signed bounds, determinant, Lean and GRH remain open. P7 checkpoint 2c2eb81b-8e05-464b-ad8d-5b1bdf5637a9 and both gates remain open.
## 8 November 2026 UTC — actual small cutoff-product amplitude
Author checked, independent review pending. See planning/small-cutoff-product-20261108/HANDOFF-v1.md and outputs/checks/2026-11-08-small-cutoff-product/ATTEMPT-v1.json. For the original ordered cutoff divisors with N(xy)<=Z, split the unused squarefree residual into an exact square-divisor expansion. Keep the full raw masks and possible overlap of the two auxiliary factors. Nonprincipal primitive Poisson for small square divisors and the inherited sixth-free norm at fixed second pivot and sixth factors give (Z/Nd)[R^2/C0+A2 Phi6(A1,C0)] after normalization. The mixed choice C0=D^(31/66), Z=D^(1/20) gives D^(493/660+eps) summed positively over d without the fixed-field power. At d=1 this leaves the complementary amplitude as the target. Across signed d the cross term remains open. Exact finite controls passed. P7 checkpoint 40021686-9f8b-4a68-ae53-a120be476a25 and both gates remain open.
## 8 November 2026 UTC — dual average for cutoff-product amplitudes
Author checked, independent review pending. See planning/dual-cutoff-product-20261108/HANDOFF-v1.md and outputs/checks/2026-11-08-dual-cutoff-product/ATTEMPT-v1.json. For each deletion factor use the shorter direct or radial Poisson length in the actual row norm, keeping the exact conductor profile, Gauss root and S-supported dual phases. With the original squarefree expansion this bounds the selected normalized positive amplitude by (Z/Nd) A2 Phi6(A1,sqrt(R)) with explicit fixed-profile seminorms. Mixed Z=D^(1/8) costs D^(53/72+eps) after harmonic summation over d, and pure bounded Z meets the target. Gaussian normalization, mask and exact scale checks passed. The larger cutoff products and signed cross term remain open. P7 checkpoint 8618fe02-8ef5-47e0-a462-20158059d259 and both gates remain open.
## 8 November 2026 UTC — joint average for cutoff-product amplitudes
Author checked, independent review pending. See planning/joint-cutoff-product-20261108/HANDOFF-v1.md and outputs/checks/2026-11-08-joint-cutoff-product/ATTEMPT-v1.json. At fixed full deletion, group the cutoff and unused factor in the direct column or keep the cutoff and Poisson frequency together through two fourth moments. Retain the exact row conductor profile, Gauss root, S-supported phases and raw masks. The selected mixed positive amplitude reaches the target up to cutoff D^(29/120). Cutoff D^(7/30) gives D^(107/144+eps) after harmonic summation over d with explicit profile seminorms. Exact direct and finite dual algebra checks passed. The larger-product amplitude and signed cross term remain open. P7 checkpoint c46107d9-ef7e-4547-8428-f3e2fcc644e5 and both gates remain open.
## 8 November 2026 UTC — full cutoff coefficient at small dual frequencies
Author checked, independent review pending. See planning/low-dual-cutoff-20261108/HANDOFF-v1.md and outputs/checks/2026-11-08-low-dual-cutoff/ATTEMPT-v1.json. Fix a low dual ideal and average the cutoff column, keeping exact Q_u, Gauss root, S-supported phases and raw masks. The scale inequality M>=RJ/D caps the effective cutoff length at min(Z,DK0/R). For the full original cutoff coefficient the direct, low, and far branches give a positive budget of mixed D^(67/90+eps) with K0=D^(1/20), and pure D^(3/4+eps) with K0=1. Fresh exact frequency-split and scale checks passed. The middle window and signed cross remain open; the whole form and published edition are unchanged. P7 checkpoint 8b1d84e9-6ac6-48ff-8b41-fdf9991ecab9 and both gates remain open.
## 8 November 2026 UTC — reduced ideal ratio in the middle dual window
Author checked, independent review pending. See planning/reduced-dual-window-20261108/HANDOFF-v1.md and outputs/checks/2026-11-08-reduced-dual-window/ATTEMPT-v1.json. Cancel only the shared ideal phase in z0/k, keeping its exact row zero. Single-factor and exact coprime fourth-moment estimates control new middle pieces for the full original cutoff coefficient: mixed powers 67/90, 11/15, 107/144; pure 3/4. Fresh exact phase, grouping, and coprimality controls passed. Large reduced factors outside the joint rectangle and the signed cross remain open; the whole bound and public book are unchanged. P7 checkpoint 6e8fe815-2ded-4bf0-bd1e-1b822f3e6200 and both gates remain open.
## 8 November 2026 UTC — sixth powers in the reduced dual denominator
Author checked, independent review pending. See planning/dual-powerful-denominator-20261108/HANDOFF-v1.md and outputs/checks/2026-11-08-dual-powerful-denominator/ATTEMPT-v1.json. Counts for the exact decomposition b=s*b0*v^6 bound two pieces of the remaining dual window. Pure unit good core and the arbitrary-core set Nv>=D^(13/120) meet 3/4 for all reduced numerators. Mixed short core with the stated numerator cutoff gives 67/90. Mixed large sixth factor with its stated cutoff gives 107/144. Fresh exact ideal, phase and scalar controls passed. The sixth-free complement and signed cross remain open. P7 checkpoint 1d46a1b9-0c33-4c3e-9be6-db90c9eea4f3 and both gates remain open.
## 8 November 2026 UTC — powered roots of dual denominators
Author checked, independent review pending. See planning/powered-dual-root-20261108/HANDOFF-v1.md and outputs/checks/2026-11-08-powered-dual-root/ATTEMPT-v1.json. On the actual residual b=s*f*c^m, average the cutoff and powered root in two fourth moments with the exact coprimality phase and original physical norm. The mixed cube and square sectors give 89/120 and 2687/3600. Pure gives 133/180 and 893/1200 at the stated cuts. Fresh exact phase, grouping and exponent controls passed. The large squarefree complement and signed cross remain open. P7 checkpoint 874b2bb8-b5c8-4952-ac98-c4c55e0b948b and both gates remain open.
## 8 November 2026 UTC — conjugate overlap in the dual window
Author checked, independent review pending. See planning/conjugate-dual-overlap-20261108/HANDOFF-v1.md and outputs/checks/2026-11-08-conjugate-dual-overlap/ATTEMPT-v1.json. After the ideal gcd, factor the conjugate overlap and preserve its exact phase and zeros. Two coupled coprimality projectors use least common multiples. Mixed integer reduced ratio bounds are 67/90, 11/15, and 107/144 on their stated ranges. The mixed equal-norm signed cross for Nd>=D^(1/5) reaches 3/4 by an absolute tail comparison. Fresh norm, phase, lcm, and exponent checks passed. The small-d cross and large integer reduced ratios with small overlap remain open. P7 checkpoint 5456a3ec-bca5-4ec2-8d47-c3e87a397bcf and both gates remain open.
## 8 November 2026 UTC — equal-norm cross with the near strip
Author checked, independent review pending. See planning/equal-near-dual-cross-20261108/HANDOFF-v1.md and outputs/checks/2026-11-08-equal-near-dual-cross/ATTEMPT-v1.json. The actual mixed equal/near signed cross reaches power 3/4 for reduced integer gaps up to D^(1/8), including the unit divisor. Up to D^(73/216) the noncomparable part, short comparable dyads, and high-length radial variation reach target cross. This uses exact sharp masks and absolute Cauchy. The wider leading term and far small-d cross remain open. Fresh exact phases, additive masks and rational profile controls passed. P7 checkpoint 6348d325-6332-41fe-bf63-2ce423bdda1d and both gates remain open.
## 8 November 2026 UTC — rational ideal content in the dual cross
Author checked, independent review pending. See planning/rational-content-dual-20261108/HANDOFF-v1.md and outputs/checks/2026-11-08-rational-content-dual/ATTEMPT-v1.json. On the actual mixed wide near strip either rational content at least D^(23/216) gives cross 3/4. When both are at least D^(13/120), this also holds across the full comparable middle range, including far gaps and the unit divisor. The proof keeps exact physical norms, phases, and masks, then uses absolute Cauchy. The remaining small-content and far noncomparable small-divisor cross is open. Fresh exact algebraic controls passed. P7 checkpoint ecd3cc89-db33-4a67-85ee-ac280eed39d7 and both gates remain open.
## 8 November 2026 UTC — grouped cutoff and deletion in the equal-norm cross
Independently reviewed research. See planning/equal-rational-sieve-20261108/HANDOFF-v1.md and outputs/checks/2026-11-08-equal-rational-sieve/ATTEMPT-v1.json. Grouping h=r e0 in the exact equal-norm middle projection gives the full mixed cross with its original complement at power 35/48, including d=1. The pure tail Nd>=D^(11/32) has power 3/4. The auxiliary cubic Gram bound L+B^(5/3) controls selected pure divisor ranges. Fresh exact algebraic checks include repeated frequencies and raw zeros. Pure small-d equality and the off-equal squared assembly remain open. P7 checkpoint 730db3a4-4a66-47f4-81fb-69ce3dd8ffd3 and both gates remain open.
## 8 November 2026 UTC — pure cutoff separation and unequal overlap
Independently reviewed research. See planning/pure-equal-cross-20261108/HANDOFF-v1.md and outputs/checks/2026-11-08-pure-equal-cross/ATTEMPT-v1.json. The pure equality cross against the full original complement has power 3/4 when the eligible conjugate overlap norm is at least D^(5/24), including d=1. This keeps the exact ordered allocations before the cubic L6 and rational L3 comparison. Grouping h against the actual unequal denominator also gives target cross for two selected unequal sectors against their entire complements. Fresh exact phases, endpoint and repeated-frequency controls passed. Small eligible overlap at the natural h scale and generic unequal interactions remain open. P7 checkpoint 4e8d8997-fec1-44be-9d36-bdc66662cfc1 and both gates remain open.
## 8 November 2026 UTC — matched and low-conductor Poisson cross
Independently reviewed research. See planning/matched-poisson-20261108/HANDOFF-v1.md and outputs/checks/2026-11-08-matched-poisson/ATTEMPT-v1.json. The matched good-pattern equality cross is bounded by R up to subpowers, and the signed defect comparison reaches target for Nq<=D^(1/2) in both critical blocks, including d=1. The conductor here belongs to the row character in the pairing. A separate exact complete allocation window has middle scalar -1 plus a subpower error, with the unit supplied by the low piece. Fresh exact phase, zero, cutoff and unit controls passed. High defects and the whole signed assembly remain open. P7 checkpoint 9cea1dc5-c60f-452a-9bfc-3319e6257913 and both gates remain open.
## 8 November 2026 UTC — directed defects and coherent erasures
Independently reviewed research. See planning/defect-order-cross-20261108/HANDOFF-v1.md and outputs/checks/2026-11-08-defect-order-cross/ATTEMPT-v1.json. The actual signed cross meets target for joint cutoff and conjugate erasures with arbitrary d residue changes, without a conductor cap, using the coherent cutoff-defect sum before row Cauchy. The local-order comparison also reaches target for every non-erasing conductor and extends the reverse unit sector to oriented D^(9/10). Exact phase, local weight and conjugate charge checks passed. Large competing insertions and erasures, the whole signed assembly and both gates remain open. P7 checkpoint 16172d73-4889-422a-8ff2-783b63473fe8.
## 8 November 2026 UTC — two coherent defect families
Independently reviewed research. See planning/two-family-defect-20261108/HANDOFF-v1.md and outputs/checks/2026-11-08-two-family-defect/ATTEMPT-v1.json. An exact additive equality projector and ideal Mobius inversion allow two coherent raw character sums with the original cutoff and zero masks. The selected comparable sector meets target at d=1 in both critical blocks. The capacity-wide powerful-increase estimate covers arbitrary mixtures of increments 2 through 5 with the stated erasures and d-residue changes. Large residue-one supports in the full divisor sum remain open. P7 checkpoint 5ebadc90-c6e3-4b74-a363-b4a2e2a42cbc.
## 8 November 2026 UTC — sharp insertion completion
Independently reviewed research. See planning/increment-one-20261108/HANDOFF-v1.md and outputs/checks/2026-11-08-increment-one/ATTEMPT-v1.json. Sharp primitive Poisson completion of the unweighted insertion array and an exact middle-equality count give comparable powers 57/80 mixed and 13/16 pure in the original selected cross. The estimates allow mixtures of positive increments 1 through 5. The pure Nd<=D^(1/8) prefix and additional localized sectors reach target. A prime-d slice shows cancellation of the outer sign by the actual equality cutoff. The remaining low-residual pure divisor range and both gates remain open. P7 checkpoint 07eb50ec-b8bd-46ab-aca4-a7b855f505e8.
## 8 November 2026 UTC — coupled pure cutoff comparison
Independently reviewed research checkpoint. See planning/coupled-cutoff-20261108/HANDOFF-v1.md and outputs/checks/2026-11-08-coupled-cutoff/ATTEMPT-v1.json. Coupling the insertion dual with the shorter actual ordered cutoff factor and grouping the powerful sparse tail improves the selected pure comparable all-divisor exponent from 13/16 to 55/68, including increments 1 through 5 and all physical comparable scales in this scope. The target divisor prefix expands from 1/8 to 9/68. Large powerful parts and additional central-scale label sectors reach target for all divisors. The exact middle equality test, original raw masks and signed identity remain. Low-residual small-powerful nonunit divisors beyond the prefix, noncomparable scales, the determinant and zero-location bridges and both gates remain open. The frozen 348-page public edition is unchanged. P7 checkpoint d51f9799-f78e-4f1a-83f8-0203fd5af64e.
## 8 November 2026 UTC — asymmetric physical cutoffs
Independently reviewed research checkpoint. See planning/asymmetric-cutoff-20261108/HANDOFF-v1.md and outputs/checks/2026-11-08-asymmetric-cutoff/ATTEMPT-v1.json. The original selected mixed cross has power 57/80 at all physical ratios and divisors, and reaches target throughout the stated critical-capacity first-pivot strip 7/24 through 439/912. In pure, short insertion and long erasure sectors reach target under the stated physical thresholds and residual selections. The exact middle subtraction keeps both endpoints and uses the two-way physical norm count. Intermediate pure rectangles, remaining signed divisor cancellation, the determinant and zero-location bridges and both gates remain open. The frozen public edition is unchanged. P7 checkpoint ba9e13cd-154c-4fd6-969f-d1f3b2a02121.
## 8 November 2026 UTC — pure intermediate cutoff ranges
Independently reviewed research checkpoint. See planning/pure-intermediate-20261108/HANDOFF-v1.md and outputs/checks/2026-11-08-pure-intermediate/ATTEMPT-v1.json. In the original selected pure cross, target holds for all erasures when K is at least of order D^(1089/1420), for all insertions when B is at least of order D^(1355/1482), and for short K up to D^(921/1652) uniformly over B at least of order R. The ordered assignment signs recombine below the cutoff and leave an exact strict correction band above it. A thin physical norm interval reaches target, while passive fixed-label Mertens gives only subpower cancellation. The coherent wide-band divisor sum, whole signed estimate, determinant and zero-location bridges remain open. P7 checkpoint 13a42796-9d69-474a-94ed-1cb595983f6c.
## 8 November 2026 UTC — paired cutoff appendix staged
See planning/cutoff-appendix-20261108/HANDOFF-v1.md and outputs/checks/2026-11-08-cutoff-appendix/STAGE_ACCEPTED-v1.json. The independently reviewed appendix, original regime figure and five PDF editions passed source-bound staging. The native candidate preserves all 10858 earlier MathML subtrees and passed 41 focused browser checks. The published edition remains 348 pages. Independent research records exact wide correction signs and fixed-label subpower cancellation, full-frequency partner small-prime compensation and an equality-side target-size prefix Nd <= D^(13/56) in the pure comparable block. The new wide notes are outside the staged appendix. The partner-only unit-pair correction, remaining coherent sum and determinant comparison are open. P7 checkpoint f31ee721-2f5b-4c28-886a-ac936316c997.
## 8 November 2026 UTC — paired cutoff and wide correction edition
The integrated candidate has 363 pages, 40 chapters and 11,657 native MathML expressions. Its focused PDFs have 88 and 257 pages. The notes have 46 and 146 pages. The appendix preserves the 10,858 previously published MathML expressions byte-for-byte and adds one original regime diagram. The wide correction sections give the exact unit and small-prime band, fixed-label subpower cancellation, full-frequency compensation and the equality-wide pure comparable prefix Nd <= D^(13/56). The staged sources, PDF and native checks are bound in outputs/checks/2026-11-08-cutoff-release/INTEGRATION_ACCEPTED-v1.json.
Separate research gives a primal reduction and two physical moments for the full-frequency unit-pair contribution. It reaches the target on the balanced split Nm,Nn at least of order D^(1/8), with all-divisor pure upper power 19/24 and the Nd >= D^(1/8) upper power 37/48. See planning/cutoff-release-20261108/research/UNIT_PAIR_FULL_FREQUENCY-v1.md and the independent review. The unbalanced unit-pair estimate, coherent selected modes, original signed assembly, determinant/zero-location and Lean obligations remain open. P7 checkpoint 21bae760-dacd-4a98-a531-c901aa431dcb.
## 9 November 2026 UTC — higher unit-pair moments
The research checkpoint is recorded in planning/unit-pair-core-20261108/HANDOFF-v1.md and outputs/checks/2026-11-08-unit-pair-core/RESEARCH_ACCEPTED-v1.json. For the original pure full-frequency unit-pair component, the reviewed all-divisor upper power is now 37/48. The fixed-moment unbalanced dyads and rational content at least of order D^(1/24) reach target size. Actual selected frequencies reach the target through D^(9/20) over all d and D^(1/2) for Nd at least of order D^(1/8), including the two exact coprimalities with empty middle equality. The separate-moment upper comparison has a 1/48 excess on the balanced edge. The joint wide core, higher frequencies, equality-wide beyond the existing prefix, determinant and Lean remain open. This checkpoint is outside the unchanged 363-page public edition. P7 checkpoint 74e89b0a-7962-4e72-820e-e0ab9ddf1433.
## 9 November 2026 UTC — joint unit-pair and equality-wide moments
The scoped proofs and evidence are in planning/joint-unit-pair-20261109/HANDOFF-v1.md and outputs/checks/2026-11-09-joint-unit-pair/RESEARCH_ACCEPTED-v1.json. The original pure full-frequency unit-pair cross has target size for all Nd up to order D^(1/6), without a content, overlap or thin-product selection. The forced radical of the grouped h^2 b_s column is stripped inside the positive physical comparison with exact raw masks. Actual selected modes admit a joint equality-frequency moment, exact label-indexed compensation and all lengths for Nd at least of order D^(27/56). Frequency content, S-part and sixth-power subsets and the equality-wide rational Gram estimate give additional scoped target ranges. The low-content unbalanced band, unrestricted selected main scale, whole signed form, determinant bridges and pinned Lean remain open. These results are outside the verified 363-page edition. P7 checkpoint 42cebb77-3938-4312-aec5-042e9367dd25.
## 9 November 2026 UTC — joint cutoff edition and middle-divisor research
The 370-page edition has 40 chapters and 12,011 native MathML expressions. The joint cutoff appendix keeps all 11,657 earlier literal expressions and all 14 canonical tools. Focused articles have 90 and 263 pages and the notes have 47 and 152 pages. The reviewed low-divisor pure full-frequency unit-pair estimate reaches target at Nd up to order D^(1/6). The actual selected frequency prefix is D^(15/32) over all d, and all frequency lengths are covered for Nd at least of order D^(27/56), with the actual selected middle subtraction. The equality-wide proof has its separate support thresholds and selected comparable prefix 111/476. Source, builds and browser evidence are bound by outputs/checks/2026-11-09-mid-divisor/INTEGRATION_ACCEPTED-v1.json.
Separate new research in outputs/checks/2026-11-09-mid-divisor/RESEARCH_ACCEPTED-v2.json covers selected middle-divisor cross-conjugate overlap, smooth main frequency bands including a good sixth-free selection, and a thin equality product strip. Signed summation over the large common ideal gives explicit subpower cancellation; the fixed 1/16 gap of that upper comparison at the core remains. Low-overlap mid-divisors, sharp main-frequency selections, the whole signed form, determinant/zero-location bridges and analytic/category Lean remain open. P7 checkpoint 09f4c3d2-4a90-4392-a2fb-28f645416e13.