# Research outcomes **No result in this register establishes twin-prime infinitude.** Read the relevant question before starting an attempt, alongside [QUESTIONS.md](QUESTIONS.md) and the current state in [G2-STATE.md](G2-STATE.md). An unsuccessful estimate can leave a useful lemma; both belong with the same question. This register records current conclusions, exact limits and the evidence needed to assess them. Git records revisions and authorship. Organize work by its mathematical question and outcome. Detailed derivations and reproducible data stay in their owning notes and scripts. ## Recording and reuse For each question or precisely scoped approach, record: - **Outcome and calibration:** exact identity, derived bound, finite measurement, conditional implication, insufficient bound, missing input, or refuted claim. State the actual scope and quantifiers. - **Established result:** what can be reused, including lemmas from an otherwise unsuccessful attempt. - **Limit or failed step:** what remains unproved, or the precise statement that fails. An unanswered estimate stays open. - **Evidence:** the owning proof, theorem inputs, validator and retained data. - **Reuse or revisit condition:** what a subsequent attempt must consume or change. Repeat a calculation only for a stated correctness concern or changed hypothesis, parameter range or implementation. Update the relevant entry in place when understanding changes. Keep useful identifiers for citations; Git preserves superseded wording. Existing S/F identifiers remain stable citations. New findings can use a question-based identifier. An ANSWERED question is not a blanket proof grade, and a green mechanical gate does not validate every mathematical statement. DERIVED uses named inputs; VERIFIED means finite checking, not an asymptotic theorem. Verify the argument behind a finding before it determines a research decision. Check successes as well as failures. In particular, a closure used to exclude an approach must establish that exclusion with the relevant hypotheses and quantifiers. A review that finds a defect or missed scope is useful work; record the correction here. Authorship and repeated agreement do not settle the mathematical claim. After each attempt, update the owning note and this register, regenerate [QUESTIONS.md](QUESTIONS.md), and keep [TODO](../TODO.md) forward-only. The closed-route table below is a compact index of scoped closures; add a route there only when a statement or specified approach is actually closed. ## By research question ### Proof specification and logical implications #### Research execution — Reviewed closeout and document ownership **Grade:** SPECIFICATION and reviewed integration only. **Question:** Q-research-execution; onboarding question Q-agent-start. **Established:** the dispatch from 1285d47 ended at 46fa948; all A/A2/B/C/D/E and V/W returns are present. [RESEARCH-EXECUTION.md](RESEARCH-EXECUTION.md) owns the integrated board and candidate next obligations. The independent 2026-09-09 review accepts A/A2 and D at their scoped estimates, repairs B/C/E quantifiers and preserves W's source limitation. Entry documents and generated indexes agree with the owning notes. **Limits:** no sufficient signed margin, proof date or fresh compute allowance. The completed dispatch is not a running assignment. **Evidence:** [review](research-round-validation.md), lane proofs and validators; incoming commits remain in Git. **Next/reuse:** select a changed bounded obligation, source match and reviewer before further agent work. The first candidate is D's corrected small-gcd moment target; A2's actual B is an independent alternative. #### Research-round validation — Independent proof and scope corrections **Grade:** REVIEWED at the exact analytic, source and finite scopes stated. **Question:** Q-research-round-validation. **Established:** A's truncation and A2's repaired low Type I estimate; B's ordinary-BV family identity; C's norm obstruction on all smooth inputs; D's block estimate and fourth residual cut; E's aggregate sieve input and certified failure of two tests for all fixed u>4. **Corrections:** D's missing moment exponent is 7/200, not 17/200; C has 1-bounded nonzero Fourier modes; E's constant is aggregate, not per-tuple; B's uniform Dickman error is 12/11!, not the specialized 24/22!; A2's two sufficient consumers are not equivalent. Bookkeeping and endpoint repairs are in the owning proofs. Broad route exclusions are replaced by the actual failed steps. **Evidence:** [review](research-round-validation.md), [regressions and rational certificates](research-round-validation.js), primary source statements and the retained JSONs. No long census rerun. **Next/reuse:** consume the corrected scope and actual coefficients. Green gates or several agreeing reviews do not settle an unproved margin. #### Transition research round — Prepared worker specs and execution contract **Grade:** SPECIFICATION only; executed and subsequently reviewed, with no signed estimate supplied by the specification. **Questions:** `Q-transition-research-execution`, `Q-transition-energy-review-spec`, `Q-transition-source-match-spec`, `Q-transition-joint-budget-spec`. **Established:** [TRANSITION-RESEARCH-EXECUTION.md](TRANSITION-RESEARCH-EXECUTION.md) is the unified handler brief at mathematical baseline `0292082`. It fixes the full coefficients, exact signed transition kernel, four-term decomposition and outside residual. The linked specs define independent [energy review](transition-energy-review-spec.md), [source matching](transition-source-match-spec.md) and [joint-budget research](transition-joint-budget-spec.md); the handler retains the central signed estimate and final mathematical integration. Each assignment has bounded work, owned report paths, source/quantifier checks, acceptance criteria and a precise failed-step return option. **Limits:** preparing the specifications is not executing the assignments. The conditional consumer template supplies no unproved lower bound. No signed saving, region, exact cut or sufficient twin margin changes. The prior 3/50 regional moment threshold is not the transition target. **Evidence:** the four specification documents and their links to the owning derivations. Baseline norm and transition conclusions remain subject to the independent audit; finite gates do not settle them. **Execution/reuse condition:** the round was executed from `cb51292` and reviewed in [transition-round-audit.md](transition-round-audit.md). Its current dispositions are in [transition-round-review.md](transition-round-review.md). Do not redispatch these completed assignments unchanged; a specified new estimate or correctness concern can justify further work. #### Cross-campaign synthesis — Reusable links and proposal corrections **Grade:** DERIVED elementary identities and bookkeeping; scoped source and proposal review. **Question:** `Q-cross-campaign-synthesis`. **Established:** [cross-campaign-synthesis.md](cross-campaign-synthesis.md) compares the full-coefficient, fold, anchored and literature routes; checks all eight local proposals and selected primary statements/proof steps; and gives the common-scale union-bound condition for combining average errors. It repairs the anchored note's reversed Euler-factor inequality, the Suen proposal's independence/atomicity and universal-closure language, and a proposal grade header that contradicted its recorded decision. All proposal grades are retained, with no publication action. **Failed inferences:** separate unbounded good sets need not intersect; the present subfamily rate cannot absorb its normalization losses through the stated absolute-average conversion; a smoothed norm and a bounded- multiplicative subfamily theorem are not estimates for the same coefficient. The reviewed quadratic-form main term with an O(D2^2) remainder does not handle the long divisor support. No exhaustive literature closure follows. **Evidence:** owning derivations and exact checks in [cross-campaign-validation.js](cross-campaign-validation.js); the external analytic proofs remain named dependencies, not numerically verified results. **Next/reuse condition:** use the sharp-transition follow-up below and its signed correlation budget, or another matched estimate with all costs. No favorable twin margin is established. #### Research-handoff — Assessment and assignment contract **Grade:** CONSOLIDATED specification and research judgment; no new arithmetic estimate or numerical evidence. **Question:** `Q-research-handoff`. **Established record:** [RESEARCH-HANDOFF.md](RESEARCH-HANDOFF.md) assembles the exact E_dagger, its sufficient positive and rescaled-average margins, the completed generic moment and the small-common-divisor target. It also specifies the complete C3 residual and centered prime–Mobius alternative, with the moving cutoff and sufficient tolerance retained. Current assignments and integration are owned by [RESEARCH-EXECUTION.md](RESEARCH-EXECUTION.md). **Assessment:** another bounded attempt is justified. The regional improvement does not establish that a full proof is near; the global margin remains OPEN. The generic estimate does not exploit additional Mobius-sign cancellation. No claim that such cancellation is necessary for every proof follows. The 0.995 budget and the greater-than-3/50 moment saving are exponents, not completion or finite-improvement percentages. A smaller signed domain is not a monotonicity statement. **Evidence and limits:** the mathematical sources are [grouped-divisor-moment.md](grouped-divisor-moment.md), [residual-coverage.md](residual-coverage.md) and [endpoint-target-audit.md](endpoint-target-audit.md), with their upstream dependencies. Their finite checks are not independent expert review. The handoff itself completes no assigned research and adds no proof. **Reuse or revision condition:** require a checked derivation or correction, a matched new estimate with all costs, and an explicit regional or global payoff before changing the assessment. Record each attempt by question; do not split histories by agent class or promote a failed estimate to a refutation of the arithmetic target. #### Independent handoff review — Completed bounded audit **Grade:** DERIVED corrections and scoped source/proof review. **Question:** `Q-handoff-review-0906`. The bounded preparation checklist is complete, with deep imported theorems retained as named dependencies. Reviews 20 and 21 check the local reduction, grouped moment and coverage steps; repair the cumulative/dyadic comparison, full-corner scope, shrinking-margin inference, equal-frequency scope and prime-power tail; and correct later window/rate and identity-piece overclaims. Joint Cauchy is priced and adds no region. No full-margin proof or universal impossibility theorem is supplied. **Evidence:** [live review](handoff-review-0906.md), [report 20](history/reviews-0906/20-independent-handoff-review.md), [report 21](history/reviews-0906/21-readiness-review.md). **Reuse:** consume the actual derivations and fixed hypotheses. [AGENT-START.md](AGENT-START.md), question `Q-agent-start`, routes incoming workers to the current handoff and execution plan. The executed round is preserved in [round-review-0906.md](round-review-0906.md); the band transfer retains its own PARTIAL scope and review record. #### S-0905-01 — Corrected the proof specification **Grade:** DERIVED logical corrections and VERIFIED finite algebra. **Question:** `Q-review-0905`. The review distinguishes sieve-method limitations from the information in the exact tile, preserves the occupancy quantifiers, and corrects the quadratic-bound deduction and the CRT alignment claim. Reuse the corrected definitions in [README Status](../README.md#status). **Evidence:** [review](history/staging/review-0905.md), [validation](review-0905-validation.js), [correction history](history/CHANGELOG.md). **Limit:** no improved bound on G₂ or on twins follows from these repairs. #### F-0905-01 — Treating all exact tile arguments as parity-blind **Verdict:** WITHDRAWN CLAIM. **Question:** `Q-review-0905`. Exact divisibility arrangements retain primality information. A limitation of a specified sieve's statistics and tolerated errors was generalized to every possible argument on the representation. **Evidence:** [review](history/staging/review-0905.md), [corrected current scope](../README.md#status). **Revisit only with:** a named method class, retained statistics, errors and quantifiers, and a proof at that scope. Existing particular failures stand. The separate assertion of a proved universal floor at 4 was also withdrawn; its owning closure is in [closed-route index](OUTCOMES.md#closed-routes). #### F-0905-02 — Inferring little-o from a constant quadratic bound **Verdict:** REFUTED DEDUCTION. **Question:** `Q-review-0905`. The comparison g≤G₂=(log x)^(2/5) for representations by 1-bounded functions on every W_L-smooth input, via growing binomial witnesses and the PNT. It is reviewed, including on 2026-09-09. **Failed steps and correction:** a finite norm example alone was not unboundedness; the growing witness now supplies that restricted result. The composite filter is not multiplicative; unweighted cancellation cannot be multiplied through its arithmetic envelope. The claim that only zero Fourier modes are 1-bounded is false: the sufficient condition |k|log W_i/log x<=1/3 admits at least k=0,±1 on the left and l=0,±1,...,±6 on the right, for every Mellin twist. Those modes still need the other hypotheses and filtered o(x/log^2 x) rate. The full family does not meet the unit-disc condition. Only a rough-pair upper bound, not a matching lower bound, is used. **Evidence:** [lane validator](full-coefficient-average-validation.js), [review](research-round-validation.md) sections 3,8, [integration controls](research-round-validation.js). **Limits/revisit:** sparse norm witnesses do not exclude density-one representations or components with paid growth. A matched filtered correlation theorem, jointly treated Fourier coefficients or a new one-sided inequality may reopen the question. No region or twin margin follows from the representation. ### Singleton fibers #### S-0905-12 — Identified the support and cost of singleton fibers **Grade:** EXACT regrouping; DERIVED mass lower bounds using prime BV, PNT and the Möbius mean estimate; finite checks VERIFIED. **Question:** `Q-singleton-fiber-audit`. The grouped remainder vanishes whenever either integer is prime. For the prescribed cutoffs, the positive and negative ungrouped singleton masses are each at least c*x*log(x) for all sufficiently large x, with c>0. Reuse the witness construction and the sign-density proof when pricing a bound. **Evidence:** [proof and parameter budget](singleton-fiber-audit.md), [independent coefficient, fiber and progression checks](singleton-fiber-validation.js). **Limit:** neither the signed singleton difference nor the longer-fiber sum is estimated. This is not a twin lower bound or a novelty claim. **Related failure:** F-0905-12. #### F-0905-12 — Discarding singleton fibers or pricing each sign separately **Verdict:** REFUTED at the ungrouped absolute-mass scope. **Question:** `Q-singleton-fiber-audit`. The actual β-weighted singleton family has both positive and negative masses at least c*x*log(x) eventually. Its absolute mass is not O(x), and neither is its negative mass alone. Dropping all positive terms and bounding the negatives separately cannot close a scale-x lower bound. **Evidence:** [derived lower bounds](singleton-fiber-audit.md), [finite partition controls](singleton-fiber-validation.js), [owning closure](#closed-routes). **Reuse:** S-0905-12. **Revisit only with:** cancellation across different fibers, a grouping that preserves useful signs before taking absolute values, or a different decomposition. The signed singleton sum and its combination with longer fibers remain OPEN; the mass lower bounds do not refute either target. ### Reusable data and retained inputs #### S-0905-11 — Recovered useful inputs from existing runs **Grade:** VERIFIED finite reconstruction. **Question:** `Q-data-reuse-audit`. The old fold CSV supplies 1,230 base primes through 10,007 and exact interval boundaries. Three complete archived rows are reproduced; bounded prefixes are reconstructed with full prime-power factors retained for future divisor weights. The saved aggregate files and producer workspaces are inventoried, so subsequent work need not repeat the data search. **Evidence:** [retention audit](data-reuse-audit.md), [validator and bound output](singleton-fiber-validation.js), [generated factors](data-reuse/factor-windows.json). **Limit:** prefix measurements are not full dyadic residuals, and the largest tested interval still has inner cutoff 2. **Related failure:** F-0905-13. #### F-0905-13 — Reading the new weighted residual directly from old summaries **Verdict:** REQUIRED FIELDS NOT RETAINED in the inspected outputs. **Question:** `Q-data-reuse-audit`. The inspected Liouville, tile-window and zone-gap outputs save aggregate statistics, not per-integer divisor signs and β weights. Twin records alone sample zero grouped coefficients of the present residual. The old fold CSV does retain prime and interval inputs, and they have been reused. **Evidence:** [artifact/producer inventory and reconstruction](data-reuse-audit.md). **Revisit only with:** an additional raw-factor or equivalent arithmetic artifact, or a specified recomputation from retained coordinates. This is not an impossibility claim about reconstructing arithmetic from integer inputs, nor an audit of unavailable backups. **Reuse:** S-0905-11. ### Signed divisor grouping and endpoint discrepancies #### Full corner coefficient energy — Sharp identity and smoothed norm **Grade:** DERIVED from exact algebra, elementary squarefree counting and the imported Graham/Barban–Vehov mean-square theorem; finite identities VERIFIED. **Question:** `Q-corner-coefficient-energy`. **Established:** [corner-coefficient-energy.md](corner-coefficient-energy.md) gives the exact sharp energy over short cofactors k,l, retaining complex conjugation, off-diagonals, all prime-log weights, non-squarefree inputs and clipping. Its counting error is O_epsilon(x^(1/2)K^2 x^epsilon), below x at the fixed corner. A logarithmic average of the lower divisor cuts yields full smoothed coefficients with squared norms O_eta(x log x), and hence the same absolute bound for their shifted product. The short-endpoint range of Graham's input is handled by an exact rescaling. Radical regrouping and coprime small-prime multiplication give exact identities; the latter produces a cutoff difference, not multiplicativity. **Limits/failed shortcuts:** Jensen cannot recover a sharp norm from its smoothed counterpart; dropping the prime singleton cancellation gives a false coefficient. The follow-up below prices the sharp norm and refutes a negligible L2 transition for sufficiently small fixed eta. A signed transition saving and the global complement remain OPEN. This is no three-log gain for E_dagger and no extension of a bounded multiplicative correlation theorem. Exact residual cuts and twin status are unchanged. **Evidence:** the owning proof, pinned Graham source statement and [cross-campaign-validation.js](cross-campaign-validation.js) with active deletion controls. The imported theorem is not proved by the finite tests. **Next/reuse condition:** consume the sharp-transition result below; a proof route still needs signed estimates and the remaining global margin. #### Sharp corner transition — Complete energy and a failed norm transfer **Grade:** DERIVED using the uniform de la Breteche–Dress–Tenenbaum mean-square theorem, PNT and exact algebra; finite identities VERIFIED. **Question:** `Q-sharp-corner-transition`. **Established:** [sharp-corner-transition.md](sharp-corner-transition.md) gives O_eta(x log^2 x) squared norms for both complete sharp coefficients and the same absolute shifted-product bound, for each fixed eta in the existing range. For every sufficiently small fixed eta>0, their squared norms on the actual dyadic intervals, and those of sharp minus smoothed, are Theta_eta(x log^2 x). The proof pays O(eta^2 x log^2 x) for all cross-prime terms against an Omega(eta x log^2 x) diagonal, and removes proper powers by an L2 bound. No diagonal-only lower bound is used. **Failed step:** an O_eta(x log x) sharp squared norm or o_eta(x log^2 x) transition squared norm fails in that small-fixed-eta range. The easy O(D^2) floor error cannot replace the uniform source input; applying the smooth sieve-weight bound at smoothing order zero also fails its hypothesis. These failures do not preclude cancellation between the transition coefficients at shift 2. **Limits:** the small-eta threshold and onset are not numerically extracted. No lower bound for a shifted product follows from individual norms. The mixed, smoothed and transition signed sums, global complement and sufficient twin margin remain OPEN. No region or exact cut changes. **Evidence:** source locators and pinned PDF in the owning note; [sharp-corner-transition-validation.js](sharp-corner-transition-validation.js) checks exact coefficient vectors, the full prime energy matrix, clipping, coprimality, singleton and proper-power corrections with active controls. An initial validator assertion distinguished JavaScript -0 from 0; normalizing signed zero repaired that test, without changing the identity. Finite checks do not establish the imported asymptotic. **Next/reuse condition:** estimate the signed transition-pair kernel with its actual arithmetic weights; retain the mixed terms and common-scale global consumer. Norms alone cannot supply that estimate. #### Moving cutoff — Published endpoint omission and a checked conditional tolerance **Grade:** source interfaces CHECKED; endpoint counterexample EXACT; dyadic repair and conditional implication DERIVED; finite controls VERIFIED. **Question:** Q-moving-cutoff-parity. **Established:** [moving-cutoff-parity.md](moving-cutoff-parity.md) reads the published Murty--Vatwani mechanism before local analysis. Its fixed-residue simplification is already explicit in print. A rearrangement on p. 654 drops n+h>ey. At x=20,y=3,h=2, the prime-only version acquires the extra positive term log(3)*(log(13)+log(19)). The PDF image and an exact formal-logarithm calculation check this. It refutes the displayed equality, not the conditional theorem. The repaired dyadic formula is S=C2*x-2*C2*M+D_y+O_A(x/log^A x), where M=sum_J Lambda(n-2)mu(n) and D_y is the fully specified centered progression discrepancy with n>ey retained. Ordinary BV handles the squarefree short term at y=ceil(x^(12/25)). The remaining cofactor moduli are at most x^(13/25), for a different sequence than ordinary prime BV. The actual discrepancy has not been bounded. **Conditional payoff:** the unconditional squarefree density and rational Euler-product enclosures give 33/200=-4x/25+o(x) on unbounded dyadic scales would give S>=x/200+o(x). This is an OPEN sufficient condition, not an estimate or proof-progress measurement. The relation D_y=B_L+2*C2*M+negligible error makes its connection to the existing residual explicit. **Source retrieval:** Vatwani's 2018-10-02 author preprint for the 2019 Math. Z. paper is now read at the owning note's stated scope. Its one-odd-exponent theorem concerns mu times squarefree indicators; its twin theorem has two unfilled near-level-one hypotheses. Its general theorem's main-term restriction must be checked before application to Lambda. A second exact control checks its formal swap before (6.23): at x=40,Q=3,h=2 the dropped cut adds 2*log(3)*log(5). Section 6's moving cuts require care; no full reproof of that theorem is claimed. **Failed payoff/limit:** no new signed bound, unconditional margin, asymptotic measurement or established literature novelty. The formula does not establish that its arithmetic estimate is easier than B_L. **Evidence:** [validator](moving-cutoff-validation.js) checks both Lambda conventions, 40 dyadic exact reconstructions, independently ordered progression/density sums and rational infinite-product enclosures. Deleting the boundary, density projection, even terms or proper prime powers changes a control. Source locators and hashes are in the note. Candidate validation was completed in this pass. **Next/reuse condition:** obtain a matched estimate for the actual centered prime/Mobius sequence and price it against (13)--(16), or give a different justified full-residual estimate. Do not repeat retrieval, the source switch or the constant calculation as a new cancellation mechanism. #### Centered discrepancy census — D_y at finite x is the classical term's error **Grade:** MEASURED, two pre-registered falsifiers (neither fired), four controls, one independent reimplementation agreeing to 1e-14, one cross-codebase check; no asymptotic content. **Question:** Q-centered-discrepancy-measurement. The sufficient input (16) of moving-cutoff-parity remains OPEN. **Result:** D_y(x), M(x), the identity pieces of (12) and the absolute sum W1 of (13) computed exactly at x=2^j, j=16..38 (3.9e9 support integers at the top, 8224 s). D_y/x stays within 0.0043 of zero for j>=26 against the threshold -0.16. W1/x decays like the random-sign control (slopes 0.827 against 0.825), within 8 percent of it at every j>=26, and reaches 0.0773 at j=38 against the threshold 0.08. Real D_y exceeds the control by up to 22 times with slope 0.863 against 0.544, and the exact identities pin D_y/x to -(T1/x - C2) within 2e-4 at every j>=30 (largest deviation 1.84e-4 at j=30), because S/x is within 1.8e-4 of C2 and the density term is negligible: the excess is the finite-x error of the classical BV/Mobius-mean term, of order x/log^2 x (independent recomputation of the identity columns: history/reviews-0907/10 §5). **Limit:** the parity object's own fluctuation is an order of magnitude beneath that error at every reachable x, so a census of D_y cannot inform the missing estimate; the absolute BV-type form is consistent with the data and the obstacle is proof, not truth. **Evidence:** [note](centered-discrepancy-measurement.md), [script](centered-discrepancy-measurement.js), [artifact](centered-discrepancy-measurement.json); code review in history/reviews-0907/02. **Reuse or revisit condition:** do not rerun or extend; there is no scale at which the census separates the two contributions. Reuse the exact identity check S=T1+T2+E_pp as a validator for any future implementation of the Vaughan pieces. #### Shifted-prime Mobius sums — A finite table of the parity input at 2^38 **Grade:** MEASURED, pre-registered falsifiers, three controls fired, one independent reimplementation and one cross-codebase check agreeing; no asymptotic content. **Question:** Q-shifted-prime-mobius-sums. The o(pi(x)) conjecture for Mobius on shifted primes remains OPEN. **Result:** sum_{5<=p<=x} mu(p+-2) and lambda(p+-2) computed to x=2^38 (10.9e9 primes, 3446 s). All four sums sit at random-sign size: the largest |U|/control-rms at any j is 2.8, and the slopes of log|U| over j>=28 (0.28 to 0.65) lie within the control's per-draw spread around 1/2. The +2 sums are negative at every j from 21 to 38; the note compares that run with the draws' runs and proposes no mechanism. The dyadic total M(x)=sum Lambda(n-2)mu(n) is at most 8.8e-5 x in magnitude for j>=29, against the trivial bound 0.374 x that moving-cutoff-parity (14)-(16) uses, so the -4x/25 tolerance is loose by four orders at these scales. Squarefree density of p+2 among primes 0.747911 at 2^38 against 2*Artin 0.7479116. pi(2^j)+2 equals OEIS A007053 at every j. **Limit:** a bound on M is itself a bound on the shifted-prime Mobius sum, the same parity object; the looseness of the trivial bound is not usable by a derivation. Searched in the owning convention (Mobius on shifted primes, Chowla two-point, Elliott--Halberstam twisted by Mobius; row in SEARCH-CONVENTIONS.md, record in history/reviews-0907/01): there is no published table to compare above x=10^4. The first run carried a wrong A2 constant in its header, caught by the data and corrected before any reading. **Evidence:** [note](shifted-prime-mobius-sums.md), [script](shifted-prime-mobius-sums.js), [artifact](shifted-prime-mobius-sums.json). **Reuse or revisit condition:** do not extend the table; a larger x sharpens nothing a proof can consume. Reuse the M column as the cross-check for any future implementation of the centered discrepancy. #### Joint correction — Known completion and the arithmetic still missing **Grade:** source interfaces CHECKED; exact reconstruction and support budgets DERIVED; formal integer/logarithm controls VERIFIED. **Question:** Q-joint-correction-source-audit. **Established:** [joint-correction-source-audit.md](joint-correction-source-audit.md) checks novelty first. Ford--Maynard Definition 7.1 already supplies prime-factor subset convolution; its complementary-sum estimate uses a separate Type II hypothesis. Mounier's signed smooth-number condition and Friedlander--Iwaniec's asymptotic sieve are additional source matches with unfilled input requirements. The full logarithmic large-factor weight D_i equals the existing short Vaughan approximant P_i minus the paid exceptional term E_i. The supports are x^(12/25) and x^(1/10); full mixed prime/approximant sums are within ordinary BV and their product has an elementary CRT error O(x^(29/50)log^2 x). This makes explicit which unrestricted weights are already controlled. **Failed payoff:** assembling the entire prime-partner/complement combination returns C_delta=S-C2*x-U_delta+o_delta(x). The surviving quantity is the already-open prime correlation/Vaughan remainder. That cancellation is not an additional estimate and gives no signed improvement. It does not close all possible joint estimates. **Branch control:** exponent vectors (12,13,75)/100 and (12,13,36,39)/100 have the same unrestricted weight -3/4 but different selected prime-cofactor weights. This distinguishes the factor-vector functions only; it makes no shifted-density or integer-information impossibility claim. The branch cannot silently inherit full support. **Source limits:** Mounier's direct application at y=x^(6/25) requires theta>19/25, beyond the currently matched input. Its signed condition is an assumption, not a theorem about this sequence. Friedlander--Iwaniec additionally requires its specific bilinear estimate and distribution beyond x^(2/3). Ford--Maynard's general Type II hypothesis is not supplied by the Pan--Ding aggregate theorem. **Evidence:** source locators, selected proof steps, PDF images and hashes in the owning note; [validator](joint-correction-validation.js) checks short/long reconstruction with repeated primes, the prime intersection term, CRT support and the formal branch control. The machinery is prior art; literature novelty is unestablished. **Next/reuse condition:** identify an additional estimate for the actual coefficients in the owning note's (12), or another justified one-sided bound for the full residual. A proposed theorem import must price its complete complement. Do not repeat this algebraic completion or harmonic optimization as if it supplied new distribution information. #### Supported coefficient — Dickman collapse and an insufficient sieve lower bound **Grade:** classical source interfaces CHECKED; harmonic identity and budget DERIVED; finite rational controls VERIFIED. **Question:** Q-supported-coefficient-dickman. **Established:** [supported-coefficient-dickman.md](supported-coefficient-dickman.md) performs the novelty check first and applies the classical Dickman convolution identity to the complete harmonic subset coefficient, with m=1 excluded and repeated-prime effects paid. At delta=1/100, H_delta=D(25/12)-1+E_delta with 0<=E_delta<=24/22!, hence -115223/4500 and locates a positive family with liminf P3/(C2*x)>1171/1125>1. The positive-family lower bound does not rely on S>=0. Both families retain prime exclusions and proper prime powers, and have unique ordered factorizations. **Failed shortcut:** liminf (N2-P3)/(C2*x)>5023/4500>11/10. The proposed pair plus the main term is insufficient. Its exact complement R_rest cannot be negligible: S>=0 forces liminf R_rest/(C2*x)>1/10. This implication supplies no improved twin lower bound. The closure is for this pair, current cutoffs and omission of its complement; broader signed grouping remains OPEN. The witnesses survive fixed delta<1/50 and all current plateau profiles. **Evidence:** owning source match, complete logarithmic and local-density budget, and [validator](paired-factor-budget-validation.js) with directed BigInt integral enclosures and sign controls. Published PDF pages, continuations and hashes are recorded. No effective onset is available. The original Pan--Ding paper and book restatement were not independently read; the import is Wu's published lemma. The machinery is prior art; literature novelty of the application is unestablished. **Next/reuse condition:** the whole supported harmonic and elementary sieve test is completed in supported-coefficient-dickman.md. Its joint prime-partner/complement estimate remains OPEN. Applying a nonnegative sieve directly to the signed coefficient would be invalid. Do not repeat either completed constant calculation for an unchanged consumer. #### Joint factor family — Type I part evaluated, signed prime partner remains **Grade:** DERIVED and reviewed identity and Type I evaluation; INSUFFICIENT separate-sign bound. **Question:** Q-joint-factor-estimate. **Established:** for the single-large-prime delta-rough family, R_F=C2*x*H_delta-Z_F^p+o_delta(x). Ordinary prime BV applies at each cofactor length >=sqrt x, to all odd moduli <=x^(1/10). The complement has Type I part -C2*x*H_delta, so these linear masses cancel. The local Mobius density constant is 2C2. Wu's squarefree-only lemma is used only for the upper sieve, not for nonsquarefree Type I moduli. **Failed step and repair:** the separate-sign coefficient 1+H_delta-(40/9)H_delta^+ is below -2/5; this is an insufficient lower bound, not a negative upper bound on the actual count. A constant 2 is still insufficient. The uniform Dickman error for 0=(H_delta-eta+kappa)C2*x+o(x), fixed 0=kappa*C2*x+o(x). Neither estimate is supplied. Generic o(x) for arbitrary bounded positive coefficients without a main term is not a valid substitute. **Evidence:** [note](joint-factor-estimate.md) sections 2,4, [lane validator](joint-factor-estimate-validation.js), [review](research-round-validation.md) section 9 and [rational controls](research-round-validation.js). **Next/reuse:** exact regrouping alone gives no new estimate, but does not exclude useful family inequalities. Reopen with a signed correlation and paid complement, different consumer or a correctness concern. #### Switching literature — Actual negative mass exceeds the main term **Grade:** DERIVED using ordinary prime BV, the nonnegative linear sieve, Mertens and exact algebra; finite certificates VERIFIED. **Question:** Q-switching-negative-mass. **Established:** [switching-negative-mass.md](switching-negative-mass.md) reads Matomäki--Zuniga Alterman's published switching framework and matches its linear-sieve interface. For the actual family n=p*q*r, p=b_L, r prime, with n-2 a W_R-rough composite, G_L=log r and G_R=-log(n-2), apart from paid proper powers. A rough lower sieve at s=4 minus a prime upper sieve at s=2 gives liminf Nhat/(C2*x)>=kappa=(10*log(3)-36/5)*I>3/2. The ordered two-factor integral I and a rational certificate are explicit. The shortest cofactor has BV power slack 1/450. The witness survives fixed delta<1/50 and all profiles with the same plateaus. The full identity forces liminf Phat/(C2*x)>1/2. **Failed shortcut:** discarding Phat and seeking Nhat=0 budget. **Source qualification:** Tao's 5 April 2020 comment identifies a gap in his blog proof of the linear-sieve theorem. The classical input remains, with the published Lemma 2.5 and book locators recorded; the Chen benchmark's citation is qualified accordingly. **Evidence:** owning proof, source pages and PDF hash; [validator](switching-negative-mass-validation.js) gives exact constant, full prime-log convolution and local-density controls. Literature novelty of the application is unestablished. **Next/reuse condition:** the proposed two-small-prime loss and three-small-prime gain are now compared in paired-factor-budget.md above. That pair is insufficient, with its complement retained explicitly. Use the follow-up's whole-coefficient assignment before another factor enumeration. Different cutoff exponents need a new budget; the present closures do not price them. #### Smooth sieve literature — Existing machinery and shifted localization **Grade:** primary-source matches CHECKED; localization and profile constant DERIVED using the named imports; finite algebra VERIFIED. **Question:** Q-smooth-sieve-literature. **Established:** [smooth-sieve-literature.md](smooth-sieve-literature.md) matches GKM's concentration results, CCHM's quadratic optimality, Green--Tao Appendix D, Goldston--Yildirim III and corrected Henriot. The C3 profile has derivative energy 700/429, versus one for the linear taper; this is a different objective from shifted absolute mass. Weakening the first logarithmic factor in the three-prime envelope gives a uniform growth-class majorant with bounded harmonic mass. For each fixed 0W_i)-F_i(s_i)*log(t_i)-E_i for the full logarithmic profiles, where s_i contains all small-prime factors, including powers. E_i is supported on inputs divisible by p^k>W_i with p<=W_i. An elementary reciprocal-sum count and divisor bounds pay its full shifted effect by O_epsilon(x^(39/40+epsilon)); proper prime powers are also paid. The complete residual is therefore a signed composite-filtered smooth/rough cofactor sum. It has no unpriced small-cofactor tail. The negative part has an explicit prime-filter correction, which is nonnegative but not estimated here. **Failed step:** the proposed F^- upper bound supported only on small-prime pairs with product >a is false. Ten small primes with every triple below a and every quadruple above b give F=-84 while that pair count is zero. The configuration has strict margins in the actual left exponent range; no shifted density for it is asserted. Linear Type I support cannot automatically be assigned to the nonlinear positive/negative part of F. **Limits:** this removes an exceptional term, not the main mixed-sign sum. Neither a new signed scale-x bound nor the sufficient twin margin is supplied. A bound on the negative part alone is sufficient, not necessary. **Evidence:** the complete derivation and [exact polynomial controls](global-factor-signs-validation.js), including logarithmic ramps, squareful inputs, prime zeros, prime-power corrections and an integer counterexample. The standard PNT input is used only for asymptotic existence of the single-input factor cell. **Next/reuse condition:** the higher-order majorant is now carried out in [global-smooth-majorant.md](global-smooth-majorant.md), giving a full absolute O(x) budget for a different admissible profile. The signed constant comparison remains OPEN, with prime filters and positive mass retained when needed. A census or pair-only classification does not provide it. #### Global cutoff averaging — Preserve the whole remainder before estimating it **Grade:** DERIVED uniform reduction, imported one-point norm and same-input cancellation; finite algebra VERIFIED. **Question:** Q-global-cutoff-averaging. **Established:** [global-cutoff-averaging.md](global-cutoff-averaging.md) averages the initial U and Y cutoffs in complete Vaughan identities, with explicit fixed margins for both prime and Mobius BV. This gives S=C2*x+mathcal R+O_A(x/log^A x), where mathcal R=sum G_L(n)G_R(n-2) includes every branch and the arithmetic previously assigned to E_out. Graham's mean square gives ||G_i||_2^2=O(x log x). For sufficiently small fixed eta, the existing sharp-corner norm then forces =-||C_i||_2^2+O_eta(x log^(3/2) x), where G_i=C_i+B_i. This is cancellation between the corner and its complement at one input. **Limits and failed inference:** it does not give the shifted cross terms, a lower margin, or a new signed O(x) bound; the latter already follows from positivity and the twin upper-bound sieve. Averaging only the corner would still leave its transition and E_out. The bounded-multiplicative subfamily theorem cannot be applied to G_i from its norm alone. **Evidence:** the owning uniformity table and derivations, primary Vaughan and Graham statements, the existing Mobius BV derivation, and [integer prime-log controls](global-cutoff-averaging-validation.js). Those controls retain prime powers, nonsquarefree inputs, independent averaging, the r=n correction and all mixed complements; they do not verify an asymptotic estimate. **Next/reuse condition:** the first factor-sign attempt is completed in [global-factor-signs.md](global-factor-signs.md), with a paid exceptional term and a refuted pair-trigger majorant. Continue from its full signed cofactor sum and prime-filter correction, retaining positive mass as needed. The priority is a research judgment, not a proved optimal route. Keep the small-cofactor transfer as a possible ingredient, with its full rate and coefficient mismatch explicitly priced. #### Cofactor progression transfer — A signed estimate on a growing restricted family **Grade:** DERIVED from named correlation and non-pretentiousness inputs; finite identities and controls VERIFIED. **Question:** `Q-cofactor-progression-transfer`. **Established:** [cofactor-progression-transfer.md](cofactor-progression-transfer.md) derives a dyadic scale-average bound O_eta(log^(2-d) X) after division by x, for some absolute d>0, for the actual prime-r coefficients restricted to remaining cofactors s,t<=floor((log x)^kappa), for some absolute kappa>0. It includes nonsquarefree inputs and every sharp, transition or smoothed profile pair. The sharp choice is the joint sum of the four restricted pieces. Modified multiplicative Euler factors represent mu(n/s) on s|n; the source is applied on the exact CRT progression modulo [s,t]. Retaining its density 1/[s,t] limits cofactor summation to O((log log X)^2). Uniform non-pretentiousness, Fourier integration, dyadic sampling, moving cutoffs, the strict transition jump and CRT rounding are priced. **Failed extension:** full cofactors reach x^(2eta+o(1)), so their progression moduli exceed the source's polylogarithmic range. Even a same-rate extension would incur O_eta(log^2 X) from termwise cofactor summation, exhausting the small saving. This rejects that application, not a collective signed estimate or another theorem. **Limits:** the restricted bound is still weaker than o(x). In the exact consumer E_dagger=E_out+K_H+J_H+negligible error, J_H contains both mixed cofactor tails and the tail product; neither J_H nor E_out has the needed one-sided control. No full-corner improvement, exact cut or twin margin is established. No numerical exponent, onset or literature novelty is claimed. **Evidence:** the owning derivation, Tao–Teravainen v2 Theorem 3.1(ii), MRT (1.12), and [finite controls](cofactor-progression-transfer-validation.js). The controls do not prove the imported asymptotic. **Next/reuse condition:** estimate J_H with its mixed terms, or retain cancellation while summing cofactor pairs. Price the rate at scale x and combine it with E_out on common scales. Do not repeat the completed small-cofactor transfer as an open assignment. #### Transition round — Direct review of the completed research **Grade:** DERIVED corrections and scoped review; finite controls VERIFIED; no signed estimate. **Questions:** `Q-transition-round-audit`, `Q-transition-round-review`. **Established:** [transition-round-audit.md](transition-round-audit.md) checks the returned arguments, validators and primary theorem statements. The original sharp energy chain survives at its stated scope. The exact cofactor and joint-cutoff formulations are retained after correcting the transition support. [The integration](transition-round-review.md) now gives one reconciled dependency and budget table. **Corrections:** the displayed cofactor kernel omitted its lower cut; the Mellin formula represented rho instead of tau; fixed-divisor singleton fibers were confused with long cofactor fibers; Theta's scalar mass did not price full T; separate upper bounds were treated as actual sizes; joint and cutoff routes were excluded beyond the inequalities proved. **Limits:** no signed improvement, region, exact cut or sufficient twin margin changed. The review does not reproduce every imported proof. **Evidence:** [targeted validator](transition-round-audit-validation.js), all four returned validators, the owning corrected reports and primary source locators in the review. **Reuse:** choose a signed cofactor or joint-cutoff interface, preserve clipping and both signs, and price a complete common-scale consumer. A source mismatch is not a closure of the broader method. #### Sharp corner transition — Independent review **Grade:** DERIVED argument retained using named imports; finite checks VERIFIED. **Question:** `Q-transition-energy-review`. **Established:** [energy review](transition-energy-review.md) retains the sharp upper norms, sufficiently-small-fixed-eta lower norms and failure of negligible L2 transition, with all cross-prime terms paid. The upper constant can be taken as 2B eta(1+o_eta(1)). Citation corrections distinguish Graham's original result from its quotation in An and correct DBT's volume. **Limits:** B, eta_0 and the onset are not extracted; ineffectivity is not established. The cross-prime support interval is only enclosing. Prefix positivity would not imply the required dyadic positivity. The unretained scratch probe is not evidence. S(z,z)=1 does not rule out sharper localized mean-square estimates; that closure is withdrawn. **Evidence:** [validator](transition-energy-review-validation.js), six active controls, and the owning energy derivations. Imported proofs remain imports, and signed correlations remain OPEN. **Reuse:** retain the eta range. Any investigation of cross-prime signs must use the actual shifted dyadic intervals and prime hypotheses. It is supporting work, not a required step toward a joint signed estimate. #### Transition source match — Inspected theorem interfaces **Grade:** DERIVED identities; scoped failed direct applications; one REFUTED relaxation. **Question:** `Q-transition-source-match`. **Established:** [source match](transition-source-match.md) retains the cofactor exchange and Theta identity. Fixed-divisor fibers are singleton at these cuts; fixed-cofactor fibers need not be. GY's displayed error is insufficient for the raw levels, GKM (1.15) is a one-point moment outside that raw range, specified divisor-function correlation theorems do not directly match T, and MRT averages shifts with bounded-function and parameter requirements. **Failed steps:** full coefficient representation, affine-form uniformity, clipping, sampling and rate costs remain unpriced. The transition Mellin kernel needs a sharp subtraction with a 1/s tail. Neither the scalar Theta expansion nor the sharp-subfamily normalization supplies a full representation for T. The rate test is representation-dependent. **Refuted:** fixed-shift cancellation uniform against an arbitrary bounded factor with one multiplicative factor, by the lambda counterexample. **Evidence:** primary locators and custody records in the note; [validator](transition-source-match-validation.js) checks finite algebra, not theorem applicability. No absence theorem or exhaustion of a method. **Reuse:** reopen a paper for a specified new reduction, proof adaptation or correctness issue; do not repeat an unchanged invalid substitution. #### Transition joint budget — Mixed terms and the combined inequality **Grade:** DERIVED identities and precisely scoped norm obstruction; OPEN signed target. **Question:** `Q-transition-joint-budget`. **Established:** [joint budget](transition-joint-budget.md) retains the exact split, no support leakage, the wider purely geometric inclusion, the cutoff-average identities and mixed second difference. Separate Cauchy upper budgets have leading order 2B eta x log^2 x. The specific average-triangle-Cauchy output is at least ||T_L||||T_R|| by Minkowski, and is saturated for sufficiently small fixed eta. The safe lower coefficient pays the proper-power and smoothed remainders first. **Corrections:** signed cutoff estimates remain open; separate control of R_00 is not necessary under every grouping. The window-average constraint uses theta_eff=min(theta,4B); a restricted-range theorem requires pricing its complement, not automatic rejection. **Evidence:** [validator](transition-joint-budget-validation.js), eight controls, plus the corrected derivations and audit countermodels. **Reuse:** a joint signed estimate may exploit cancellation before taking absolute values. Smoothed norm lower bounds are optional supporting work. No consumer input, signed saving or E_out bound is supplied. #### Transition signed estimate — Corrected cofactor and cutoff interfaces **Grade:** DERIVED identities and absolute budgets; OPEN local input. **Question:** `Q-transition-signed-estimate`. **Established:** [signed attempt](transition-signed-estimate.md) gives the single-base split and an exactly clipped two-affine-form cofactor average with tau_i=rho_i 1_(D_i,z_i), both gcd branches and prescribed residues. The existing proper-power support argument gives R_11=R_11^(PP)+O_epsilon(x^(39/40+epsilon)); the returned 79/80 exponent was weaker. Cauchy gives 2B eta x log^2 x(1+o_eta(1)); cofactor triangle gives the weaker 4eta^4 x log^4 x budget. The cutoff-difference identity retains both boundary windows and exact floor-dependent logarithms. **Limits:** no uniform positive clipped fiber length, no signed saving. Input M states a local one-sided O(x) bound on an unbounded dyadic set. Mixed and outside estimates must separately fit a common-scale budget, or be replaced by a joint or admissible averaged estimate. **Evidence:** [validator](transition-signed-estimate-validation.js) and [audit controls](transition-round-audit-validation.js). The returned validator already included the cut omitted in the printed formula. **Reuse:** attempt the weighted cofactor average or joint cutoff correlation with full normalization. Norm-only repetition cannot supply the signed gain. #### S-0905-13 — Controlled a signed region across singleton fibers **Grade:** DERIVED using the classical quantitative Möbius mean estimate; finite algebra and saved-factor tests VERIFIED. **Question:** `Q-signed-divisor-grouping`. For every fixed H, the actual remainder restricted to de≤x^(7/10) is O_H(x/log^H x). Its density term cancels by a uniform excluded-prime Möbius convolution; the summed CRT endpoint errors are O(x^(99/100) log³x). Both signs in this region still have mass at least c*x*log(x) eventually, so this is signed cancellation across fibers. **Evidence:** [derivation and parameter budget](signed-divisor-grouping.md), [independent validator](signed-divisor-validation.js), [saved signed blocks](data-reuse/signed-grouping.json). **Limit:** the complementary product range remains OPEN. No practical asymptotic onset, improved twin count or novelty is claimed. **Related failure:** F-0905-14. #### S-0905-14 — Isolated the remaining signed endpoint sum **Grade:** EXACT CRT and discrete partial-summation identities; DERIVED cancellation of the complementary density term. **Question:** `Q-signed-divisor-grouping`. The remaining region has an explicit inverse-residue fractional-part error E_>, with S(x)=C₂x+E_>(x)+o(x). Reuse its precise coefficients, composite moduli, product cut and endpoint weights when auditing an import. **Evidence:** [endpoint interface](signed-divisor-grouping.md), [exact discrepancy and weighted-summation checks](signed-divisor-validation.js). **Limit:** identifying E_> does not estimate its sign or prove its sufficient one-sided hypothesis. The required scale-x improvement is OPEN. #### F-0905-14 — Extending the unsigned endpoint budget to the remaining range **Verdict:** INSUFFICIENT BOUND, not a refuted signed estimate. **Question:** `Q-signed-divisor-grouping`. After the β expansion, separate absolute CRT errors cost O(L*V*Z*log³x) up to product L. This gives a saving for every fixed product exponent below 71/100, but supplies no scale-x saving when extended without additional cancellation. On the full complementary rectangle the same allowance is O(x²log²x), worse than the existing O(x log⁴x) absolute bound. The complementary density term does cancel; the unestimated part is its signed finite-interval discrepancy. **Evidence:** [error budget and exact remaining sum](signed-divisor-grouping.md), [finite CRT checks](signed-divisor-validation.js). **Reuse:** S-0905-13–14. **Revisit only with:** a signed bound across endpoint phases, a sharper summed discrepancy estimate, or a different decomposition with its own error budget. The exponent 71/100 is not a universal method barrier, and the sufficient one-sided endpoint hypothesis remains OPEN. #### Endpoint-target-audit — A shrinking positive margin is sufficient **Grade:** DERIVED quantitative error accounting and CONDITIONAL twin implications using classical inputs. **Question:** `Q-endpoint-target-audit`. The complete reduction has error O_H(x/log^H x) for every fixed H. Consequently C₂x+E_>(x)≥c*x/log^K x on unbounded dyadic scales, with fixed c,K>0, suffices for infinitely many twins. A logarithmically rescaled average also suffices. Both endpoint hypotheses remain OPEN. The fixed-fraction target remains a stronger sufficient option; its failure would not refute the smaller-margin target. The recorded uniform-gap Kloosterman comparison does not establish an obstruction at these different coefficients, positions and quantifiers. **Evidence:** [rate accounting, conditional derivations and scope check](endpoint-target-audit.md). **Reuse:** retain the smaller margin after removing the additional rectangle in [endpoint-fourier.md](endpoint-fourier.md). **Limit:** the target audit itself adds no cancellation estimate, count, effective onset or independent referee validation. Its Fourier follow-up controls a region, not the full E_>. Recheck a decisive prior claim at its exact scope before using it to exclude an approach. #### Endpoint-fourier — Aggregation makes a further rectangle controllable **Grade:** EXACT coefficient identity; DERIVED regional estimate from Vaaler, Bettin–Chandee and the uniform Möbius mean; finite checks VERIFIED. **Question:** `Q-endpoint-fourier`. Combining small prime-power terms on each expanded divisor yields coefficients bounded by 2log l. For original ranges d~x^delta, e~x^nu, put a=delta+6/25, b=nu+1/20. The complete Fourier budget gives O_H(x/log^H x) for each rectangle with (17/20)(a+b)+(1/4)max(a,b)<1, delta>6/25 and nu>1/20. In particular d in (floor(x^0.27),2floor(x^0.27)] and e in (floor(x^0.46),2floor(x^0.46)] are controlled. Its endpoint error has a power saving and its density cancels at every fixed logarithmic precision. The whole rectangle lies outside de≤x^0.70 eventually. **Unsuccessful steps:** applying the same bound separately to the small prime-power factors has first norm exponent 2279/2000 in this example; aggregation reduces it to 1989/2000. Dropping all nonconstant Vaaler majorant terms is invalid: a retained finite counterexample has actual truncation error 0.859687064 versus the zero-only allowance 0.2. On full expanded supports of order x, the generic first exponent is 39/20, so this estimate still supplies no full residual bound. That upper-bound shortfall is not a lower bound or impossibility theorem. **Evidence:** [derivation, sources and complete budget](endpoint-fourier.md), [independent algebra and exact exponent checks](endpoint-fourier-validation.js), [retained checks and reused prefixes](data-reuse/endpoint-fourier.json). **Limits:** a rectangle, not a uniform product cutoff at x^0.73; no proportion of mass controlled, effective onset, novelty or twin lower bound is claimed. The written theorem application has been checked here; the finite tests do not establish its asymptotic claims. **Reuse or revisit:** preserve aggregation, g=2, the endpoint perturbation and every majorant frequency. The [coefficient follow-up](coefficient-structure.md) supplies a smaller nonsquarefree norm and a squarefree norm lower bound. The [paired endpoint follow-up](endpoint-pairing.md) improves this budget and controls another rectangle. To continue, exploit the remaining arithmetic phase correlation or use a justified one-sided argument. A coupled product cut requires a separate treatment. Both sufficient twin margins in the target audit remain OPEN after removing the controlled rectangle. #### Coefficient-structure — Sparse exceptions and the squarefree core **Grade:** EXACT coefficient classification; DERIVED support bounds, norm lower bound and regional endpoint reduction using classical inputs; finite identities and sector reconstruction VERIFIED. **Question:** `Q-coefficient-structure`. On I=(D,2D], the aggregated contribution splits as L_0(l)log n+L_1(l)+Q(l)+B(l). The low terms lie in I. Q is a squarefree Möbius sign times a nonnegative prime-log sum; B has at most one repeated prime and is explicitly plus or minus its logarithm. The nonsquarefree norm is O(sqrt(D)*W^(1/4)*log^(3/2)(2DW)), allowing its endpoint sectors to be controlled in a larger range. For d~x^0.277, e~x^0.467, all endpoint sectors except QQ have a power saving: their largest first exponent is 19913/20000, whereas the separate-endpoint QQ budget has 20163/20000. The total rectangle density cancels at arbitrary fixed logarithmic precision, leaving R_{I,J}=E_QQ+O_H(x/log^H x). A separate excluded-prime argument also controls the QQ density, so its raw correlation is an equivalent target at that precision. **Failed step and its scope:** a uniform power saving in the individual squarefree norms is impossible. Ordinary PNT and an elementary squarefree count give sum_{DW/2= c*DW*log W for fixed positive power scales D,W. Reinforcing prime contributions supply this lower bound. It does not exclude cancellation in the signed inverse-residue correlation, special ranges, logarithmic gains or one-sided bounds. **Evidence:** [classification and proofs](coefficient-structure.md), [exact vectors, norm witnesses and 16-sector validator](coefficient-structure-validation.js), [retained checks and archived prefixes](data-reuse/coefficient-structure.json). **Limits:** the further rectangle is reduced, not controlled, and the full twin margin remains OPEN. The exact finite QQ endpoint retains its density subtraction; raw prefix QQ sums are not endpoint estimates or full-interval averages. The nonsquarefree terms are not identically zero: both signs have exact witnesses. **Reuse or revisit:** use the prime-factor identity (13) in the note, preserving the prime averages and CRT parity branches. The [paired endpoint budget](endpoint-pairing.md) reduces the relevant deficit to 61/20000 near h~MN/x on the top boxes, and controls their lowest frequencies. That note also controls the entire majorant by an elementary divisor-count argument. A one-sided strategy can use a different sufficient budget. Do not retry uniform squarefree norm power savings without changing their stated scope. #### Endpoint-pairing — Interval cancellation and the transition frequencies **Grade:** EXACT polynomial integral identity; DERIVED regional bound from the existing Vaaler, Bettin-Chandee and Mobius inputs; finite identities and rational budgets VERIFIED. **Question:** `Q-endpoint-pairing`. Keeping the two interval endpoints together saves min(1,Ax/(MN)) in the polynomial sum. The positive error majorant does not have that factor; its nonzero frequencies instead retain their coefficient A/T. The complete first exponent becomes 3/20+(7/10)(a+b)+(1/4)max(a,b), with the second term also controlled. This bounds the full rectangle d~x^0.275, e~x^0.465 to O_H(x/log^H x) for every fixed H. Its endpoint first exponent is 3999/4000<1, whereas the separate-endpoint estimate gives 4017/4000>1. The balanced sufficient boundary is a=b<17/33. An independent elementary bound controls the entire positive majorant by O_epsilon((MN/T)*x^epsilon): group nearby CRT solutions by their integer value and count divisor pairs. Both negative integers and the exceptional values 0,2 are accounted for; the latter contribute no pairs because both divisor indices exceed 2. **Remaining failure:** on the larger squarefree pilot a=b=0.517, the transition h~MN/x~x^0.034 still has first exponent 20061/20000>1. The relevant deficit is 61/20000 before strict losses. The low-frequency polynomial on these top boxes has exponent at most 19993/20000 for h<=x^0.03. Thus focusing on h~1 as their obstruction would repeat an avoidable loss. A direct fixed-q application of Wright v2 Theorem 2.1, followed by triangle summation over q, has exponent 1.01895 even with a free perturbation cost, and does not close this deficit. The advertised improvement in arXiv:2601.00292 does not follow after its authors' erratum (v2, missing factor L^2 in (2.53)) and is not an input. Neither failed import excludes a retained-prime-average estimate. **Rejected shortcut:** applying the interval factor to the whole positive majorant is false at sawtooth jumps. The retained witness has error 0.859687064, full allowance 0.968575681, but incorrectly paired allowance 0.110694364; the note gives a family disproving any universal such factor. Nonzero majorant modes cannot be discarded either. **Evidence:** [derivation and versioned sources](endpoint-pairing.md), [validator](endpoint-pairing-validation.js), [retained checks](data-reuse/endpoint-pairing.json): 35,640 polynomial identities, 757,350 complex harmonic checks, 8,910 exact CRT count checks, both gcd branches, and 79,728 exact rational envelope inequalities. The whole-tail check adds 6,376 Fejer bounds and 2,307 integer divisor-count bounds, including 897 negative integers. The existing 251-check numerical audit and all strict QC/self-test checks also pass. **Limits:** this is a rectangle with product of order x^0.74, not a uniform product cutoff, a mass proportion, an effective onset or a twin lower bound. The full complementary endpoint remains OPEN; the larger QQ pilot is controlled by [prime dispersion](prime-dispersion.md). Finite tests do not prove the asymptotic inputs; no novelty is claimed. **Reuse or revisit:** seek a correlation estimate near each relevant box's transition, retaining p,q averages and the endpoint phase. Consume the divisor-count bound for the full majorant rather than requiring a further oscillatory saving there, or justify a different one-sided consumer. Do not repeat the lowest-frequency or norm-only attempts at unchanged scope. The [prime-band follow-up](prime-band-completion.md) now localizes the prime ranges and treats completion; [prime dispersion](prime-dispersion.md) consumes it to control the larger pilot, retaining right-prime cross terms. #### Prime-band-completion — Localized core and the full dual spectrum **Grade:** DERIVED prime-strip and uniform low-h removals using the paired classical-input budget; EXACT completion, Gram identity and diagonal lemma; finite checks VERIFIED. **Question:** `Q-prime-band-completion`. On the unresolved pilot, aggregation removes p<=x^0.235 or q<=x^0.045 with first exponent 19991/20000<1. An envelope uniform in the expanded divisor product removes h<=x^0.029, with exponents 19991/20000 and 193/200. The residual prime bands are (x^0.235,x^0.24] and (x^0.045,x^0.05], and the positive harmonics satisfy x^0.029h_i>0, equality h_1/q_1=h_2/q_2 forces identical pairs. This identifies only that diagonal, not every possible coincidence in a further dispersion argument. **Evidence:** [derivation and source audit](prime-band-completion.md), [validator](prime-band-completion-validation.js), [retained artifact](data-reuse/prime-band-completion.json). Checks include 8,192 divisor/prime comparisons, 8,468 exact phase splits, 28 completion identities with the endpoint perturbation, 462 Gram entries and 36 norm-attaining checks. There are 1,134 exact rational envelope checks and 17,424 equal-fraction comparisons. Both gcd branches are included. All fourteen strict QC gates, the verifier self-tests and the existing 251-check numerical audit pass; these do not prove the asymptotic inputs. **Data reuse and limits:** all three retained archived prefixes have an empty high-band core because at least one rounded prime band is empty. This is a finite cutoff effect, not evidence of an asymptotic rate. The toy high-band endpoint is -9.232843649 and its density is 5.326894578; neither is silently set to zero. No twin lower bound, effective onset, mass proportion or novelty is claimed. **Reuse or revisit:** consume the completed dispersion argument below, which changes the placement of Cauchy while retaining the actual phases. The full-spectrum closure and coefficient-separation caveat here remain valid. A larger run is justified only by a specific falsifier and sufficient input fields. #### Prime-dispersion — A second moment controls the full pilot **Grade:** DERIVED bound using classical Ramanujan and Weil inputs; EXACT regrouping and second moment; finite checks VERIFIED. **Question:** `Q-prime-dispersion`. The full d~x^0.277, e~x^0.467 rectangle now satisfies R_IJ=O_H(x/log^H x) for every fixed H. This consumes the established coefficient-sector, density, prime-strip, uniform low-h and entire-majorant bounds; it does not replace those arguments with a numerical observation. The decisive choice is to aggregate m=dp before Cauchy, keeping a logarithmically bounded coefficient on an interval of length M. For each e, retain both right-prime and harmonic averages in the second moment. Its only zero-phase pairs are identical (q,h). Same-prime different-h pairs have modulus eq and gcd loss O(A); distinct primes have modulus e*q1*q2 and gcd loss O(AZ). Interval completion keeps the Ramanujan full-period term, and partial summation prices the true endpoint weight. The diagonal's final exponent is 1989/2000=0.9945; the largest off-diagonal exponent is 1173/1250=0.9384, before an arbitrarily small loss. The remaining high-h polynomial is O(x^0.995). **Unsuccessful placement and scope control:** fixing p before Cauchy leaves a much shorter d interval. The present uniform Weil bound then loses to the trivial cross kernel, returning exponent 1.034 after the outer sums. This is an insufficient bound, not an impossibility result. Dropping the e factor from the cross-prime modulus is also invalid in general: e=5, q1=11, q2=13, h1=h2=1, theta=1 yields modulus 715, and the phases at m=1,144 differ despite equality modulo 143. **Evidence:** [derivation and classical inputs](prime-dispersion.md), [validator](prime-dispersion-validation.js), [retained artifact](data-reuse/prime-dispersion.json). Checks include 6,144 coefficient identities, 1,024 parity zeros, eight original-versus- grouped endpoint polynomials, 54 endpoint-weighted second moments, 606,096 exact modulus comparisons, 90 interval completions (30 longer than their moduli), and 180 rational exponent comparisons. Both gcd branches and all pair classes occur. The validator's first run caught a same-prime exponent addition error: the corrected row is 0.8884; the leading diagonal and cross-prime budgets were unaffected. All fourteen strict QC gates, the verifier self-tests (58 positives and 47 controls), and the existing numerical audit (251/251) also pass. **Limits:** the controlled rectangle has product scale x^0.744, compared with the previous x^0.74 rectangle. These are rectangle statements, not a uniform product cutoff or measured proportion of the residual. The global endpoint remainder and sufficient twin margin remain OPEN. The retained prime table supports algebra checks; the archived high-band cores are still empty at their finite cutoffs. No effective onset or asymptotic rate is measured, and no novelty is claimed. **Reuse or revisit:** the [range follow-up](dispersion-range.md) supplies a different orientation and averages the gcd loss, controlling a larger region. Consume its explicit constraints along with the pilot here. Do not repeat the norm-only attempt or the full-frequency operator closure at unchanged scope. #### Dispersion-range — Averaged gcd losses and a region of rectangles **Grade:** DERIVED regional bound with classical inputs, EXACT reciprocal and second-moment identities; finite checks VERIFIED. **Question:** `Q-dispersion-range`. Reversing inverse reciprocity shifts both interval endpoints by 2 and moves the larger left distinguished-prime average into the second moment. Cauchy in the right expanded divisor and then in the original left divisor allows the bound sum_(d~D) gcd(r,d)^alpha << D*tau(|r|), for r!=0 and 02h. Complete periods and the whole positive Vaaler majorant remain priced separately. This derives the uniform bound in the owning note (12). **Range:** combining this estimate with the existing left-prime dispersion and every other coefficient sector controls d~x^0.275, e~x^0.541, a full rectangle at product scale x^0.816. The leading BB and left-prime cross budgets are 0.99945 and 0.99925. The product-exponent supremum of the stated sufficient inequalities is 1368/1675, about 0.816716, approached at a=1727/3350, b=3961/6700. BB and the left-prime cross moment both become tight there. This is not a uniform product cutoff or a bound on the full remainder. The actual signed cross moment is written before the final kernel triangle inequality; an upper saving x^(-gamma) in its moment reduces the block exponent by gamma/2. This stronger estimate remains OPEN, and alone would not remove the BB constraint. **Failures and controls:** the preceding paired sparse-norm bound gives 1.01195 on the new rectangle, insufficient for a saving. Reusing q>2h would discard needed harmonics, since T can reach exponent 0.10602 while q<=x^0.05. Explicit witnesses refute the resulting false zero-pair exclusion and the omitted distinguished-prime gcd factor. The preceding notes' prime cuts validate their narrower use of those assertions. A finite-difference check initially failed at large phase because its second-order truncation error exceeded the comparison tolerance; a fourth-order Richardson comparison checks the same analytic derivative. This was a numerical diagnostic issue, not a change to the derivative or the claimed variation bound. **Evidence:** [derivation, full region and signed target](sparse-dispersion.md), [validator](sparse-dispersion-validation.js), [retained artifact](data-reuse/sparse-dispersion.json). The checks use the saved prime table, actual B coefficients, both gcd branches, harmonics below and above the primes, exact moment partitions and rational budgets. They measure no asymptotic cancellation rate. The completed run includes 8,722 B identities, 1,388,040 gcd factorizations, 24 exact moments, 1,251,432 pair phases and 22 paired-sector budgets. All fourteen strict QC gates have zero findings, the verifier self-tests pass (58 positives, 47 controls), and the independent existing numerical audit passes 251/251 checks. **Reuse or revisit:** the opposite BQ orientation and frequency matching are completed at this scope. [Prime-power dispersion](prime-power-dispersion.md) now splits the left B coefficient and prices its entire odd-power core. It also uses the paired, all-harmonic first-power moment to remove the older prime cut. Consume that follow-up's region and limits before proposing another BB calculation. Do not pair the positive approximation tail or apply a global norm discount pointwise on a smaller box. #### Prime-power-dispersion — Exceptional coefficient and all-harmonic moments **Grade:** DERIVED regional bound with classical inputs; exact coefficient, modular and moment identities, with separate finite validation. **Question:** `Q-prime-power-dispersion`. The full remainder and sufficient twin margin remain OPEN. The exact left B coefficient splits into odd prime-power sums and an exceptional part with norm O(sqrt(D) log^2(2DV)), retaining support O(DV). For every fixed power k inside the moment, power bands R<=p^k<2R make the constants uniform as k grows. The common-divisor gcd average and harmonic averages of the full prime-power gcd handle all nonzero numerators; equal rational frequencies supply an explicit additional zero class. Paired endpoint weights control that entire zero class, with the complete-period term and the full positive majorant retained. The k=1 case consequently needs no small-prime/harmonic cut. Both reciprocal endpoint shifts and gcd branches remain explicit, and all right coefficient sectors are allowed. **Range:** C(a,b)<28/25 and W_L(a,b)<1 suffice for the full rectangle bound O_H(x/log^H x), with the original cutoffs and a=delta+6/25, b=nu+1/20. The example delta=0.247, nu=0.612 has product scale x^0.859 and leading budgets 0.9998 and 0.99975. The exact supremum of product exponents in these inequalities is 5757/6700, about 0.859254, approached at a=816/1675, b=1109/1675. The remaining tight terms are the exceptional coefficient's paired bound and the signed first-power cross moment. This is not a uniform product cutoff or full remainder bound. **Failed shortcuts:** at the new rectangle the preceding BB norm budget is 1.0473 and the unpaired diagonal budget is 1.029; those estimates are replaced, not assumed small. Replacing gcd(h,p^k) by gcd(h,p) fails already at h=25,p=5,k=2. Base-prime dyadic bands can widen by 2^k and are not uniform for growing k; power bands resolve that issue. The validator also demonstrates failures of ordinary Number residue arithmetic above the safe integer range. Its actual kernels use BigInt. The old 1.01195 budget in the preceding entry is correctly described as paired; correcting that wording changes no numerical value or bound. **Evidence:** [derivation and full region](prime-power-dispersion.md), [validator](prime-power-dispersion-validation.js), [retained artifact](data-reuse/prime-power-dispersion.json). The completed 3.3-second run checks 1,883,736 coefficient identities, 1,152 harmonic gcd/counting bounds, 80 weighted moments, 1,434,208 pair phases and 197,736 shifted endpoint identities. It includes powers through eight, 10,612 distinct-prime zero terms, 217,616 nonzero prime-power gcd terms and every right coefficient sector. There are 56,368 kernels above the safe integer range; the deliberately unsafe arithmetic disagrees with 55,284 exact residues. These are finite algebra tests, not measured asymptotic rates. Harmonic subsets in the large-power examples are declared; the all-harmonic conclusion comes from the written proof. All fourteen strict QC gates have zero findings, verifier self-tests pass (58 positives, 47 controls), and the independent numerical audit passes 251/251 checks in 165.2 seconds. Together with the new validator this is under three minutes of numerical runs, within the authorized budget. **Reuse or revisit:** consume the uniform power-band moment and the two-inequality region. The [coverage follow-up](residual-coverage.md) now handles the first branch inside a shorter-base moment, includes the prime-free terms, and separates exact coupled cuts. Its full region no longer has the exceptional norm condition. Use that audit's remaining sum and edge-specific budgets before another local optimization. #### Residual-coverage — Exact global complement and first-branch dispersion **Grade:** DERIVED bounds from the existing classical completion inputs; VERIFIED finite twisted identities, moment algebra and exact masks. **Question:** `Q-residual-coverage`. The sufficient twin margin and full endpoint remainder remain OPEN. The first branch B(p^k m), with pm in the original interval, has base length D/P on a power band p^k~R, P=R^(1/k). Retaining that shorter interval bounds its cross term by F*D^(5/4)*W^(3/4), up to the stated losses, while preserving zero and complete-period terms. A prime-free moment controls the low coefficients, and a separate singleton calculation retains the actual p=2,g=2 reduced power. Both orientations consequently cover the whole coefficient without an exceptional norm constraint. **Range and complement:** full boxes are controlled when delta+nu<19/25 or 5delta+2nu<123/50. Independent divisor twists have uniform endpoint bounds; an explicit truncated Perron argument therefore separates de and d^5*e^2 cuts, including their intersection. The untwisted density is bounded directly on its original intervals before separation. Every fixed de<=x^theta with theta<19/25, and d^5*e^2<=x^lambda with lambda<123/50, has arbitrary fixed logarithmic savings, as does their union. This is a uniform product improvement over 7/10 (and the elementary allowance theta<71/100). The regional product supremum 87/100 is a different claim. The concrete remaining E_* in (16) has de>floor(x^(3/4)) and d^5*e^2>floor(x^(49/20)); all margin strips and endpoint conventions are retained. S=C2*x+E_*+O_H(x/log^H x) preserves the open one-sided and rescaled-average twin consumers. **Insufficient bound and next target:** at d~e~x^(2/5), the right cross/zero/period budgets are 179/200, 51/50, 21/25. Improving the cross term alone does not fix its zero-frequency deficit. The extra Cauchy inequality in the original divisor loses its Mobius signs. Equations (19)--(20) retain those signs and give a specified first-power sector target: a signed upper bound with moment exponent below 34/25, versus the current 7/5, requiring more than a 1/25 saving. The other sectors and the full complementary domain remain to be handled. Neither this failed estimate nor the nonnegative zero class excludes another proof. **Evidence:** [derivation and exact consumer](residual-coverage.md), [validator](residual-coverage-validation.js), [retained input checks](data-reuse/residual-coverage.json). The finite decomposition covers 146,016 twisted coefficient identities and 1,710 power-of-two reductions. Exact reciprocity checks reject the omitted-shift control. The moment tests include twists, both orientations, both gcd branches, the low terms and larger prime powers. A rational grid tests 117,711 budget points. The archived factor data supply 273,177 weighted divisor pairs for exact inclusion-exclusion, with nonzero overlap as the double-counting control and BigInt monomial masks. None of these finite checks measures an asymptotic cancellation exponent. The final validator runs in 1.4 seconds and includes forty moment checks with 535,900 expanded pairs. All fourteen strict QC gates have zero findings; verifier self-tests pass all 58 positives and 47 controls. The independent existing numerical audit passes 251/251 checks in 164.7 seconds. Numerical verification remains under three minutes. **Reuse or revisit:** the coverage audit, exceptional first branch, prime-free moment and cut separation are completed at their stated scope. The [literature audit](structural-literature-audit.md) derives the cross-divisor kernels, lcm moduli, gcd factors and rational-frequency coincidences before the second Cauchy step. Their improved analytic estimate remains open; use the audit's full-moment test next. The right zero term limits the uniform-product edge; the left cross term limits the other edge. Choose the next estimate by its effect on the explicit complement, not by maximizing a regional product exponent. #### Structural-literature — Theorem interfaces and the exact cross-divisor kernel **Grade:** SOURCE-CHECKED theorem interfaces; DERIVED kernel identities, collision count and conditional parameter comparisons; VERIFIED finite algebra. **Question:** `Q-structural-literature-audit`. No reviewed modern theorem directly closes E_*. The [grouped-divisor follow-up](grouped-divisor-moment.md) derives the proposed full moment from classical completion and controls an additional region. The global twin margin remains OPEN. **Reusable result:** with u_i=j*ell_i, j=gcd(u1,u2), the paired inverse phase has modulus c=j*ell1*ell2 and numerator -theta*(h1*ell2-h2*ell1). Its zero class is exactly h_i=t*ell_i, requiring j>N/(2A) on u_i~N,h_i~A. The number of ordered collisions is at most 8*N*A*H_floor(2*min(N,A)). On the live squarefree first-power sector the original sign product is mu(ell1)*mu(ell2); squarefreeness and other common-factor weights remain. The nonzero gcd inequality in (6) includes repeated prime powers. Equation (8) gives the exact endpoint-weighted completion and exposes coupled coefficients. **Useful matches and insufficient imports:** Pascadi's dispersion-coefficient spectral theorem has the same linear frequency difference, but at the natural diagnostic level q~L², L=N/J, its extra factor is trivial when J<=x^0.27 at the benchmark. This does not exclude another spectral factorization. Blomer--Pascadi's July 2026 composite-modulus bilinear estimate and Pascadi's factorization-sensitive estimate require a short separated bilinear form that has not been constructed. A headline saving cannot be multiplied into the present moment. Trace-function estimates require prime moduli and kernel hypotheses; the reviewed Mobius estimate has only logarithmic savings. Triple-convolution dispersion subtracts small-conductor characters, whose contribution must be handled separately. Well-factorable progression weights, divisibility-graph expansion, quantitative Chowla correlations and the asymptotic sieve each have explicit transfer or input gaps recorded in the note. These are failed direct imports, not refutations of those methods or of the twin target. **Evidence:** [source-linked audit and derivations](structural-literature-audit.md), [validator](structural-literature-validation.js), [retained finite checks](structural-literature-validation.json). The validator independently checks modular phases, original-divisor sign products, collision parametrization, gcd bounds, endpoint-weighted Fourier completion and rational budgets. Incorrect modulus, missing theta and unrestricted squarefree-sign controls are detected. These tests do not prove a saving or validate the papers' proofs. **Next and revisit conditions:** the full grouped-moment test is now completed in the next entry. It includes the coefficient sectors, harmonic gcd averages, periods, endpoint variation and twists, and removes the old 1.02 loss at the benchmark without assuming Mobius cancellation. Match a complete spectral formula or a separated bilinear piece for the follow-up's remaining small-common-divisor kernel. Centered dispersion and graph transfers remain alternatives with concrete obligations. Do not repeat the unmatched theorem comparisons or the completed moment. #### Grouped-divisor-moment — Full moment, exact additional cut and next kernel **Grade:** DERIVED from classical completion and elementary gcd averages; VERIFIED finite identities and rational bookkeeping. **Question:** `Q-grouped-divisor-moment`. The full endpoint margin and twin-prime infinitude remain OPEN. **Established result:** the full aggregate coefficient, with arbitrary complex divisor twists of nonnegative real part, stays inside one Cauchy moment. Equal rational frequencies contribute at most B²*x^epsilon*f²*MN/A. Nonzero kernels with common divisor j~J have Weil budget N³/J^(3/2) and complete-period budget M, up to small losses. The general harmonic gcd average includes repeated prime powers. Endpoint variation costs f²*(1+Ax/(MN)). This proves the proposed full moment B²*x^epsilon*[x+x^(3*tau)*(N³+M)] on retained harmonic bands. No cancellation estimate for Mobius signs is assumed in this moment. **Coverage:** every coefficient sector is included by the bounded full convolution coefficient, in both gcd branches. First Cauchy gives right block budgets (1+a)/2,a/2+3b/2,a. The additional full region is delta<19/25 and delta+3nu<161/100. The entire d~e~x^(2/5) benchmark has budgets 41/50,199/200,16/25 before small losses, and is controlled to arbitrary fixed logarithmic precision. The full positive majorant and the untwisted density are consumed separately. Uniform divisor twists justify the exact added cut d<=floor(x^(151/200)), de³<=floor(x^(321/200)), and all intersections with the previous cuts. The current E_dagger has the domain in (19). E_*-E_dagger=O_H(x/log^H x), preserving the sufficient twin consumers. **Limits and failed inferences:** the vertical cutoff cannot be dropped from this application; a reaching one leaves the zero budget at one. The combined uniform product threshold remains every fixed exponent below 19/25, with an explicit unhandled boundary witness. A smaller summation domain does not imply monotonicity of its signed value. At delta=8/25,nu=9/20 the grouped Weil budget is 103/100; the old left cross and right zero budgets also exceed one. Neither estimate controls this nearby box. These are failures of stated upper bounds, not lower bounds on the actual correlation or a universal obstruction. **Evidence:** [proof and exact consumer](grouped-divisor-moment.md), [validator](grouped-divisor-validation.js), [retained finite checks](grouped-divisor-validation.json). The validation checks 16,384 general harmonic gcd averages and 14,894,880 ordered completion-proxy pairs in 360 common-divisor bands. It reconstructs 96 full-coefficient moments, including 104 repeated-prime coefficient entries, and their 16,864 pair kernels. A rational grid checks 117,711 parameter points. The existing archived factors supply 273,177 signed divisor pairs for exact cut partitioning; 28,318 are in the additional cut only. These counts are finite validation, not estimates for the proportion or asymptotic mass of the residual. Missing periods, conjugation, endpoint shift, vertical cutoff and overlap corrections are detected by separate controls. **Next target:** the restricted bound controls all nonzero pairs with j>x^(1/20) at the next box, with block budget 397/400. Its zero and period terms are already controlled. Equation (21) retains the real signed sum over j<=x^(1/20), R!=0, and asks for a one-sided upper bound with exponent below 36/25, versus the generic 3/2: more than 3/50 must be saved. Complete the actual Fourier-weight separation and price its coefficients before using a modern bilinear or spectral input. The large-common-divisor ranges are not the remaining bottleneck. A solution at this box alone would still not establish the global twin margin. #### Structured dispersion estimate — Common prime-power factor retained through Cauchy **Grade:** DERIVED and independently reviewed block and regional bounds; finite inputs VERIFIED. **Question:** Q-structured-dispersion-estimate. **Established:** for b_u=sum_(q|u)beta(u/q)lambda(q), prime-power q~Q, |beta|<=1, Cauchy over (m,q) gives (D1), with a factor Q^(-1/4)+A^(-1/4) on the cross term. Lemma H prices the harmonic gcd average. The 2026-09-09 review rechecks completion (7), its Pascadi inputs, the necessary separate coefficients, gcd branches and the Perron/density application. The target-box exponent falls from 41/40 to 407/400, still above one. **Region:** delta<71/100, delta+3nu<327/200. The concrete cut C'={d<=floor(x^(141/200)), de^3<=floor(x^(1631/1000)), de>floor(x^(77/100))} is removed in addition to the three previous cuts. It is a fourth simultaneous complement condition, not a substitution. The resulting exact residual differs from the previous residual by O_H(x/log^H x). The uniform product threshold remains 19/25. **Corrected target:** at the top sector Q^(3/2)E^3 has exponent 57/40; MQ has exponent 61/100. A block saving requires the moment below 139/100, hence saving greater than 7/200. The reported 59/40 and 17/200 were arithmetic errors. The j_e^(-3/2) factor pays the gap for j_e>x^(7/300+epsilon), fixed epsilon>0; the remaining small-gcd range is OPEN. Native u-pair and fixed-q e-pair targets are distinct sums. **Scoped closure, return #714 (2026-09-16).** The 7/400 block deficit at this box is not reachable from the objects (D1) is built from. Four probes are closed and two named entries replace the earlier open-ended condition. (i) The four endpoint classes of the pair weight collapse exactly to 2e(varsigma R/(cm))[cos(delta R/(cm)) - cos(delta S/(cm))], so the inflation factor f = min(1,v0) is the size of the *difference* of two endpoint weights: the class split is free at the binding band (v0 = 1) and costs v0^(-2) away from it, and the deficit is class-uniform, not carried by the mixed classes. (ii) The variation / l-infinity to Parseval / l^2 substitution inside (6)+(7) is annulled on the same line: (7)'s t != 0 branch uses the weight's variation and the Dirichlet l^1, not a t-sum of the weight, so there is no insertion point, and the substitution moves the block exponent from 407/400 to 519/400. (iii) A G-relative improvement of the Weil input is inert, because the gcd averages (8)-(9) are flat in the exponent, and a cap G <= x^gamma is worth at most 1/400, seven times short. (iv) The improvement is needed on the whole dual range, not on a bounded zone of t (the arithmetic input is uniform in t while the weight's cost is flat per octave, so every octave carries the same power; the exemption allowance is x^(-7/400)), and the (t,r,c)-local structure supplies only G*x^o(1): the degenerate p-adic scaling reproduces the corpus's sqrt(cG) input, the coset vanishing and the quadratic support give a density G^(-1), and the mean square sum_t S(t,r;c)^2 = c*phi(c) is r-blind. Two named entries: **R1** an absolute, uniform-in-t improvement |S(t,r;c)| << sqrt(cG)*c^(-eta) with eta >= 7/190, for the sup over t and on the whole t-range; **R2** the dual form of Corollary 8.1 of arXiv:2511.08445v2 with a coefficient allowed to depend on both indices (pair-rank <= 2), whose payoff in the coprime e-pair class is 19/80 with the legal windowed norm. Reopen with R1 or R2, not with the four closed probes above. **The two route-level refutations, and why one supersedes the other (added by the audit of 2026-09-17).** D was attacked at the level of the obligation along two routes, and the second supersedes the first, so the operative obstruction is single. **T5** asked whether Theorem 1.2 of arXiv:2511.08445v2 absorbs the completed (D1) moment and answered **NO for the two-sequence class**; **T791** asked the same of Theorem 7.1 / Corollary 7.9 and answered **INAPPLICABLE**, the failing hypothesis being the product-shaped coefficient (`alpha_t = w_c(t)` still depends on `R`, and the kernel admits no weight depending on both indices). T791's read of the v2 LaTeX source **withdrew three of the dependencies T5 had named as obstructions**: the completion's `t`-sum *is* the mechanism (`sum_t S(t,r;c) e_c(-tm) = c*1_{(m,c)=1} e_c(r mbar)`, so `alpha_t` is pair-free), `(m,c)=1` is produced by that identity, and the third index from `c = j_e*l1*l2` is absorbed by grouping the pair sum by `(u1,u2)`. **One obstruction survives, and it is T791's.** The count is therefore **two route-level refutations with a single operative obstruction**, plus the refuted named entry **R1** and the five refuted sub-claims (the four probes above and the class split): eight refutations in the record, two at the level of the obligation. Recorded with T5 as well: squarefree quotient plus balanced divisor is *not* a hypothesis of Theorem 1.2 -- it is the sufficient condition of Example 1.3 that produces the `c^(-1/12)` economy. **Evidence for the count:** return #714, file `job1506-T5-verdict.md` (sha256 dfda7c9116b62b4c3935b704668edc7c92c7d39152bb770aeac61c4cc9a886a5) and file `job1506-T791-verdict.md` (sha256 e68d8a80bb83455f49fb392ed80525948fe89a3c2ed1d3239323f29b51c20dd6); register audit return #715 (this revision's base) and return #750 below the two-route reading. **Evidence:** return #714 and job1506-D1-obligation-dossier.md (sha256 fa94c5edb7442a58d9ceee7bc87e16562936429e07c76577f50733b61e683e87; sections 3, 5 and 7), with the served instruments listed there: job1506-class-split-check.py, job1506-dual-len-check.py, job1506-gain-threshold.py and the three job1506-kloost-*.py checks. **Evidence:** [note](structured-dispersion-estimate.md), [lane validator](structured-dispersion-estimate-validation.js), [review](research-round-validation.md) section 10, [exact exponent regression](research-round-validation.js). **Limits/revisit:** upper block bound only, no sign of E_dagger. A smaller residual domain is not a monotone improvement of its value. Target-box success would still leave the corner and global complement. Reopen with a mechanism paying the small-gcd saving and all costs; no claim that the unused Mobius signs are necessary for every possible proof. #### Small-divisor-kernel — Reciprocity separation and priced Kloosterman-fraction interfaces **Grade:** DERIVED separation and exact rational pricings; VERIFIED finite identities. **Question:** `Q-small-divisor-kernel`. Target (21) of the grouped moment and twin-prime infinitude remain OPEN. **Established result:** at delta=8/25, nu=9/20 the j<=x^(1/20) kernel of [grouped-divisor-moment (21)](grouped-divisor-moment.md) separates exactly. Reciprocity splits the completed phase index by index, e_c(theta R mbar) = e_m(theta h1 u1bar - theta h2 u2bar) times a correction phase whose sup norm and total m-variation are O(A/(MN))=O(1/x) on this box; removing it costs O(x^(14/25)) absolutely. Mellin separation of the endpoint factor costs x^epsilon with unimodular twists and L1 mass min(1,Ax/(MN)), free exactly because v=Ax/(MN)<=1 on every retained band here. The block reduces to a pure trilinear Kloosterman fraction T = sum alpha_m beta_u nu_h 1_{(m,u)=1} e(sigma theta h mbar/u) with alpha_m=A_left(gm), beta_u=A_right(gu), |nu_h|<=C/A. The small-common- divisor kernel is therefore the pair expansion of the left orientation with modulus m; reciprocity interchanges the two endpoint conventions. Both gcd branches, both endpoints, all coefficient sectors and the divisor twists are carried. **Limits and failed steps:** Bettin-Chandee (arXiv:1502.00769v1, Theorem 1 with Remark 1; every hypothesis met) gives block exponent 129/125 at this box, above the required 1 by 4/125 and above the classical completion budget 103/100; its second bracket term alone is 399/400, so the first term (AMN)^(7/20)(M+N)^(1/4) is the whole failure. DFI (1.1) with the band summed trivially gives 1267/1200. The spectral diagnostic factor of the literature audit is trivial for all common divisors up to x^(19/50), so the structured spectral advantage is absent where the deficit sits. No separated short bilinear subexpression exists for the composite-modulus route; the only pairing is the full-length dual variable against a modulus that is itself a summation variable. On a 19005-point rational grid every box Bettin-Chandee controls is already controlled. The band majorant N^3/J^(3/2) puts 64.6 percent of its mass at j in [1,2): the whole 3/50 deficit is present at common divisor j=1, so this is a coprime-pair problem, not a small-common-divisor problem. These are failures of the stated upper bounds and say nothing about the size or sign of the actual correlation. Dong-Robles-Zeindler arXiv:2601.00292 stands at v2 with an author erratum (missing factor L^2 in its (2.53)); its claimed improvement does not follow and is not imported. **Evidence:** [proof and pricings](small-divisor-kernel.md), [validator](small-divisor-kernel-validation.js), [artifact](small-divisor-kernel-validation.json). Negative controls reverse the reciprocity sign, replace the lcm by the product, drop theta, and replace the region by one of its three conditions (722 containment violations detected). **Reuse or revisit condition:** any future bound on T plugs into (7) of the owning note with the thresholds it must beat: the (M+N) exponent below 27/140 or the (AMN) exponent below 9/28, holding the other factor fixed. Revisit only with a corrected reading of Bettin-Chandee, a Kuznetsov formulation with modulus-independent coefficients, a modulus factorization exposing a genuinely short separated bilinear form, or an argument that does not pay the first Cauchy in the left divisor (see the reachability record below). #### Left-divisor-signs — Type I/II decomposition of the left coefficient instead of the first Cauchy **Grade:** DERIVED lemmas and exact rational pricings, reviewed once: three conservative slips found and repaired (corner per-block requirement eta'>1 not 2; Lemma II's hypothesis B_219/100, violated on 4790 of the 9943 counted grid boxes, conservative for the zero-added region; one boundary lattice point at sigma=3/100); the 41/40 headline and the zero added region survive. VERIFIED finite identities (82 exact rational assertions, 6 negative controls). **Question:** `Q-left-divisor-signs`. Target (21), the corner and twin-prime infinitude remain OPEN. **Established result:** the left coefficient A_left(gm) of the block has three sectors: A_0 and the -mu log term supported inside the original divisor interval, and A_1's prime-power term, which is exactly the convolution (mu (.)^(-s) 1_I) * (Lambda 1_[2,W]). Parametrising the left prime power by r~x^rho and the right one by q~x^sigma gives a=delta+rho, b=nu+sigma sector by sector rather than the uniform top values. Lemma I (Type I): a piece of mu(m) carrying a long smooth variable has block budgets alpha_1+3b/2 and a with no zero-frequency term, because no second moment is formed; it replaces sqrt(M) by the length of the arbitrary block, decisive only when that block is below x^(1/4). Lemma II (Type II): Cauchy in the ORIGINAL divisor d rather than the expanded m=dr, uniform in the Perron twist heights because the twist rides on the variable removed by Cauchy. At (8/25,9/20) the grouped bound already controls every sector with rho+3sigma<33/100, so the A_0 and -mu log sectors are controlled; Lemma II controls rho+5sigma/4<111/400; together every sector with right prime power q<=x^(3/100) is controlled, the survivor at sigma=1/20 is 43/200<=rho<=6/25, and the worst block exponent there is 41/40 against the previous 103/100. Deficit 3/100 becomes 1/40 at this box. **Limits and failed steps:** the balanced Heath-Brown piece (two variables at x^(7/25)) is covered by neither lemma and returns to Cauchy in m at 103/100; the deficit sits there. Lemma I is not uniform in the Perron twist heights (its smooth variable carries b^(-s) with total variation of order |Im s|), so a Type I route must redo the cut separation at height x^(o(1)). At the corner a=b=1 every priced route exceeds 1: grouped 1/2/1, Lemma I >=3/2, Lemma II 15/8, Bettin-Chandee and Wright 15/8, and even a hypothetical per-block square-root cancellation in m with trivial summation over u gives 3/2, because the corner needs a per-block saving u^(-eta') with eta'>(a+b-1)/b, which is eta'>1 at a=b=1, while square-root cancellation at M=u is eta'=1/2 (twice the ceiling; the note's first version said eta'>2, corrected on review). Cauchy on the right divisor instead gives 109/100 at the target box and the identical 1/2/1 at the corner; the U=V versus Y=Z asymmetry is invisible at a=b=1. Fouvry-Kowalski- Michel Theorem 1.7 (arXiv:1211.6043v3) needs a prime modulus and would give at most 1/48 against a required 3/25; Wu-Xi arXiv:1603.07060v5 needs squarefree friable moduli and short sums; Wright arXiv:2608.27732v1 Theorem 2.1 reduces to Bettin-Chandee at full dyadic supports because its subdyadic hypothesis is on the two inverted variables, not on the harmonic band, and manufacturing subdyadicity costs 1157/1000; Guria arXiv:2410.10856v2 needs two variables of weight exactly 1 to Poisson-sum and a genuine prime, neither present. All are failed upper bounds, not refutations of the target. **Evidence:** [note](left-divisor-signs.md), [validator](left-divisor-signs-validation.js). On a 36481-box grid Lemma II controls 9943 boxes, none outside the region already controlled. **Reuse or revisit condition:** reuse Lemma II wherever a Cauchy in the original divisor is admissible and twist uniformity is needed. Revisit the balanced piece only with an estimate for a bilinear form in two variables of length x^(7/25) against the Kloosterman-fraction phase; the corner needs an argument that is not a per-block bound in m. #### Reachability-coverage — Fixed-margin ceiling and hypothetical kernel savings **Grade:** DERIVED rational bookkeeping at a prescribed fixed margin; shrinking-margin extension UNPROVED. **Question:** `Q-reachability-coverage`. The sufficient twin margin and twin-prime infinitude remain OPEN. **Established result:** for p=19/25-delta and q=19/20-nu, the right budgets are 1-p/2, 2-p/2-3q/2-gamma/2 and 1-p; the left swaps p,q. At prescribed margin eta0 the zero budget leaves S_0={p<2eta0,q<2eta0} out of reach of that margin for every nonzero-kernel saving. Uniform gamma=2 covers its complement at that margin. This is far stronger than the one-box gamma>3/50 target. S_0 lies inside W_dagger for eta0<1/400. The a=b=1 endpoint keeps zero budget one for every margin; fixed interior points can become controllable with a smaller margin. No general impossibility theorem follows. **Correction from independent review:** substituting eta0=C loglog x/log x is not justified with the present x^epsilon losses and uncontrolled epsilon-dependent constants. The corresponding O(loglog x) box counts are geometry, not proved analytic coverage. The original fixed-margin cuts remain intact. Proper prime powers are still negligible after replacing the incorrect lower bound on their base by sum_{p^j>W,j>=2}p^(-j)<1 and non-squarefree branches. Its sharp mass constants were not rechecked in this independent pass. **Evidence:** [owning note](reachability-coverage.md), [finite budget validator](reachability-validation.js), [artifact](reachability-validation.json), [independent review F3 and F5](history/reviews-0906/20-independent-handoff-review.md). **Reuse:** distinguish prescribed fixed-margin failure from failure at every positive margin. Prove uniform losses before shrinking eta with x. The first-Cauchy alternatives have already been attempted in signed-moment.md; Lemma A's unequal R=0 pairs remain uncontrolled. #### Determinant-corollary — Bettin-Chandee Corollary 1 applied directly to dk-et=2 **Grade:** DERIVED transfer and exact rational pricing, negative, reviewed once with no defect (Corollary 1 confirmed verbatim at the source; all vertices and areas reproduced by an independent clipper; the DFI containment needs the line (1/40)S-(1/48)max>=0, now added); VERIFIED finite checks (28 checks, 4 negative controls). **Question:** `Q-determinant-corollary`. E_dagger and the twin margin remain OPEN. **Established result:** the residual R(x) equals T_11-T_10-T_01+T_00 exactly, each a sum over m_1 n_2 - m_2 n_1 = 2 with two arbitrary coefficients and two smooth weights, after deleting the vacuous cutoffs k>V, t>Z (beta_W vanishes below W) and writing beta_W(k) = log k - sum_{r|k, r<=W} Lambda(r) with the bijection (d,r) -> dr of endpoint-fourier §2. The literature scout's recorded blocker, smoothness of the cofactor weights, is therefore removable. The parity branches g in {1,2} never arise: the corollary's main term over (n_1,n_2) | 2 is exactly the compatibility condition, and its density factor and modulus coincide with the subtracted M_> of signed-divisor-grouping §5. The sharp interval dk in J_x is separated by Perron at height x^kappa with eta of order x^kappa; R=O(1) because m_1 n_2 = dk is of order x. Every hypothesis of Corollary 1 (read at the primary source, arXiv:1502.00769v1, statement p. 4, proof §9) is met. **Limits and failed step:** the error term costs x^((17/20)(A+B) + max(A,B)/4 + 3kappa/2) with A=delta+rho, B=nu+zeta over the bands rho<=6/25, zeta<=1/20, so the all-pieces region is 22 max(A,B) + 17 min(A,B) < 20, forcing A+B<40/39 and hence delta+nu < 2869/3900 at the top band, strictly inside the region already controlled. Added area: exactly zero. The log-log piece alone would add 12.97 percent of the domain, but the region for R is governed by the worst piece. At (8/25,9/20) the top band fails by 41/1000; at the corner by 39/20, and no smoothing repairs it because the corner's mass sits in the top band with coefficient exponent 1. Since every band budget (this corollary, grouped (14) per band, residual-coverage (9) per band) is nondecreasing in rho and zeta, an r-size split covers a box iff the top band does, so the split region equals the current region. The constraint a+b<40/39 is the same one endpoint-fourier's use of Theorem 1 forces; Corollary 1 inherits Theorem 1's bracket. DFI's older determinant corollary (statements via the BC text; both DFI primary texts unreachable today) has diagonal 48/95 and adds nothing. No note had applied Corollary 1 before; the structural literature audit does not cite BC. **Evidence:** [note](determinant-corollary.md), [validator](determinant-corollary-validation.js). **Reuse or revisit condition:** the four-piece decomposition and the main-term identification are reusable for any determinant-equation input. Revisit only with an error term whose (N_1 N_2) exponent is below 33/106 at kappa=1/4, or whose (N_1+N_2) exponent is below 99/560 at gamma=7/20, or with a Delta-averaged version (unavailable at fixed Delta=2), a different decomposition of beta, or a joint-band estimate. #### Kernel-sign-control — Measured Mobius signs against random signs in the kernel moment **Grade:** MEASURED, pre-registered falsifier, four negative controls fired; no asymptotic content. **Question:** `Q-kernel-sign-control`. **Result:** at (delta,nu)=(8/25,9/20), x=2^j for j=18..30 (M from 1082 to 114104, N from 274 to 23170, harmonic bands A in {1,2,3}, 8 seeded random-sign draws per configuration, 1031.7 s wall clock), the slope of log2(|X_small(actual)|/|X_small(draw)|) against log2 x is -0.057+-0.174, +0.209+-0.207, -0.144+-0.224 and -0.081+-0.302 in the four families (top box g=2 at the top band and at A=1, top box g=1, low box g=1); mean plus sd is nonnegative in every case, so the pre-registered falsifier fires: no heuristic support for Mobius-sign cancellation in the small-j kernel along this route. The moment ratio actual/random is 1.027 to 1.014 with slope -0.0024+-0.0035. Signs help against absolute values by about 3x (geometric mean 0.316); Mobius signs do not help against random signs. The R=0 class carries 96 to 103 percent of the moment, complete periods are proved empty at these sizes, and X_small is 0.02 to 2.88 percent of the moment. Measured moment slopes 0.98 to 1.00 against budget slopes 1.33 to 1.50. **Limits:** the reachable regime is diagonal-dominated, so the Weil term that (21) concerns is invisible; J0=floor(x^(1/20)) is 1 or 2, so small common divisor means gcd 1 or 2; Z=2 collapses the prime-power sector to r=2; one box, native orientation, two bands. Finite scaling proves nothing asymptotic in either direction. The embedded tail carries a forced stamp explained in the note: the first binding predated the fourth table and the guard refused the changed output, with 0 of 546 figures failing to reproduce. **Evidence:** [note](kernel-sign-control.md), [script](kernel-sign-control.js), [artifact](kernel-sign-control.json). **Reuse or revisit condition:** a regime where the nonzero kernel is a measurable share of the moment needs x>=2^60 for floor(x^(1/20))>=3; do not rerun below that without a changed question. #### Mobius-bv-derivation — Provenance repair for the Mobius Bombieri-Vinogradov input **Grade:** DERIVED from numbered published theorems, reviewed once with no defect: all four Koukoulopoulos statements and numbers confirmed verbatim in the author's preliminary version, the identity proved, every hypothesis checked, the bookkeeping found to over-estimate safely by log powers; VERIFIED identity checks. **Question:** `Q-mobius-bv-derivation`. No twin-prime statement changes. **Established record:** no published theorem statement of sum_{q<=T^(1/2)/log^B T} max_{(a,q)=1} |sum_{m<=T, m=a(q)} mu(m)| <<_A T/log^A T was located on five channels, searching in the owning convention of [SEARCH-CONVENTIONS.md §1](SEARCH-CONVENTIONS.md) (Mobius function in arithmetic progressions, Bombieri-Vinogradov) (Granville-Shao Adv. Math. 350 assert it without a locator; Koukoulopoulos GSM 203 ch. 26 proves the prime case only; Iwaniec-Kowalski §17.2 and Opera de Cribro §9 unreachable; Tao Notes 3 has only the L^2 form for mu). The note derives the estimate, with max over y<=T inside, from Koukoulopoulos Corollary 13.4 (Siegel-Walfisz for mu with coprimality), Theorem 26.2 (Type I), Theorem 26.6 (Type II large sieve over primitive characters) and equation (26.3), using the identity mu = mu_{>U} * mu_{>V} * 1 - mu_{<=U} * mu_{<=V} * 1 + mu_{<=U} + mu_{<=V} (valid at every n, checked exactly to 10^5), with U=V=T^(1/5): Type I T^(9/10) log T, small conductors by Corollary 13.4 class by class, large conductors by the dyadic split and Theorem 26.6, yielding level Q<=T^(1/2)/(log T)^(A+6), ineffective constant, reduced classes only. The mesh argument in shifted-prime-decomposition for interval endpoints becomes unnecessary but stays correct. **Limits:** theorem numbering read from the author's preliminary version of GSM 203 by two readers, not the printed book; that version's final display in ch. 26 has a harmless misprint (min{U,V} for sqrt U, sqrt V). **Evidence:** [note](mobius-bv-derivation.md), [validator](mobius-bv-validation.js). **Reuse or revisit condition:** replace by a page citation if Iwaniec-Kowalski Theorem 17.4 or Opera de Cribro Theorems 9.16-9.18 is confirmed to state the Mobius case. #### Corner-correlation — Full coefficients and the prime-cofactor subfamily **Grade:** DERIVED classification and fixed-corner proper-power bound; finite identities VERIFIED. **Question:** `Q-corner-correlation` (PARTIAL: the sharp full-corner estimate and global complement remain OPEN). **Established record:** on the exact fixed-eta S_0, the excluded-prime Mobius mean makes its density contribution O_H(x/log^H x), uniformly in the lower corner cuts. Its raw residual is exactly sum_n C(n)C'(n-2). Every prime power r above the threshold occurs with Lambda(r), including multiple powers of the same base. The proper-power terms on the left and right cost O_epsilon(x^(22/25+epsilon)) and O_epsilon(x^(39/40+epsilon)), both negligible for fixed small epsilon. The remaining prime-r terms still have small cofactors s,s'>=1. Only the s=s'=1 subfamily has the mu(n)mu(n-2) representation with nonnegative prime-band weights and exact moving cuts. Its representation error is O_epsilon(x^(19/20+epsilon)). Neither that representation nor proper-power removal eliminates all non-squarefree full-coefficient inputs. **Reuse and failed interfaces:** the prime-dilation average is not an additive-shift average, so an averaged-shift theorem does not directly apply. The subsequent prime-band-transfer DOES give a bounded multiplicative lift and a small continuous and dyadic scale-average saving for the subfamily. The earlier abstract-only/no-purchase summary is superseded by that source-matched derivation. Neither rate gives o(x) or handles the full coefficients. The specified Bettin-Chandee and prime-averaged completion interfaces fail at their recorded hypotheses and budgets; no universal correlation obstruction follows. The sharp term-wise envelopes do not measure the signed remainder, and fixed-eta estimates do not justify shrinking eta with x without uniform loss control. The sign and non-squarefree measurements remain finite proxy observations, not asymptotic lower bounds or cancellation theorems. **Follow-up:** [corner-coefficient-energy.md](corner-coefficient-energy.md) retains all branches in a sharp one-point quadratic form and an auxiliary smoothed norm. [sharp-corner-transition.md](sharp-corner-transition.md) then prices the sharp energy and its non-negligible norm transition; signed/global obligations stay open. A one-sided corner estimate must use a proved complementary bound at the same scales to imply the consumer. **Evidence:** [owning classification](corner-correlation.md), [validator](corner-correlation-validation.js), [prime-band transfer](prime-band-transfer.md), [round review](round-review-0906.md), and the energy follow-up above. #### Consumer-comparison — Our sufficient consumer against the published conditional routes **Grade:** literature readings CHECKED in primary sources where stated, two elementary DERIVED specification remarks, no estimate. **Question:** `Q-consumer-comparison`. The consumer remains OPEN. **Established record:** Murty-Vatwani (JNT 180, 2017) Theorem 1.1 needs EH_Lambda(x^theta log^C x) and EH_{mu_h}(x^(1-theta)); at theta=1/2-eps the first is Bombieri-Vinogradov, so their route reduces to the single hypothesis EH_{mu_2}(x^(1/2+eps)), for which no case is known; their conclusion is a positive proportion (1-A(2)) = 0.2521 of the Hardy-Littlewood main term on [1,x], which implies our consumer at K=0 by the elementary cumulative-to-dyadic argument. A proof transfer to every dyadic interval is unnecessary for the unbounded-scale consumer. Tao's 2016 asymptotic-sieve notes: only 0<=delta_x<=2 is known, conditional on EH, with delta_x defined up to o(1); the correct dictionary is E_dagger(x)/x = C_2(2 delta_x - delta_{x/2} - 1) + o(1), so the K>0 consumer lies below that normalisation's resolution. The block-summed form on unbounded scales is equivalent to the dyadic consumer at fixed K, with a possibly different positive constant and witnessing scales. At K=0 this is cumulative scalar positivity on unbounded scales; eventual positivity also implies it. The former strict comparison is withdrawn by independent review F1. Every integer-side conditional route from a two-point Mobius or Liouville object to twins (Pintz 2012 at level >3/4 for five sequences; Murty-Vatwani) needs equidistribution in progressions, not a bound; Sawin-Shusterman over F_q[T] derives twins and Chowla from a common geometric input; Siegel-zero routes (Heath-Brown 1983, Friedlander-Iwaniec, Tao-Teravainen 2022) imply our consumer at K=0 by applying Cor. 1.8(i) at x and x/2, so a proof of the consumer must be consistent with the Siegel-zero world and would not disprove exceptional zeros. Prior art for the corner shape: Tao states on GEH the equivalence of the twin asymptotic with sum mu(n)1_R(n)mu(n+2)1_R(n+2) = o(x/log^2 x) and why Chowla progress does not transfer; Maynard ICM 2022 Question 17 poses the determinant-2 sum with arbitrary coefficients as open. The corner identification is an independent rediscovery with a different weight (large prime factor, not rough) and a log^(4+eps) relative saving against Tao's o(1). **Limits:** Heath-Brown 1983, the Vatwani Math. Z. journal copy, Opera de Cribro chapter bodies, Bombieri 1975/76 and Ng 1992 were not reachable; every negative is "not found on the channels used" in the owning conventions of SEARCH-CONVENTIONS.md. No published statement identifies the Vaughan Type II remainder with weighted two-point Chowla; the shape is published, the identification is not. **Source follow-up:** [moving-cutoff-parity.md](moving-cutoff-parity.md) now retrieves the 2018 author preprint for the Math. Z. paper, reads its named hypotheses and repairs the relevant 2017 dyadic switch. The preprint body is accessible; its equivalence to the journal copy was not checked. The source equality omits a moving endpoint, but the repaired fixed-interior-cutoff conditional consumer survives. The new note owns that validation and its scope. **Evidence:** [note](consumer-comparison.md). **Reuse or revisit condition:** any brief asking for "a two-point cancellation bound" must say whether it means a bound or uniformity in q; the published routes need the latter. Cite Tao's post and Maynard's Question 17 in any write-up of the corner. #### Corner-measurement — The corner correlation at finite scales against a random-sign control **Grade:** MEASURED, pre-registered falsifier, four controls fired, one independent reimplementation agreeing exactly; no asymptotic content. **Question:** `Q-corner-measurement`. The corner remains OPEN. **Result:** K(x) = sum_{x/21 branch is essentially unmeasured (s=1 is forced whenever x^(2 eta_0)<2). The absolute target's log^(2+eps) x saving is invisible over the factor 1.8 in log x available, so a negative was the expected outcome even if the target holds. Standard errors are scratchpad-grade. **Evidence:** [note](corner-measurement.md), [script](corner-measurement.js), [artifact](corner-measurement.json). **Reuse or revisit condition:** do not rerun a direct corner sieve; the band degeneracy makes it uninformative below x=2^70. A different finite probe would need a model with populated bands and a stated falsifier. #### Corner-log-average — Corrected transfer assessment **Grade:** DERIVED support/mass and quantifier corrections; scoped source comparison. **Question:** `Q-corner-log-average`. Pure bands are invariant under outside-prime dilation. Exact cofactor windows are not, but their support is not a single dyadic block and a fixed dilation can retain terms. The natural-average unsigned swap cost is 2*eta*(1-log 2)*x*log x+o_eta(x log x); this does not lower-bound the signed error. Uniform-prefix logarithmic bounds transfer by Abel; c>3 is one sufficient exponent for o(x) at the stated envelope, not a necessary condition for every argument. Pilatte's 1/(96e) is arithmetic for one displayed parameter choice, not a universal method ceiling. **Evidence:** [owning note](corner-log-average.md), review 21 F6--F7 and the finite proxy validator. Historical T4 predictions are not a proof of the corrected leading constant. No full-corner bound follows. **Reuse:** direct fixed-parameter substitutions still need their stated hypotheses; do not use nonmultiplicativity alone to exclude a lift. #### Round-review — Review of the completed round and the next research specification **Grade:** REVIEW of local deductions and checked source hypotheses; DERIVED elementary sampling consequence; finite algebra checks. **Question:** `Q-round-review-0906`. The fixed-band Fourier lift and continuous moving-window estimate survive review against Tao–Teräväinen v2 Theorem 3.1(ii) and MRT (1.12). For a bounded fixed sequence, normalized sums over (N,2N] vary by O(h+1/N) on [N,(1+h)N]. Integrating these disjoint neighborhoods and then using the existing moving-band mesh gives a dyadic scale-average saving with exponent c_*/2 for |A_w(N)|/(N log^2 X). The bare exceptional-set argument did not justify excluding this sampling route. Proper-prime-power terms already have total absolute bound O_epsilon(x^(22/25+epsilon)+ x^(39/40+epsilon)) on the fixed corner; they were incorrectly listed as an open obligation. The driver's integer-parity split does not measure per-term CRT gcds; its total coefficient identities survive. Universal diagonal lower bounds fail under arbitrary bounded coefficients. The conditional reachability intersections require six scalar cutoffs, and the fixed-margin bulk can disappear before gamma=2 even though edge strips persist up to that threshold. **Limit:** the dyadic saving is relative to N log^2 X, with a small unextracted exponent; it is not o(N), every-dyadic control or a twin margin. Remaining prime-r terms with s>1 or s'>1, their non-squarefree inputs and the complement remain OPEN. No unconditional region or exact cut changes. This review is not a reproof of every upstream reduction or of the deep imported theorems. **Evidence:** [review and sampling derivation](round-review-0906.md), [finite validation](round-review-validation.js), the corrected owning notes and rerun driver/transfer checks. The full number audit and QC self-tests pass; these do not certify the analytic argument. **Reuse:** price a stronger averaged rate or an estimate for the remaining full coefficients against the global consumer. Do not require an every-scale theorem merely to obtain a dyadic average, reassign the completed fixed-corner prime-power tail bound, or infer a lower bound from an upper budget. Reopen a finding for a precise correctness concern or changed hypotheses. #### Prime-band-transfer — Bounded multiplicative lift and scale accounting **Grade:** DERIVED from an imported primary-source theorem; PARTIAL campaign payoff. **Question:** `Q-prime-band-transfer`. An integrable Fourier superposition represents mu L_B/log X by bounded multiplicative functions. Their squared pretentious distances differ from mu's by O(1) uniformly in Fourier parameters. Tao–Teräväinen v2 Theorem 3.1(ii), Tonelli and an explicitly priced log-scale mesh give a small unspecified log saving in the continuous scale-average of the absolute prime-cofactor sum divided by N*log^2 X. The exact moving cofactor windows are included at fixed eta. The coordinating review's interval-stability lemma additionally yields a dyadic scale-average saving with exponent c_*/2, recorded as (7) in the owning note. **Limit:** no o(N), every-dyadic bound, s>1 or s'>1 estimate, full-corner bound, regional extension or positive twin margin. Proper prime powers are separately negligible on this fixed corner. The source theorem is consumed, not reproved; no novelty claim is made. **Evidence:** [complete local derivation](prime-band-transfer.md), [sampling consequence](round-review-0906.md), [finite ingredient checks](agent-readiness-validation.js). **Reuse:** the two reviews retain (1)–(6), with the scale discussion qualified by the sampling lemma. A stronger global claim needs a matched estimate and all omitted branches. #### Next-transfer-review — Adversarial check of the band transfer at its scope **Grade:** REVIEW of the continuous estimate; scope corrected by the coordinating review. **Question:** `Q-next-transfer-review`. Checks a–k on prime-band-transfer.md (1)–(6) survive: multiplicativity, Fourier identity, uniform pretentious-distance comparison, MRT exponent 7/24, exceptional-set structure, Tonelli, CRT error against N>=sqrt X and mesh count. The source's technical condition (3.2) belongs to case (i), not the consumed case (ii), independently confirmed at source. Four unstated steps (shift assignment, frozen-band uniformity, the two-integer lower-edge set and width-2 strip) are now written in. **Correction:** check l did not consider bounded-interval stability. A weaker dyadic scale-average estimate follows as proved in round-review §2. The no-o(N), no-every-dyadic and no-margin limits stand. Proper prime powers already have a separate fixed-corner bound; they are not open merely because the lift does not represent them. Imported proofs were not independently reproduced. **Evidence:** [review](next-transfer-review.md), [finite validator](next-transfer-review-validation.js), [coordinating correction](round-review-0906.md). **Reuse:** consume the explicit derivations with their hypotheses and check any inference that determines a new research decision. A review label is not a substitute for the argument. #### Corner-branch-diagnostic — Exact branch-preserving enumeration of C(n)C'(n-2) **Grade:** VERIFIED finite algebra; tooling only. **Question:** `Q-corner-branch-diagnostic`. A driver computes each side by the direct divisor formula and by the expanded prime-power identity independently, compares them as integer prime-log coefficient vectors, and retains the 3-by-3 (P,S,Q) contribution matrix, both integer parities and the four squarefree classes, with signed, term-absolute, grouped-absolute and support totals reported separately. Five controls are ACTIVE on a disclosed fixture and on the archived window q=9973 (8192 sides, 0 mismatches; deleting S or Q breaks the identity; repeated powers counted separately; every integer reconstructs from its stored factors). The factor artefact is read, never written, with its source and window hashes printed. **Limit:** the parity split is gcd(n,n-2), not the per-term CRT gcd partition (round-review R3). The fixed-eta corner is empty on the retained prefix (0 integers with both sides nonempty at x=2^27, eta_0=1/500); Elo also exceeds E0 across the entire J_x; the embedded matrix uses the handoff cuts on 6.1e-5 of J_x and is a proxy window, not S_0. No sign trend, saving or closure is inferred; the right cutoff Z=2 is degenerate. **Evidence:** [specification](corner-branch-diagnostic.md), [driver with two embedded tails](corner-branch-diagnostic.js). **Reuse:** use it to enumerate branches for a later named mechanism on retained factors; a larger census needs the condition in [data-reuse-audit.md](data-reuse-audit.md) §5. #### Next-correlation-source-map — Scale and coefficient quantifiers beyond the transfer **Grade:** scoped NEGATIVE for three primary statements read at source. **Question:** `Q-next-correlation-source-map`. Tao-Teravainen 1809.02518v2 Cor. 1.13/1.14 has fixed functions, no rate and a logarithmic-density-zero exceptional set, weaker on every axis than the statement already consumed. Guo 2608.23500v4 Thm. 1.1/1.8 holds at every large scale and covers shift 2, but only for the Liouville pair in logarithmic normalization at rate (log x)^(1-c), below the natural-average conversion threshold of [corner-log-average.md](corner-log-average.md) §3; its proof was not examined. Tao-Teravainen 2512.01739v2 Thm. 3.1(i) shares 3.1(ii)'s exceptional set: its real-valued condition fails for a general complex lifted factor, and its g1(p)=1 condition fails for mu. **Limit:** none of these statements supplies the required rate and full coefficients. An every-scale input is one possible sufficient form, not a necessary gate: the coordinating review already derives a weak dyadic average from the consumed source. Averaged alternatives must price their actual sampling and normalization losses. The prime-r terms with s>1 or s'>1 remain unestimated; proper-prime-power terms already have a separate fixed-corner bound. This is not a claim that no suitable theorem exists. **Evidence:** [source map](next-correlation-source-map.md) with fetch records and hashes; [coordinating review](round-review-0906.md). **Reuse:** match a further candidate against the actual missing rate, coefficient class and consumer quantifiers before another lookup round. #### Signed-moment — Alternatives to the first Cauchy inequality **Grade:** DERIVED elementary bounds and joint-Cauchy pricing; one CONDITIONAL budget and one HEURISTIC ceiling remain distinct. **Question:** `Q-signed-moment`. No region or twin margin is added. Lemma A includes only u1=u2,h1=h2, not unequal proportional R=0 pairs. Reviews 20--21 retain its endpoint integral/Abel repair and check Lemma B's geometric-series argument. Lemma B's exponents are (1+a+b)/2, b+1/2, a+b/2, b+a/2; its usable region is analytically contained in the already controlled product region. Joint Cauchy in (m,h) is now priced: a+b/2, a/2+3b/2, a. It gives 103/100 at the target and 2 at the corner, adding no region. Modulus larger than the interval does not prevent finite Fourier completion. The true diagonal has only an upper bound under arbitrary coefficients; the old general Holder floor fails for sparse coefficients (review F9). The random-matrix lower-bound model remains heuristic; no actual coefficient lower bound follows from the available upper estimates. **Evidence:** [note](signed-moment.md), [readiness review C](history/reviews-0906/21-readiness-review.md), [signed-moment validator](signed-moment-validation.js) and finite readiness controls. **Reuse:** a new attempt must change an estimate or hypothesis, not reprice joint Cauchy or infer a universal barrier. #### Heath-brown-edges — Formal pieces, not an intrinsic obstruction **Grade:** DERIVED identities/classification with corrected scope. **Question:** `Q-heath-brown-edges`. The identity and subset-sum lemma survive review. The exponent simplex has sup a=1 at the balanced j=K configuration; exponent-zero free variables may be subpower, not just bounded. This prices separately bounded pieces and does not exclude identity-piece cancellation. The Vaughan log and low-prime terms can cancel exactly on some formal edge configurations. The bounded-cofactor correlation retains affine multipliers l,l', so it is not literally F(n)F(n-2). Finite F-coefficient masses are not asymptotic lower bounds for that signed correlation. A direct counting proof shows the fixed Linnik truncation has a large term-wise absolute tail; the signed tail remains unestimated. BV's exponent-1/2 endpoint needs logarithmic slack, and divisor-bounded coefficients need actual normalization costs. A weighted-cofactor theorem cannot inherit budgets by a formal exponent substitution. **Evidence:** [owning note](heath-brown-edges.md), review 21 F8 and its existing finite identity validator. Maynard Question 17 and Lemma 18 were checked again at source; they are not impossibility statements. **Reuse:** only the exact identities and scoped classification. No full alternative reduction, new controlled region or twin margin is claimed. #### Centered discrepancy estimate — Exact flip retained; top-range truncation proved at its scope **Grade:** exact identities DERIVED and finite checks VERIFIED; the top-range bound T^top=O_(A,eps)(x/log^A x) DERIVED 2026-09-08 and read twice independently the same day (handler, lane V; reader V2, [reviews-0908/01](history/reviews-0908/01-second-reader-A-B-C.md)), accepted at its stated scope with ineffective constants. **Question:** Q-centered-discrepancy-estimate. **Established:** D_y=D^(e_1)+T^top-P^top exactly at the stated finite cutoffs. The top divisor flip uses mu(e)mu(em)=mu(m)1_((e,m)=1) for squarefree e and retains m>y. The density projection is controlled by the existing weighted Mobius means. Section 3a of the note proves, for every fixed A>0 and 0=-4x/25+o(x) is equivalent to D^(e_1)>=-4x/25+o(x). **Earlier failed step (a5e1244, repaired):** the first proof changed n>=e_1*m to a strict endpoint, omitted odd b, and inverted the reciprocal-totient factor while losing (m',g)=1. The proposed H_band interface remains withdrawn. **Review scope:** the reader reconstructed (BV*), (3a.6)-(3a.9), (3a.14) and (3a.16) by hand. The retained integration validator checks four (b,g) pairs at u=1e6; other earlier diagnostic counts are not its recorded invocation. The 2026-09-09 reading also repairs the floor factor 2 and the log^(L+3) error count, absorbed by fixed power slack. Finite checks do not prove the rate; the analytic steps rest on the two cited Tao statements. **Evidence:** [note](centered-discrepancy-estimate.md) sections 1–3a; [worker validator](centered-discrepancy-estimate-validation.js) section 4, [review](research-round-validation.md) sections 2 and 7 and [regression checks](research-round-validation.js). **Reuse/revisit:** reuse the exact flip and the truncation (3a.1) at the stated scope. It is a Murty–Vatwani-type freedom to fix the endpoint at x^(1/2+eps), not a signed estimate: D^(e_1), the twisted sequence Lambda(n-2)mu(n) at odd moduli e=3 is imported. The band obstacles in note section 4 remain open. The reviewed A2 continuation below pays the low Type I term without bounding the whole fixed-endpoint discrepancy. #### Fixed-endpoint discrepancy — Reviewed low Type I estimate; exact remainder OPEN **Grade:** DERIVED and reviewed analytic estimate with finite identity controls. **Question:** Q-fixed-endpoint-discrepancy. **Established:** Vaughan decomposition gives D^(e_1)=2C2*M+T_I^low+B+O_A(x/log^A x), B=T_II^low+P_band exactly as (2.9). The repaired T_I^low=O_(A,eps')(x/log^A x) survives the independent 2026-09-09 reading, so S=C2*x+B+O_A(x/log^A x) is accepted. **Failed step repaired:** e[r,g]<=e_0UV was false; the retained witness has q=43681>1908. Truncating g<=G=(log x)^(A+13) pays both tails and keeps body moduli below x^(1/2-eps'/3)G. The density sum_(g|e)mu(g)/phi(e[r,g])=1_((r,e)=1)/(e phi(r)), multiplicity tau(q)^4 and the uniform Mobius mean at (k,e) complete the estimate. The review corrects a log^4 weighted count, the repeated-prime local exponent and the power-of-two atom for every 0=-4x/25+o(x). It implies, but is not equivalent to, B>=-(C2-1/200)x+o(x). General H_B has fixed c0>0 on unbounded dyadic scales. All remain OPEN. The all-modulus absolute statement (4.9) is one stronger sufficient band input, not a necessary condition. An elementary O(x log^5 x) bound is used for B; the claimed O(x log^4 x) was not justified by applying Brun–Titchmarsh to every divisor class. **Evidence:** [proof](fixed-endpoint-discrepancy.md) sections 2,4, [lane validator](fixed-endpoint-discrepancy-validation.js), [independent review](research-round-validation.md) section 12 and [rational density/consumer controls](research-round-validation.js). **Reuse/revisit:** the Type I review obligation is complete. A new attempt must estimate the actual B with paid coefficient costs, supply a changed representation/consumer or identify a correctness concern. The inspected decompositions leave B unestimated; they do not rule out every other one. #### Review of the 2026-09-07 session — six ranked claims read by an outside reader **Grade:** REVIEWED at textual and finite scope; no asymptotic content. **Question:** Q-review-request-0907. **Result:** Claims 1, 2, 4, 5, 6 verified within stated scope, claim 1 conditional on the unread 1974 page of Halberstam and Richert; claim 3 verified with its tolerances corrected (max |D_y/x + (T1/x - C2)| over j>=30 is 1.84e-4, recomputed by the integrator as well). Lemma 1 and Corollary 1 of Kalmynin and Konyagin read at the arXiv TeX and at two page images; Richert Theorem 11.3 and the 1971 Mémoire Theorem 3 read at page image; the Dover OCR index re-fetched; all Brun form. The Corollary 1 representative slip sharpened by exact finite algebra: with representatives in [0, p-1] the printed encoding selects the empty set whenever some Omega_p is not contained in {1}, and selects the avoiders of -Omega in the remaining case; the repair r' = r (mod p), r' = 1 (mod P(z;p)) selects the -Omega avoiders pointwise in all nine tested configurations. The cross-codebase M column agrees at all 23 shared j to 5.4e-12 relative. **Limit:** the 1974 page image is unread; Richert's proof of 11.3 is a pointer, re-derived only in sketch; the Lean build is not reproduced. A finite check does not prove Corollary 1; the statement is consumed from a refereed source. **Evidence:** [report](history/reviews-0907/10-independent-review-0908.md); scratch `crt-check.js` and `a2.js` under the round-0908/V scratchpad (not embedded). **Reuse or revisit condition:** revisit claim 1 only with the 1974 page or a printed proof of Theorem 11.3; revisit claim 2 only if a representative convention other than [0, p-1] is shown to be the authors' intent, which does not change the statement consumed. ### Finite algebra and numerical validation #### S-0905-09 — Reproduced the algebra and existing numerical checks **Grade:** VERIFIED finite computation only. **Questions:** `Q-review-0905`, `Q-chen-fold-benchmark`, `Q-prime-detection-inputs`, `Q-shifted-prime-decomposition`. The campaign's exact validators cover the relevant fold, weight, CRT, Vaughan, parity and determinant identities. Their output was generated and reproduced with the embed tool. Full-gate runs and their scope are recorded in [the changelog](history/CHANGELOG.md). **Evidence:** [review validator](review-0905-validation.js), [Chen validator](chen-benchmark-validation.js), [prime-detector validator](prime-detection-validation.js), [shifted-prime validator](shifted-prime-validation.js). **Reuse:** existing scripts for regression or a stated falsification check. **Limit:** agreement on finite inputs proves no asymptotic estimate and does not referee every argument in the repository. ### Fold ledger and the parity table #### Fold-arithmetic-bridge — Aggregate sieve input and certified failure of two tests **Grade:** DERIVED and independently reviewed identities and bounds; conditional source-uniformity reading explicit. **Question:** Q-fold-arithmetic-bridge. **Established:** T=P_odd*P'_odd/S+S*cov/4-N_odd3 for S>0, with a valid denominator-free form. Above (2X+2)^(1/3) a killed composite is a semiprime; X^(1/3) alone is insufficient. Proposition 3 derives BV for k-fold X^(1/u)-rough products, fixed k,u and squarefree moduli, from Wu Lemma 2.3 read as uniform over bounded coefficients. Proposition 4 gives shifted-prime contamination <=(4D_k(u)+o(1))2C2*X/log^2 X. This is aggregate over factor tuples, not uniform per tuple. Its cardinality form retains an additive arbitrary-log error even at k=u. **Failed tests:** the two displayed union-bound tests with that aggregate constant fail for every u>4: Q_cov<1 and c*_real<4. The 2026-09-09 review replaces floating-point proof bounds with directed rational certificates, while retaining the old grid as a measurement. The certified sub-2 ranges are (4,4.8] and (8,infinity). **Evidence:** [proof](fold-arithmetic-bridge.md) sections 3a–4a, [lane validator](fold-arithmetic-bridge-validation.js), [review](research-round-validation.md) sections 4,11 and [rational certificates](research-round-validation.js). Wu page 6 was re-read at the rendered PDF; the original Pan–Ding proof remains unread. **Limits/revisit:** the failure does not decide Cov_u or Dec_1, require a second Liouville hypothesis or exclude a joint bound retaining the partner's parity. Reopen with an improved input/consumer or a correctness concern. No sufficient twin-prime margin follows. #### Theorem 2c source review — every hypothesis of the substitution read at source **Grade:** DERIVED (the manuscript's Theorem 2c), not refereed; source review VERIFIED within stated scope. **Question:** Q-two-class-theorem2c-source-review (TODO W); prior Q-kk-substitution, Q-paper-kk-draft. **Established:** the chain of paper/kk-lower-bound.md Theorem B, re-derived in history/reviews-0907/11 section 3, is proved from its cited sources once Kalmynin and Konyagin's Corollary 1 (arXiv:2302.00459v2, TeX read in full; Izvestiya 88:2 read at extracted text) is granted at its printed statement. The smooth-number input is Hildebrand and Tenenbaum 1993 Theorem 1.2 / Corollary 1.3 (page image), range slack; the band-3 count is Rosser and Schoenfeld (3.5), (3.6) (page image). kappa = 4 enters only the implied constant. No progression, Brun–Titchmarsh, Chebotarev or Galois input is consumed. Proposition 1 holds exhaustively at y = 5, 7, 11, 13; the three-band CRT composition produces twin-free runs at toy parameters with zero twin slots inside (finite check only). **Limits:** Halberstam and Richert Theorem 2.2, cited by the source's Lemma 1, is unread at the page (shared with Theorem A). A second bounded access pass on 2026-09-08 (thirteen channels, [history/reviews-0907/12](history/reviews-0907/12-halberstam-richert-second-access.md)) reached no page image and no verbatim secondary; the OCR-read hypotheses (Ω_1), (Ω_2(κ)) or (Ω), (R) and both clauses are discharged by the substitution with A = 1, A_1 = κ + 1, A_0 = κ; the owed item is a page image of pp. 68 to 69 (Academic Press 1974 or Dover 2011), obtainable by an archive.org loan or a library copy. Implied constants unpriced; y_0 = 10^{134.1} is a floor with constants set to 1 and H2 binding. Band 2 is empty at every computable y, so no finite run exhibits the asymptotic construction. Not refereed. **Defect found:** the explicit Mertens error quoted as Rosser and Schoenfeld "Theorem 20", 1/(10 ln^2 x) + 4/(15 ln^3 x) for x >= 286, is Dusart's Theorem 6.10 (x >= 10372); Rosser and Schoenfeld's (3.18) has 1/(2 ln^2 x) for x >= 286 (confirmed by the integrator at the page image). Provenance only; no number of Theorem 2c moves. **Evidence:** [report](history/reviews-0907/11-two-class-theorem2c-source-review.md) (hypothesis table, chain, artifact hashes, scratch script). **Next/reuse condition:** a reader holding Halberstam and Richert 1974 quotes Theorem 2.2 at the page; an explicit-constants pass turns y_0 into a value; an outside number theorist reads the substitution. No twin-prime payoff is implied. ## Closed routes **Scope of closure.** A counterexample to an inequality, a bound on a specified certificate class, a finite failed forecast, and a method with a missing input are different outcomes. Each row closes its named attempt at its cited scope. These rows do not collectively prove a universal obstruction to every argument on the exact tile. In particular, the lambda-ledger's constant-sign result at crystallization and its failed finite anomaly test (`history/staging/attack-lambda-ledger.md`) do not rule out useful conditional decorrelation before crystallization; see `history/staging/review-0905.md` §4. TPC-equivalent targets require an independent derivation; equivalence alone is not a refutation. Each row names an attempted route, its scoped verdict, the failed step, and the evidence. A surviving lemma can still be reused. Read its record before revisiting; a new attempt must change the hypothesis or mechanism responsible for the closure. Dates identify the recorded checks. | route | verdict | why, in one clause | closed | record | |---|---|---|---|---| | separating the four endpoint classes of the (D1) weight as a route to the 7/400 block saving | CLOSED, class-uniform | the four classes collapse to 2e(varsigma R/(cm))[cos(delta R/(cm)) - cos(delta S/(cm))], so the corpus's inflation factor f = min(1,v0) is the size of the DIFFERENCE of two endpoint weights: the split is free exactly at the binding band (v0 = 1) and costs v0^(-2) away from it, so no class carries the deficit alone | 2026-09-16 | return #714; job1506-D1-obligation-dossier.md sections 3 and 7 | | replacing the weight's variation by an l^2 / Parseval norm inside (6)+(7) | REFUTED at the corpus's own lines | (7)'s t != 0 branch bounds the weight by its variation and the Dirichlet l^1, not by a t-sum of the weight, so Parseval has no insertion point there; the substitution moves the block exponent from 407/400 to 519/400, i.e. it pays the factor instead of winning it | 2026-09-16 | return #714; job1506-D1-obligation-dossier.md section 5 | | improving the Weil input relative to G, or imposing G <= x^gamma | REFUTED for this chain | a G-relative gain is identically zero because the gcd averages (8)-(9) are flat in the exponent, and a cap G <= x^gamma is worth at most 1/400 on the A-piece, seven times short of 7/400 | 2026-09-16 | return #714; job1506-D1-obligation-dossier.md section 5 | | localising the required c^(-eta) improvement to a bounded zone of the dual index t | REFUTED at this box | the arithmetic input is uniform in t while the weight's cost is flat per octave, so every octave carries the same power of x and the exemption allowance is x^(-7/400): the improvement is needed up to c^(1-o(1)) | 2026-09-16 | return #714; job1506-D1-obligation-dossier.md section 5 | | the (t,r,c)-local structure of the Kloosterman sum as a source of the saving | CLOSED | the degenerate p-adic scaling reproduces the corpus's sqrt(cG) input exactly; the coset vanishing and the quadratic support give a density G^(-1), hence a gain G*x^o(1) on the binding class; and the mean square sum_t S(t,r;c)^2 = c*phi(c) is r-blind, so no structured mean exists | 2026-09-16 | return #714; job1506-D1-obligation-dossier.md section 5 | | Theorem 1.2 of arXiv:2511.08445v2 applied to (D1)'s pair block after completion | REFUTED for the two-sequence class, SUPERSEDED by the row below | the completed moment still carries `alpha_t = w_c(t)` depending on `R`, so the coefficient is product-shaped and the theorem consumes no such object; the route that reads the same question, Theorem 7.1 / Corollary 7.9, is the row immediately below, and its read withdrew three of the dependencies this route had named as obstructions | 2026-09-16 | `job1506-T5-verdict.md` (sha256 dfda7c91...) | | Pascadi Theorem 7.1 / Corollary 7.9 (arXiv:2511.08445v2) as printed, applied to (D1)'s pair block | INAPPLICABLE at that scope | their kernel admits no weight depending on both indices and after completion the first sequence carries the R-dependence; the three dependencies previously named as obstructions are absorbed by the completion identity | 2026-09-16 | return #714; job1506-D1-obligation-dossier.md section 4 and job1506-T791-verdict.md | | expanding the coprime condition in (D1) by Mobius to reach the e-pair class | CLOSED, scoped obstruction | priced as the route states it, the expansion's modulus-growth channel has ceiling 407/400 - 1/400 = 203/200 > 1 at the top sector, and the 7/100 it needs is seven times the largest it admits (1/100) | 2026-09-16 | return #625; structured-dispersion-estimate.md | | the two submitted Liouville parity-table ratio tests | CLOSED at every fixed u>4 for the displayed aggregate-constant-4 tests; rational certificate reviewed 2026-09-09 | Q_cov(u)<1 and c*_real(u)<4 for all u>4 with the discharged inputs and the corrected contamination constant 4; conditional on the uniformity reading of Wu Lemma 2.3; not a closure of the decorrelation hypotheses or of a joint bound retaining the partner's parity | 2026-09-08 | [fold-arithmetic-bridge.md](fold-arithmetic-bridge.md) sections 3a–4a; [research-round-validation.md](research-round-validation.md) section 11 | | using the printed p. 654 Murty--Vatwani divisor swap without its moving inner endpoint | REFUTED for the displayed equality; conditional theorem not refuted | d>y requires n+h>ey after switching, and x=20,y=3,h=2 gives an exact positive counterexample | 2026-09-06 | [moving-cutoff-parity.md](moving-cutoff-parity.md) sections 2--4, with a dyadic repair | | treating complete prime-partner/complement reassembly as an extra signed estimate | NO NEW ESTIMATE from this algebraic operation | short terms are already within BV/CRT; the uncancelled term is S or the existing Vaughan remainder | 2026-09-06 | [joint-correction-source-audit.md](joint-correction-source-audit.md) sections 2--5 | | seeking a positive main-plus-supported lower bound from the specified separate-sign sieves, with the complement omitted | INSUFFICIENT BOUND for this specified consumer | H_delta<-2/3 makes its lower constant K<-80/9; this does not give an upper bound on the actual residual | 2026-09-06 | [supported-coefficient-dickman.md](supported-coefficient-dickman.md) section 4 | | keeping only the specified M2/M3 pair and main term, or declaring its complement negligible | REFUTED for this pair and current plateau cutoffs | liminf (N2-P3)/(C2*x)>11/10 with prime exclusions paid; S>=0 forces liminf R_rest/(C2*x)>1/10 | 2026-09-06 | [paired-factor-budget.md](paired-factor-budget.md) section 4 | | discarding every positive contribution and bounding the current full negative mass below C2*x | REFUTED for the present cutoff exponents and plateau profiles | a three-prime/rough-composite family with its prime exclusion paid gives liminf Nhat/(C2*x)>3/2, also at fixed delta<1/50 | 2026-09-06 | [switching-negative-mass.md](switching-negative-mass.md) section 5 | | applying GKM Lemma 10.5's cube/m! formula to a general ordered-prime weight | REFUTED at this scope | a bounded antisymmetric function gives zero cube integral but positive ordered-prime mass; symmetry repairs the paper's own application | 2026-09-06 | [smooth-sieve-literature.md](smooth-sieve-literature.md) §3.1 | | replacing the three-smallest-prime majorant by the multiplicative function with its prime values | REFUTED as an identity or pointwise upper bound | w_T(210)>w_T(30)w_T(7) when T>log 7; the general nonmultiplicative upper theorem remains applicable | 2026-09-06 | [global-smooth-majorant.md](global-smooth-majorant.md) §4 | | bounding the full F^- by a constant times the number of small-prime pairs with product >a | REFUTED for this specified majorant | a ten-prime cell has F=-84 and zero such pairs, with strict admissible left exponent margins | 2026-09-06 | [global-factor-signs.md](global-factor-signs.md) §3 | | averaging sharp-window coefficients, then triangle and Cauchy, to obtain o(x log^2 x) | REFUTED for sufficiently small fixed eta | Minkowski bounds this exact norm output below by the transition norm product; signed cutoff estimates are untouched | 2026-09-06 | [transition-joint-budget.md](transition-joint-budget.md) §3.3; [audit](transition-round-audit.md) | | bounding one 1-bounded multiplicative factor against an arbitrary 1-bounded factor at a fixed shift | REFUTED by counterexample | g_2=lambda, g_1(n)=lambda(n-2) gives sum g_1(n)g_2(n-2)=X+O(1); the shift average in Matomäki–Radziwiłł–Tao Theorem 1.6 is load-bearing | 2026-09-06 | [transition-source-match.md](transition-source-match.md) §5 | | transfer the smoothed corner through a negligible L2 error | REFUTED for sufficiently small fixed eta>0 | sharp and transition squared norms are Theta_eta(x log^2 x); this does not rule out signed shift-2 cancellation | 2026-09-06 | [sharp-corner-transition.md](sharp-corner-transition.md) sections 4–5; cross-prime budget retained | | strictly weakening the unbounded-scale dyadic consumer merely by cumulative summation | REFUTED DEDUCTION, at fixed K and existential positive constants | nonnegativity and sum_{i<=j}2^i/i^K=(2+o(1))2^j/j^K give the converse implication with a changed positive constant; both estimates remain OPEN | 2026-09-06 | `consumer-comparison.md` §2; `history/reviews-0906/20-independent-handoff-review.md` F1 | | replacing the distinguished prime-power gcd by its radical in the moment | REFUTED as an arithmetic identity | gcd(25,5^2)=25 but gcd(25,5)=5; the full factor can still be handled by harmonic averaging | 2026-09-05 | `prime-power-dispersion.md` §4 | | excluding every distinct-prime zero numerator without a prime-greater-than-harmonic condition | REFUTED at this unrestricted-frequency scope | (q1,h1)=(5,5), (q2,h2)=(7,7) have the same rational frequency; counting such collisions still permits a sparse dispersion bound | 2026-09-05 | `sparse-dispersion.md` §3 | | omitting distinguished-prime factors from gcd(r,eq) once harmonics exceed q | REFUTED at this unrestricted-frequency scope | e=6,q=5,h1=5,h2=10,theta=1 gives gcd(r,eq)=5 but gcd(r,e)=1; harmonic averaging retains the missing factor | 2026-09-05 | `sparse-dispersion.md` §3 | | a uniform fixed-power operator-norm saving for the complete dual-frequency matrix S(t,lambda*h;q), with arbitrary coefficient vectors | REFUTED at this full-spectrum scope | for prime q and at least two distinct nonzero h rows, the Gram matrix is q^2 I-q J and the norm is exactly q, even after the zero dual column is removed; structured coefficient cancellation remains OPEN | 2026-09-05 | `prime-band-completion.md` §4 | | automatically dropping the common e factor from a cross-prime second-moment phase | REFUTED as a general phase identity | e=5, q1=11, q2=13, h1=h2=1, theta=1 gives modulus 715 and numerator -2; m=1 and 144 agree modulo 143 but have different phases. Gcd-dependent reductions require their own calculation | 2026-09-05 | `prime-dispersion.md` §3 | | multiplying Vaaler's full positive endpoint-error majorant by min(1,T times interval-length / CRT-modulus) | REFUTED at this pointwise scope | fixed-degree polynomials cannot reproduce a sawtooth jump across a unit interval as the modulus grows; the polynomial pairing identity remains valid | 2026-09-05 | `endpoint-pairing.md` §3 | | a uniform fixed-power L2 norm saving for the aggregated squarefree coefficient Q on all expanded-divisor boxes | REFUTED at this norm scope by a derived lower bound | for D=x^delta and W=x^w at fixed positive exponents, the top-band energy is at least c*DW*log W; cancellation in the signed phase correlation remains open | 2026-09-05 | `coefficient-structure.md` §3 | | discarding the determinant-2 singleton family in absolute norm, or bounding its entire negative mass by O(x) | REFUTED at the ungrouped mass scope | with the prescribed cutoffs, each sign's singleton mass is at least c*x*log(x) eventually; the signed difference and cancellation across fibers remain open | 2026-09-05 | `singleton-fiber-audit.md` §§2–6 | | o(x) absolute reconstruction of the shifted-prime remainder from normalized joint local folds of polylogarithmic modulus | REFUTED at this model class | uniformly for even squarefree Q≤(log x)^L with fixed L, Σ_{n∈(x/2,x]} abs(w(n)−r_Q(n))≥(1/2+o(1))x; the signed reconstruction error can still cancel and is not refuted | 2026-09-05 | `polylog-fold-transfer.md` §5 | | the localized merge chain, telescoped to level x | REFUTED | it dies on averaging before any sieve question is reached, short by a factor 9.6 ln x | 2026-08-17 | `localized-04-maxsum.md` §7, §10 | | weakening the gate to M ≤ α·p | REFUTED | the bootstrap map B ↦ 2.4·R·B·ln x is expanding at every x, so there is no fixed point at any α | 2026-08-17 | `localized-04-maxsum.md` §10 | | bounding the increment rather than the level | REFUTED | true increments outrun the gate by an order of magnitude | 2026-08-17 | `localized-04-maxsum.md` §10 | | composing the chain in blocks to reach M ≲ 3.3 x ln x | REFUTED | blocks do not compose; the composed bound exceeds its own gate by 13.2 ln x | 2026-08-17 | `localized-04-maxsum.md` §7 | | the accumulating-index family on the tile (A4, A10's m_eff) | CLOSED | the Overshoot Budget, MEASURED: 0.88 to 1.18 nats of lifetime slack against x² (1.16 to 1.47 against the programme's x′² − 2) versus a counting tool loose by 3.5× to 6.7× at one fold, plus the measured bust at fold 31; the m_eff half dies on step 4's refutation, not on the budget | 2026-08-17 | `gate-multiplies.md` §5, §6 | | TODO 0b as stated, ln c ≲ 2 ln²u/u ⟹ G2(u) < u² | REFUTED | false by a factor ln u; the sharp rate is ln c(p) ≤ 2 ln p/p and the measured multiplier spends 77% to 214% of it on this corpus's ladder (mean 1.30 over p = 7..37) and 106% on Ziller and Morack's | 2026-08-17 | `gate-multiplies.md` §7 | | the Maier matrix as a route to the origin | REFUTED | the informative and origin-distinguished regimes are disjoint | 2026-08-17 | `maier-matrix.md` §5 | | the origin as a distinguished position at S = x′² | REFUTED, and reversed | origin/mean = ρ(2), at the minimum of the survival curve: measured 0.79303 to 0.79922 against e^{2γ}/4 = 0.79305, HL-conditional in its exact value, and below the mean by exact count at y = 13, x = 19 | 2026-08-17 | `origin-excess.md` §2, §5 | | the certificate route (the θ ladder) as a road to TPC | RETIRED | the sharp maximal law that would finish the route is itself TPC-implying, need/z² = 0.49 to 0.61 flat at z = 13..43; the reported z = 29 crossing was separately found to be a fixed-window artifact and is not among the reasons this closed | 2026-08-18 | `theta-ladder.md` §5b; `history/staging/phase1-T4-maximal-law.md` | | extending h2 past 21 terms as the high-value computation | WITHDRAWN | infeasible and non-diagnostic: precision was never the constraint, and the bias is a property of the range | 2026-08-17 | `h2-scoping.md` §5, §6 | | the two-class driving-term route (Holt's method adapted) | route REFUTED, lemma stands | it certifies gaps of size O(x) and cannot reach x² | 2026-08-17 | `two-class-lower-bounds.md` §7 | | Hagedorn's algebraic construction h(n) ≥ 2p_{n−1}, adapted | CLOSED | the same O(x) ceiling as the driving-term route, already free from G₂ ≥ F(x) + 1 ≥ x′; past it the construction meets the first-twin-slot question, which is the postulate | 2026-08-17 | `two-class-lower-bounds.md` §7 | | the proportional-bias exponent correction | REFUTED | the h2 ≥ h floor kills it | 2026-08-17 | `exponent-control.md` §3 | | any twin-specific discrepancy law | REFUTED | the random-class control reproduces it at the one level tested (k = 2, top prime 23, 24 draws, the twin set at the 71st percentile in sd and the 79th in sup), and the growth law's integer 3 is R_k(p) → k+1, class-independent | 2026-08-17 | `discrepancy-two-class.md` §6 | | GS Corollary 1.4 as an obstruction | WITHDRAWN | η is capped at 1/100, so u ≥ 50,000 and log x ≥ (5·10⁶)^{200} | 2026-08-17 | `maier-matrix.md` §8 | | Lemma V's mean-square form as the missing factor | CLOSED | B was never the binding term, the min taking the B2 branch at every z from 13 to 47; and B ≤ 9A²(E−1) = O((log z)⁸) is a theorem with no hypothesis, proved in-repo and re-derived adversarially 2026-08-29, with Opera de Cribro 6.18 a published neighbour for a one-dimensional signed analogue at s ≥ 9 against this corpus's s in [2.0, 3.4] | 2026-08-18 | `history/staging/attack-AB-bounded.md`, `history/staging/attack-tail-maximal.md`, `history/staging/redteam-0829-theorem1.md` | | `u_sup` as an unconditional worst-position bound | CLOSED | measured rising at all nine steps over ten levels (z = 13..47) on a flat ~2.05 per-prime factor, and not a basis artifact, the (h,m) basis being strictly more expensive by a flat 7.26 | 2026-08-18 (basis-independence added 08-19) | `sift-limit-attack.md` §7e; `history/staging/attack-tau-repricing.md`, `history/staging/attack-hm-basis.md` | | Brüdern–Fouvry's left factor as the thing to beat | CLOSED | it is unimodular, so Parseval kills it free; the price is exactly the quantifier | 2026-08-18 | `sift-limit-attack.md` §7e | | re-splitting Brüdern–Fouvry's four side conditions | CLOSED | the apparent extra 0.0295482779 of exponent violates side condition (iii); their point is a vertex of an optimisation already performed | 2026-08-18 | `history/staging/attack-bf-split.md` | | the loss-budget LP at x = 43 | CLOSED | calibrated floor 3.3152 over 33 readings, error bar 0.62 to 0.98 and not converged, with DP1 carrying 71% of the loss; inside the LP's information class θ is the only place left to push | 2026-08-18 | `research/lp-push-x43.js`; `history/staging/lp-push-x43.md` | | fractional retention (Brady / Runbo Li) | CLOSED | the upper-bound window (α_κ, β_κ+1) is empty at every κ ≥ 2, since α₂ = 5.35773 > 5.26645, and the lower-bound half that does apply is worth a projected 3.83e-4 of exponent, 0.017% of the gap | 2026-08-18 | `history/staging/scope-fractional-retention.md` | | "a floor at 4", and the band (4, 4.2665] | REFUTED | 2κ is Selberg's conjectured target and not a proven floor, already beaten for ½ < κ < 1 and open for every κ > 1 per Brady p. 3, and Blight never writes the number; what survives is only that no κ = 2 limit below 4.2665 is exhibited, per `SEARCH-CONVENTIONS.md` §4 | 2026-08-18 | `sift-limit-attack.md` (Blight passage); `SEARCH-CONVENTIONS.md` §4 | | improving β₂ itself | CLOSED | closed for us, not proven unimprovable: unimproved since Diamond–Halberstam 2008 and everything after is worse at κ = 2 (Blight 4.45, Franze 4.516, Opera de Cribro 4.83) | 2026-08-18 | `dhr-verification.md`; `PRIOR-ART.md` | | exact-strata re-insertion as a proof technique | CLOSED | infinite regress: it terminates at every finite z and at no uniform z; it survives as a finite theorem giving G₂(19#) ≤ 210, G₂(23#) ≤ 420 | 2026-08-18 | `history/staging/attack-beta2-03-exact-strata.md` | | the covering economy asymptotically, and the hybrid | CLOSED | it dies at x = 13 where Σ 2/p crosses 1, and β_pure diverges like 7.182 lnln x (2a with a = 3.5911 the root of a·ln(a/e) = 1; the 7.19 that stood here used a = 3.594, corrected 2026-08-29); the exact head reaches only x₀ = O(ln x), so no exponent moves | 2026-08-18 | `sift-limit-attack.md` §7; `history/staging/attack-hybrid-bound.md`, `history/staging/attack-beta2-05-covering-pruning-bound.md` | | Brady's thesis Problem 3 as a licence on the exponent | CLOSED | it is not our covering problem — one class per prime, value-shift, measurably different optimum | 2026-08-18 | `history/staging/attack-np-licenses.md` | | prior art on the `u_sup` representation | CLOSED | the Fejér-mass identity is the classical Σ sin²/sin² = m(n−m) folklore and the mean-square half is published (Opera de Cribro 6.18, Friedlander arXiv:2607.05707), so the representation is an assembly of standard parts; the title-level zeros are not an absence claim and SEARCH-CONVENTIONS has no owning-convention row for it | 2026-08-18 | `history/staging/attack-prior-art-last-ground.md` | | the FKMPT corrigendum alarm (MR4592874) | WITHDRAWN | the corrigendum was read on 2026-08-18, one day before the alarm claimed otherwise; 22 of 25 citations clean and every corrected constant already carried | 2026-08-19 | `covering-dive.md` §2.3; `history/staging/verify-fkmpt-corrigendum.md` | | Ford–Halberstam's dual decomposition, carried out after 26 years | REFUTED | one χ⁺ per component forces D⁻ = D⁺, giving 2(1+√e) = 5.2974 against 5.1581, and 4.3591 against their own published 4.1560 | 2026-08-19 | `history/staging/attack-ford-halberstam.md` | | "Hough's Lemma 3.2" as a separate source | SUPERSEDED | it is BBMST arXiv:2211.01417 Lemma 3.2, confirmed at source | 2026-08-18 | `history/staging/verify-fkmpt-corrigendum.md` §4 | | `L ≤ 111` as the adjudicated block bound | SUPERSEDED | coverability of [1, L] is downward closed, so the bound is first-infeasible-minus-one: `L ≤ 62` stands and 111 is 1.79× weaker | 2026-08-19 | `research/block-L-first-dead.js`; `sift-limit-attack.md` §7a | | the A/B-coupling depth axis as a proof route | CLOSED | the alternation lemmas add exactly 0 beyond exact per-prime counting, since K_p is already an exact maximum; depth 3 = 38 does clear 529's L ≤ 51, so the counting route is open at block 1 and closed on reach, x = 41 being unreachable at every depth run | 2026-08-19 | `sift-limit-attack.md` §7a-ter; `history/staging/attack-ab-coupling.md`, `history/staging/verify-ab-coupling.md` | | certificate monotonicity in L as a live defect class | CLOSED | every first-valid-L in the live layer and the three documents it cites comes from a consecutive scan; the one survivor was an H₁ bisection, wrong at 3 of 5 levels, now boxed as SUPERSEDED; about 40 staging files were not swept | 2026-08-19 | `history/staging/monotonicity-sweep.md` | | the greedy oracle as an exact solver past x ≈ 53 | CLOSED | at one fixed uniform budget, exact at 13 of 13 and 14 of 14 sealed-scope terms but 1 of 8 on the published optima x = 47..79, log-log slope −0.0235 ± 0.007 in ln x; 43 and 53 were budget (4× restarts closed them) and three of the eight misses were still moving with budget | 2026-08-19 | `research/greedy-oracle-validation.js`; `history/staging/greedy-oracle-validation.md` | | a constant-factor submodular certificate for a finite-level G₂ ceiling (import-map row 11) | CLOSED | greedy's coverage is at least L(1 − ∏(1−2/p)) for every L — 0.800 to 0.959 along the ladder, 0.9167 to 0.9965 at the decisive lengths — so no α ≤ 1 − 1/e test ever fires | 2026-08-19 | `research/row11-closure-01-coverage.js`; `history/staging/row11-closure.md` | | the ℓ¹ → ℓ²√log conversion on Θ_e(a)'s arithmetic (import-map rows 5 and 6) | CLOSED | ‖Θ·S_H‖₂ = rms(R_H) exactly, so the conversion IS the sharp maximal law and its true constant (0.56–0.92) sits below the TPC line (1.36–2.23); the available gain is the C^{π(z)} loss itself | 2026-08-19 | `history/staging/import-l1l2.md` | | the local-lemma family (Shearer's exact criterion, Moser–Tardos, resampling oracles, entropy compression) as a route past the Mertens threshold | CLOSED, by three separate mechanisms | on a complete dependency graph Shearer's exact criterion is the union bound; Moser–Tardos gains nothing (chordal graph, mutually exclusive dependent pairs); the variable model that escapes needs structure costing H ≥ x#; and Achlioptas–Iliopoulos is not touched by tightness at all but arrives two levels earlier than Shearer, at x = 7, because this object's causality digraph is complete (INFERRED) | 2026-08-19 | `history/staging/import-shearer.md` §4, §6, §8 | | fold-succession damping (the moiré question) | CLOSED | the succession anti-correlation is level-tracking and nothing more; kill succession is memoryless and record assembly is cooperative | 2026-08-19 | `history/staging/fold-succession-autocorr.md` | | the 0c×0e composition (level-selected residue-deleted maxsum) | CLOSED | the chain is circular at m = 1 and alignment goodness is universal, so there is nothing to select | 2026-08-19 | `history/staging/attack-0c0e-level-selection.md`, `history/staging/verify-monotone-depth.md` | | the maxsum bridge as a 0c→L converter | CLOSED | maxsum_k ≥ G₂ floors the bridge output at ≈ 0.183x for any maxsum bound whatsoever | 2026-08-19 | `a3-05-bound-L.md` §7; `history/staging/attack-foldL-05-maxsum-direct.md` | | chaining the Tail-Count Transport on the tile | CLOSED | a fixed window index certifies a constant against a diverging truth (108/180/240/330 at every fold 13..29 against a truth of 66..258), and forcing the index to grow prices out at A5's coordinate-free cap K ≤ 1 + θ/(3q); the single-fold instrument stays exact because the alignment sum is the sum over copies and realises every merge | 2026-08-19 | `history/staging/attack-foldL-03-transport.md` §4, `history/staging/verify-tailcount-transport.md` | | gap-genealogy amortization on the tile | CLOSED | births are maximal, not scarce; the ledger fails by exactly the factor ln²x | 2026-08-19 | `history/staging/attack-foldL-04-amortized.md` | | the B-free/Sarnak dynamical import as a route | CLOSED (an exclusion in print) | Araújo's 2026 multi-class generalisation names the Mertens-divergent regime "not an interesting system" — the limit comb is {−1}, entropy 0, not Toeplitz; the dictionary is exact at period scale and silent where Assumption A and H″ live; H″'s dynamical translation predicts the opposite of the measurement | 2026-08-19 | `history/staging/import-bfree.md` | | recognizability-radius route: R_x < x′²−3 ⟹ Zone Postulate | REFUTED | R_x ≥ G₂ − 1 makes the reformulation strictly stronger than the target, and R_x/x′² crosses 1 at x = 11 (1.1006) and reaches 2.5429 at x = 17, so it fails at four consecutive levels | 2026-08-19 | `history/staging/import-bfree.md` §3.2 | | generic chaining against the maximal-law union bound | CLOSED | the entropy integral of the true metric already exceeds the union bound at every level (1.04–1.10×) — chaining's ceiling is below the union bound's floor — and the C^{π(z)} price was never over positions but over the arithmetic of Θ_e(a), an ℓ¹→ℓ²√log statement no geometry on Z/W reaches | 2026-08-19 | `history/staging/import-chaining.md` | | bounding the anchored δ by the forced scale (or any constant) as a lemma | REFUTED as a target | the identity 1+δ = S(0)·N̄/(L0·R0) makes δ > −1 equivalent to S(0) > 0, which at infinitely many x is TPC by `anchored-note.md` Prop 2, so ANY bound \|δ\| ≤ c < 1 is TPC-strength, the fourth wrong-direction arrival; the split identity (1+δ)(1+F) = 1+ρ then shows ρ/F is not bounded by 1 (2.001 at @11, 1.266 at @19) and δ takes the sign opposite to Hardy-Littlewood at those same 2 of 7 levels, while \|δ\| < F at 7 of 7 | 2026-08-19 | `history/staging/import-suen.md` §1, §4 | | a stochastic coupling dominating CRT thinning by independent thinning | CLOSED (argued) | the merge event is a function of the gap value (merges only at g ≡ 0, ±2 mod p), mutually singular with every solvable variant; and H″(m=2) in moment form implies the Zone Postulate, the third wrong-direction arrival | 2026-08-19 | `history/staging/import-thinning.md` | | the L = 1 residue count as a smaller target than the postulate | REFUTED | it IS the Zone Postulate in residue notation: the residue condition is the kill condition with zero slack, and the chain sum vanishes iff G₂ < θ (three lemmas; T* = G₂ measured at four of four windows) | 2026-08-19 | `history/staging/attack-l1-residue.md` | | the instrument-slack channel of the i.o. licence (dial 4's last) | CLOSED | the loosest direct-G₂ instrument oscillates 0.3747 nats against a need of 6.700, no sharp-level signature survives a permutation test at any of 19 instruments (Šidák p = 0.8865), and the two signatures with a named mechanism select finite sets — dial 4 is FINISHED, a measured margin of ~0.7 nats total | 2026-08-19 | `history/staging/ioslack-survey.md`; `research/attack-ioslack-survey.js` | | the one-parameter extinction density law as a fitting model | REFUTED | two calibrations of the same shape disagree by 60%; the two-parameter form is the measured, predictive one | 2026-08-19 | `history/staging/attack-foldL-06-scaling.md` | | the bounded-differences family (McDiarmid, Azuma, Talagrand's convex distance, Warnke, Kutin, Kim–Vu) on the anchored deficit | CLOSED | the first scour prime's worst-case effect d(q₁) misses the theorem's own slack by a factor growing like √N (174 to 70,576 over x = 11..23); the published repair's exponent is MEASURED below 1 and falling at eight levels (0.5323 at @11 to 0.1682 at @37), with a closed form q₁Π/16 converging onto it from below; and independently, the target is a named anchored word on which an ensemble bound of any strength is inert | 2026-08-20 | `history/staging/row7-recon.md` | | the lonely runner / view-obstruction import, and the Birkhoff/bounded-remainder branch (import-map row 12) | CLOSED | the fully general multiplicity version is answered in print with the union bound exactly tight (per the 2025 survey quoting Schoenberg 1976, himself NOT REACHED), the shifted variant is false from n = 5, the field names prime velocities as ITS obstruction, and the BRS branch fails its hypotheses and is dominated 5454× by the Level Ledger at x = 29 | 2026-08-20 | `history/staging/row12-recon.md` | | the distortion method as a route to an interval bound (import-map row 8) | CLOSED | the criterion speaks at two classes and its economy is convergent, but its measures live on a CRT product, so the certified window is at least a primorial and the only interval bridge is exponential in the progression count | 2026-08-19 | `history/staging/import-distortion.md` | | the BGT interpolation machine for 1d limit existence (import-map row 9) | CLOSED as a machine, reframing banked | it closes on hypothesis H1 in the coordinate the exponent lives in, since π(st) − π(s) − π(t) is never 0 at st ≥ 25 (minimum 2, maximum 12 over 104 pairs), for the rotation ensemble as well; the one coordinate where H1 does hold, n = π(x), has limit +∞; the surviving target is the bounded superadditivity defect of S(x) = ln(x²/Ĝ(x)), whose explicit-constant form is TPC-implying below ln C = 1.3946 | 2026-08-19 | `history/staging/import-interp.md` | | the sofic first-moment shape L ≍ p/ln p (import-map row 2's rate claim) | REFUTED | killed by its own pre-registered flatness criterion on the map's own formula (t = 2.69 at seven points) and, independently of any flatness, by arithmetic: granting perfect flatness the measured ratio 2.1346 still misses the requirement by 1.618 to 2.640; the corrected qualifying fraction falls faster than 3/p by an order of magnitude and the first-moment L is MEASURED polylog (ln L = −1.884 + 2.892 lnln x, R² = 0.972, 42 corrected census points) | 2026-08-19 | `history/staging/import-sofic.md`; `history/staging/fdecay-deep.md` | | the tail factor √(2 ln D), and its repair √(2 ln θD), in the maxsum growth law | REFUTED | the measured level sits ABOVE the θ ≤ 1 ceiling at small m and the residual widens with D — refuted by magnitude and by sign | 2026-08-19 | `history/staging/scanstat2.md` | | the linear exponent rule H = a + b·ln D (pre-registered 0.220511 + 0.006140, refitted 0.220795 + 0.006146) | REFUTED, twice blind | validated one level out at T₂₉ and then missed both sealed bands, T₃₁ by 4.96 and T₃₇ by 6.90 band standard errors (1.99 and 3.63 of each measurement's own s.e.), two engines; H is a grid-dependent summary of a curve and no single exponent should be quoted | 2026-08-19/20 | `history/staging/scanstat2.md`; `history/staging/scanstat-t37.md` | | the Chen–Stein route to the 3.8 and the extinction constants, and the super-W decomposition of the deficit (import-map row 4) | CLOSED / REFUTED | one shared uniform residue draw per object leaves no error term: b₃ reaches 0.850 to 0.9998 of its own ceiling on the multi-kill events and 12.7 to 73.4 times b₁ + b₂ on the strike process; and the deficit does not live in the super-W part (δ_tot/s = 0.0195 against a measured 0.0341 at @23) | 2026-08-19 | `history/staging/import-stein.md` | | class-uniform joint caps at the origin (closing 34 → 45 with bounds quantified over scour classes and concluding a class-uniform number) | CLOSED | advmin@11 = 16 exact, two disjoint proof stacks — 18 BELOW the staircase's anchored floor 34, so any class-uniform conclusion is instantiated by the adversarial witness and capped by 16; only anchored-aware caps remain (at @13 at least 155 of the 197 floor-to-truth points are anchored-only) | 2026-08-20 | `history/staging/attack-advmin-1113.md`; `history/staging/redteam-0820-structural.md` §2 | | manufacturing the smooth profile from sieve weights (completion/decomposition on the Rosser factor) | CLOSED | Möbius-signed weights are Fourier-flat: the aggregated low-frequency mass share sits at or below the flat null at every occupied block and at 31 of 31 fixed-s slices (0.57 to 1.15 times the null, median 0.82) where a usable smooth component needs it at scale C̃/K ≫ 100, and the largest single mode, k = 0 included, carries 2.8e−3; the exact blocking term is the high-frequency remainder of the completed expansion, which keeps 1 − o(1) of the ℓ² mass and is exactly as inadmissible as the original | 2026-08-20 | `history/staging/smoothness-front.md` §4; `history/staging/redteam-0820-math.md` §2.3 | | the Kowalski–Michel–Sawin branch for Lemma V | CLOSED | the modulus hypothesis fails on its own (both papers scope themselves to a single fixed PRIME in the authors' own words, read at source, while ours is z-smooth squarefree and is summed over), and the kernel and interval-support hypotheses are jointly unsatisfiable for the rank-1 ratio phase; the granted fantasy gives no saving at all on the binding h ≍ 1 block and H^{−0.011127} on the largest, under a third of Bettin-Chandee's standing H^{−0.035608}; revival needs the squarefree-smooth-moduli extension KMS 2017 §1.5.2 names as open | 2026-08-20 | `history/staging/smoothness-front.md` §5; `history/staging/redteam-0820-math.md` §2.4 | | the Y_N/coefficient axis of the DI/Pascadi frontier past 0.393922 | CLOSED | Y_N enters Cor 18's ℐ² only through the exceptional factor, and the Y_N-free regular-spectrum main term CS(C+DR)(RS+N) binds at 0.393922 by exact arithmetic (1.212157 + 2σ ≤ 2), so no coefficient information moves the frontier further | 2026-08-20 | `history/staging/smoothness-front.md` §3.4; `history/staging/redteam-0820-math.md` §2.1–2.2 | | literature-owned special-level constructions as a source of favourable levels | CLOSED (recon) | the field's level engineering is entirely on the DEVIATION side: no surveyed construction certifies smallness or regularity of a max-type object at its engineered level, and the only offensive i.o.-over-levels quantifier (Heath-Brown's, citation unverified at source) hypothesises its levels rather than constructing them; the absence rests on the family's own surveys, the indexed 11N13/11N25 owning-convention sweep never having run | 2026-08-20 | `history/staging/special-levels-recon.md` | | mirror symmetrization as an improvement channel at the anchored cap | REFUTED | gains exactly 0 — every cap family in play is itself mirror-covariant, and the covariance σ(K_q(a)) = K_q(w−a) with the comb fixed is a one-line theorem, so the zero is structural, not empirical | 2026-08-20 | `history/staging/attack-anchored-01.md` §3; `history/staging/redteam-0820-night-proofs.md` §1 | | K = Ψ/Φ² (the Fold Moment Identity's deviation) as the M_p field mechanism | REFUTED as a field predictor | K spans a factor 1.009 across folds 101/211/421 while M̂ spans 2.28 — the deviation does not carry the field | 2026-08-20 | `history/staging/mp-derivation.md` §3; `history/staging/redteam-0820-night-empirical.md` §T1 | | modelling the twin tile as a repulsive point process (Hermitian determinantal, negatively-associated, strongly Rayleigh) or as an attractive one (permanental, positively-associated) | CLOSED, both signs | its pair correlation takes the value 0 or a value ≥ 2.3812 and nothing in between, so no association hypothesis of either sign holds: g(2) = 0 by mod 3, g(6) = 6∏_{5≤p≤x}(1−4/(p−2)²) = 2.6617 at x = 11 | 2026-08-28 | `history/staging/import-repulsive.md` §2; import row 17 | | the QR refinement of the anchored-cap transplant (immune offset classes in the caps) | CLOSED | the registered surplus test is VOID on its own matched control (a 0.6% design bias at 6.63 s.e. where no class is QR-immune at all), and what stands is the registered placement test at Z = +0.49 plus a POST-HOC design-free class test in which the immune classes deliver exactly the guardrail's r/(r−2) and nothing past it (per-prime Z −0.48..+1.66 over 3.4M twin openers, differenced primary +0.00046 ± 0.00127); as a certificate they are dominated by simply naming the two classes each small prime kills, which certifies 1219/1219 anchors against the immune subset's 1126 | 2026-08-28 | `history/staging/attack-z3-immune-01.md` §3, §4; `research/attack-z3-immune-01.js` | | the K*-product doubling certificate Ĝ(2s) ≤ (K*+1)·Ĝ(s) as the bridge for item D at any C₂ in the legal band [4, 19.2455) | CLOSED | dead on the chain's own exact data plus the red-team-confirmed floor K* ≥ π(2s) − π(s): certificate 18 against target 8 at s = 16 where the true ratio is 5.2727, then ≥ 14 at s = 64 and ≥ 24 at s = 128, past the whole band; the maxsum certificate (sup 6.6364 at s = 16) survives as the tightest proven per-step bridge and its all-s form is the open inequality, so the doubling target itself is untouched | 2026-08-29 | `history/staging/attack-0829n-doubling-bridge.md` §§0, 3; red-teamed `history/staging/redteam-0830-doubling.md`, which finds the row UNDER-claims: Lemma 1 at s = 128 alone clears the whole band (N+1 = 24 > 19.2455), so the exact walk corroborates rather than carries it | | a law D* ≈ Q^c for the parity adversary's killing level, as a candidate for unnamed structure at the wall | CLOSED (sealed prereg, 5 HIT 1 MISS) | no law in Q: the Q-slope moves 1.446 → 1.204 → 0.575 across ranges while ln D* on ln(width) holds 1.097 ± 0.012 with a quarter of the scatter, D*/width ≈ 0.5, and the independent-thinning null reproduces the LN-Q slope within 0.025 (corrected 2026-08-30, `redteam-0830-records.md`: 0.025 is the ln Q slope difference, not the width law's; on the same 84 matched anchors the WIDTH difference is 0.031); D* is where a modulus first reads a single position of the window, the wall survey's remainder statement in the stretch coordinate, and names nothing new; eleven anchors seen around the seal are disclosed in the note | 2026-08-29 | `history/staging/attack-0829n-parity-dstar.md` §§4, 6; red-teamed `history/staging/redteam-0830-records.md`, which confirms D* EXACTLY at 299 of 376 anchors on a BigInt rational simplex (0 contradicted, 77 above its row cap) and re-derives the 172-figure gate from arithmetic, and which rules the closure well powered on clause (b) (paired se 0.0401 against a 0.15 band) | | the Skeleton Equidistribution door (fixed-modulus equidistribution of (q mod M_T, ⌊W/q⌋ mod M_T) over the scour primes) as the route to the all-x aggregate 30-skeleton bound; TODO item 4 retired | CLOSED as a route; the Collapse Theorem and the six-level certificate stand | over every scour prime the branches on which the door is a fixed-modulus question carry 9.2%, −0.9%, 0.2%, 0.5% of the skeleton at @13..@23 (the −10.8% and 5.5% on record were 9% and 2% subsamples), so the door as named moves G30_agg by at most 0.0102 and NOT ALWAYS DOWNWARD (the closable block is −0.0009 at @17, so removing it RAISES G30_agg from 0.1011 to 0.1020; sign corrected 2026-08-30 per `redteam-0830-imports.md`), against an open part of 0.094 to 0.126; on the open side M_T > lB ⟺ q > W/M_T, the phase never wraps and no equidistribution statement remains, only the inequality itself; the one literature instrument (Saffari–Vaughan Thm 10) is a marginal on a growing window and certifies no level; and nothing on the live board consumes G30_agg < 1/2, calm being decorrelated from survival | 2026-08-30 | `history/staging/decide-0830-skeleton-door.md`; `natal-cap-36-skeleton-door.md` §Proposition E; `history/staging/import-fracparts.md` §§5-6 | | the deep-ladder Buchstab transfer (item 8a) by the sharp sieve functions at the depths the run levels reach | CLOSED | the transfer is an asymptotic for a ratio of two κ = 2 sifting functions sharing z = q, so any sieve bracket enters the ratio undivided: F₂/f₂ = 3.668 at @23's head σ = 4.7088 (NOT "the best σ any level has": @29's head σ is 5.5785 and @97's 17.1422; corrected 2026-08-30); the source's level-free implication "|B − 1| ≥ c forces σ < u_c − 1" is REFUTED as a derivation (`redteam-0830-engine.md`: it bounds every u in a weighted average by the largest and mistakes the denominator, and fails exhaustively over all 1,512,930 (q, K) pairs at @23 for the model B at c = 0.10/0.05/0.02 and for measured B at every c), and what survives is the MEASURED statement at @23: worst σ = 2.159 against β₂ = 4.26645, while a lower bound needs σ > β₂; the iterated Buchstab identity with F₂ is negative from K = 1 at q = 37 @23 and Jurkat–Richert fails Ω(1) as stated; the F₂ bound beats the trivial cap at 0 of 1,512,930 (q, K) pairs, floor −1,733,138 at every K | 2026-08-30 | `history/staging/attack-0830-buchstab-deep.md` §§2, 4; red-teamed `history/staging/redteam-0830-engine.md` (26 of 26 comparisons agree; one derivation refuted, conclusion surviving as a measurement at @23) | | the per-fold composition of L across the primes in (s, 2s] as an upper bound on the doubling kill-run (item D) | CLOSED (truth gap) | the composition is a PRODUCT, K*+1 ≤ ∏(1 + L_j) (PROVEN by a nesting chain of survivors; the sum form is false at 7 of 14 steps), so the composed index is ≥ 2^N over N folds and 2^N·ḡ(s) alone beats 8·Ĝ(s) at chain rungs 16, 32, 64 with no walk; the composed certificate exceeds the allowance at 8 of 14 enumerable steps (20.83× at s = 16) while the truth sits under it everywhere (sup w/a = 0.6591 for the truth; the 0.8295 this row carried until 2026-08-30 is the CERTIFICATE's ratio, `redteam-0830-doubling.md`); it is the Tail-Count Transport chain with the index forced to grow, failing at the opposite end of the index dial | 2026-08-30 | `history/staging/attack-0830-doubling-killrun.md` §§2, 3; red-teamed `history/staging/redteam-0830-doubling.md` (product theorem stressed to 21,641,346 nesting links over 6,012,804 killed runs, 0 failures) | | the ρ maximal law / REC(s, u₀) route to a two-class exponent below β₂ (item 0's Lemma V → sup-over-positions arrow) | CLOSED as a TRUTH GAP at rung DERIVED, red-teamed twice on the growth half and once on the exact half | a pointwise floor on the vector-sieve remainder, cc(r) = −(A₁A₂ + A₁B₂ + B₁A₂) as signed counts of Rosser exit chains, is PROVEN (identity holding at every position, 0 mismatches in 20.5M, extended to z = 113), and its DERIVED growth Ω ≫ z^{16s/9}/ln⁸z (s ≤ 3; corollary 2 × capped-LP-max > β₂ for every s > 2.3999) survived two adversarial passes on independent code (2026-08-30, 2026-09-04) on the Rosser vector-sieve lattice at every admissible level split (1643 splits, minimum margin 0.64; no other admissible weight system is derived); if the growth holds then REC(s, u₀), F1, F2 and RML(α) are FALSE for every u₀ < 16s/9 > β₂, so the route is false rather than hard, and the proof-gap reading is refuted because REC quantifies over the remainder of the very certificate the floor bounds; the falsity lives beyond a level that depends on (s, u₀): z ≈ 10^15.3 at s = 3.0 and u₀ = β₂, 5.6e31 at s = 2.698721 and u₀ = 4.2165, and no computation reaches any of them; two failed breaks on a derivation are two failed breaks, so the row is worded at rung DERIVED and not above it | 2026-08-30, reworded 2026-09-04 | `history/staging/attack-0830-rec-cheapest.md` §4; `history/staging/redteam-0830-floor-growth.md`; `history/staging/redteam-0830-floor-sign.md`; `history/staging/attack-0829n-rml-proof.md` §3 (REC's statement); `history/staging/redteam-0904-floor-growth-2.md`; `history/staging/redteam-0904-item0.md` | Recorded script outputs that cite [REFUTED.md](REFUTED.md) resolve to this table through a compatibility pointer; all outcome updates belong here.