# Documentation supporting records — demo-v1

Date: 2026-10-12. Source revision: `01ae3bdac6f61ecbabcd72e0116ebb2860e93157`; input as-of: 2026-09-18; fictional dossier: HER-scenario-1. This is executor readiness work, not independent assessment or institutional validation. The selected OpenProse agreement is release 0.4.6-rc.1. Scope: inspect frozen evidence and create documentation; no financial-model script was executed.

## Record 1 — Section plan and requirement mapping

This final section plan maps content obligations to concrete locations. Requirement IDs remain in this supporting record. “Met” means the documentary obligation is satisfied at the specified location; it does not establish that missing actual institutional evidence has been supplied. The individual judgments below are the executor's assessment of the final artifact. Sections and table numbers are locators in `document.md`.

| Requirement | Status | Document/report locator | Evidence and remaining work |
|---|---|---|---|
| R-01 | Met | Front matter; §1 first two paragraphs | Plain-language discounting and projection, fictional commission and decision; BRIEF Model, INST Inventory |
| R-02 | Met | §§1–2 | Historical demonstration, proposed sandbox, no approved use; products/currency/collateral/horizons and exclusions distinct |
| R-03 | Met | §3.1–3.2 | Deposit and lagged OIS equations, curve/forward relations, interpolation/solution and exact date conventions; source functions cited |
| R-04 | Met | §4.1–4.2, Tables 1–2 | Real fixing, public prints, derived estimates and fictional facts separated; timing, unit, filters, transformations and limits |
| R-05 | Met | §7.1, §7.4, Tables 5–6 and 11 | Historical developmental checks, all repricing rows, convention perturbations; current arithmetic distinguished |
| R-06 | Met | §7.2–7.3, Tables 7–10 | A implementation check separate from method comparison; spot/forward cohorts, residual signs and common adjustment explicit; no licensed benchmark |
| R-07 | Met | §§5.1, 7.4, 8–9 | One date, sparse long end, locality, extrapolation, event steps, conventions and stress limitations |
| R-08 | Met | §8, Table 12 and paragraph after table | Each material limitation linked to proposed control or explicit absent/dedicated-control gap; no effectiveness assertion |
| R-09 | Met | §9, Table 13 | Construction and quote assumptions with reason/evidence, consequence and monitoring/unresolved work |
| R-10 | Met | §8 Table 12 and cadence paragraph; §10 exception process | Thresholds, owners, candidate/quarterly/event cadence and hold/review triggers from fictional dossier |
| R-11 | Met | §10 and §8 gap markers | Fictional inventory/tier/roles/change/findings/sign-off; actual institutional facts visibly unavailable |
| R-12 | Met | §§6 and 11 | Named scripts, historical versions and shared schedule/data dependence; customized B and comparison limitations |
| R-13 | Met | §12 plus locators throughout | Every citation key expanded to public origin; no private repository used as an answer source |
| R-20 | Met | §§3–5.3, 7.1–7.4, 9; Record 2 below | All seven principal choices and 27 smaller entries preserved with reasons/alternatives/limits; unspecified alternatives marked |
| R-21 | Met | §5.1 opening equation/Table 3; §7.3–7.4 | Single-cash-flow maximum, signs and units; smoother-forward versus wider-risk tradeoff; no “profit” inference |
| R-30 | Met | §6, Table 4; §12 source links | Source identities and historical receipts; full offline raw-data reproduction explicitly unavailable |
| R-31 | Met | Captions/source lines for Tables 1–13 plus 3a; Figures 1–4 | Dated calculation identities, units and populations; Table 5 current arithmetic and Table 10 arithmetic confirmation labeled |
| R-40 | Met | Claim locators throughout; Record 3 | Material relationships checked against exact sources and code; attribution-only external claims separated; no claim that unchecked papers were read |
| R-41 | Met | §11 second paragraph; §12 external references; Record 3 external status | Only registered bundle evidence used; unavailable/unretrieved primary contents not asserted as independently established |
| R-50 | Met | Front matter; §§2, 8, 10 (14 gap markers) | Fictional dossier clearly simulated; every selected actual institutional slot unavailable or covered by actual new document history |
| R-60 | Met | Front matter status; §§1, 10; Record 4 | Readiness, self-check, independent assessment and actual authorization kept separate |
| R-61 | Met | Front matter blockquote | Exact house disclaimer retained; applies to extracts too |
| R-62 | Met | This table; agreement/house mapping below; Record 4 | Each requirement separately assessed; unresolved substantive real-world evidence listed without treating disclosure as supply |

### Agreement and house requirements beyond the base IDs

| Selected obligation | Status and exact home | Remaining substantive work |
|---|---|---|
| Kernel: selected identity, nested adoption and origin retention | Met; `result.md`, `input-manifest.json`; six pinned definitions read and declared hashes matched | Release is experimental; no general conformance certification |
| Documentation contract: complete methodology rather than decision note | Met; all twelve numbered sections, about 7,195 words excluding table rows/reference list at draft check | Independent content assessment still required |
| Documentation contract: section plan, choice register, citation audit, readiness and result | Met; four clearly separated records here plus `result.md` | No implication of independent assessment |
| Model-description contract: purpose, inputs/outputs, method, assumptions/limits and revision | Met; front matter and §§1–4, 8–9 | Actual deployment facts unavailable and marked |
| Numerical-evidence contract: exact populations, intervals, units, comparison baseline, precision and interpretation | Met; §§5 and 7, especially Table 7 and §7.3 statistical conventions | Raw-source reproduction outside scope; untested regimes remain unknown |
| Claim-evidence contract: relationship checked, missing/conflicting claims visible | Met; Record 3 and source-located document | External paper contents unchecked; used only as attributed developer rationale |
| Institutional-facts contract: simulated facts and exact missing-fact marker | Met; §§2, 8, 10 | Actual institutional inventory, feeds, exposures, roles, policies, benchmark, assessment and signatures remain unavailable |
| House: exact title/version/date/status/inventory and three-sentence conclusion | Met; front matter | None for documentation presentation |
| House: twelve exact numbered `##` headings | Met; §§1–12 | None |
| House: captions, units, dates, calculation identities, local figures | Met; all tables and Figures 1–4 | Plots are retained historical images, not newly generated |
| House: all 22 input errors and 0.01 bp criterion | Met; Table 5, current frozen-column arithmetic | This is not a fresh calibration run |
| House: ≤1 bp held-out median criterion; exact A implementation grid and max/mean; distance bins | Met; §7.2–7.3, Tables 7–10 | Criterion passage does not give production acceptance |
| House: paper checks, round trip, sensitivities, continuity, collar, extrapolation, locality and failed QuantLib variant | Met; §§5.2, 7.1, 7.4 | Primary paper interpretation and solver generality are not independently established |
| House: complete republication notice and nonaffiliation | Met; §12 final subsection | Terms linked; no external legal conclusion offered |
| Program: saved document/reviews/result, local figures and input manifest | Met; `result.md` artifact table | Actual institutional authorization remains separate |
| Program: no raw reconstruction, model rerun, installs, source changes, publication or approval | Met; result execution/effects record | No such effect observed or claimed |

## Record 2 — Modeling choice register

Reason means the developer's stated rationale. A brief citation establishes that the developer made that choice; it does not independently prove the cited paper's contents or establish a theorem. All BRIEF locators refer to `inputs/model-brief.md` with public origin preserved in the manifest. “No alternative stated” is an explicit absence, not an invented rationale.

| ID / choice | Supplied reason | Alternatives considered | Limitation / evidence | Document locator |
|---|---|---|---|---|
| P01 Monotone-convex unameliorated B, collar, simultaneous fit | Forward projection shape; smaller short broken-date benefit; developer attributes recommendation to HW2008 | A raw/log-linear; QuantLib defaults and paper-setting bootstrap; cubic splines | Less local; sparse long-end dependence; default variant differs, paper-setting solver fails, cubic not built. BRIEF choice 1; RESULTS variants/shape/locality; CHECKS paired rows | §§3.2, 5.1–5.2, 7 |
| P02 Close-adjusted median prints | Remove intraday drift from the available free/date-specific public source; median robustness | Raw all-day or last-hours medians; licensed vendor/clearing close | One factor, interpolated loadings, execution estimates, common held-out adjustment; sole-source assertion is developer rationale. BRIEF 2; QUOTES | §4.2, §5.3 |
| P03 Eligible par-print filters, ≥5 prints and IQR ≤2 bp | Align fixed rate with instrument par rate; package/upfront/schedule differences matter | Include packages, convert other schedules, count floor 10 | Missing 9M/25Y; retained off-market prints; widest 20Y–30Y gap. BRIEF 3; QUOTES filter counts | §4.1, §5.2–5.3 |
| P04 Last-known-fixing O/N and T/N | Pin as-of-to-spot from a published available input | Start at spot; use later published as-of fixing | Assumes four-day constant fixing; ex-post match and small anchor sensitivity don't prove future behavior. BRIEF 4; FIXING | §§3.1, 5.2, 7.1, 9 |
| P05 OIS only, no futures | OIS span to 30Y from one retained report; missing futures data and convexity complexity | Short-end futures, developer's CME example | Limited front-end detail. BRIEF 5; external CME/HW assertion not independently checked | §§2, 5.2 |
| P06 Single curve | Deliberate scope; developer cites multi-curve literature for broader setting | Multi-curve framework | SOFR projection only. BRIEF 6 | §§1–3, 5.2, 9 |
| P07 Flat terminal instantaneous forward | Developer's HW2006 p.109 attribution; continuous last-node extension | Long-rate assumption; institutional extrapolation policy | No quote after last node, B's end rule determines held level; policy missing. BRIEF 7; RESULTS extrapolation | §§3.2, 5.2, 7.4, 8 |
| C01 B positivity collar | Developer states positivity for positive discrete forwards | No collar, built | Can flatten near nodes; no binding here; negative regime unassessed. BRIEF Smaller choices/Construction; RESULTS collar/variant | §§3.2, 5.1–5.2, 9 |
| C02 Simultaneous SciPy hybrid Powell system | B depends on neighboring intervals; sequential solver may fail | Iterate nodewise solves to fixed point | No convergence proof; loud residual failure >1e-11 rate. BRIEF Construction; HW-CODE bootstrap | §3.2, §5.2 |
| C03 A as transparent comparison | Developer attributes mistake-finding comparison recommendation to HW2008 p.73 | QuantLib default also measured; cubic not built | A is not demonstrated best competitor. BRIEF Construction | §5.2 |
| C04 Common QuantLib instrument dates | Conditional comparison isolates interpolation | Separate schedule code for B | Shared schedule defect remains; pricer agreement checks valuation not dates. BRIEF Construction; BOOT hw_instruments | §§3.1, 5.2, 7.2, 8 |
| C05 Last-payment-date nodes | Keep default helper pillar and common node set | Final accrual-end node | Quote does not independently pin final lag days; 1M node rates differ. BRIEF Construction; NODES | §3.1, §5.2 |
| C06 Act/365F time, continuous zeros | Developer's paper notation/common neutral time rationale | Act/360 time | Descriptive rates change with convention, DFs do not. BRIEF Construction | §3.1, §5.2 |
| Q01 Median regression | Match median quote treatment of off-market prints | Least squares; trimmed least squares | No least-squares comparison supplied. BRIEF Close adjustment; DERIVE QuantReg | §4.2, Table 3a |
| Q02 Path ≥2Y, loading one | Long-tenor fitted loadings near one; short tenors moved less | Separate loading each tenor | Ignores 0.97–1.08 movement differences. BRIEF Close adjustment; QUOTES loadings | §4.2, Table 3a |
| Q03 ≥8 prints in ≥5 hours to fit loading | Need time coverage to separate level from movement | Fit every tenor | 2M loading interpolated. BRIEF Close adjustment | §4.2, Table 3a |
| Q04 Loading clipped to [0,1] | Developer views negative or above-long loading as noise on this day | Unconstrained | Fitted 1M −0.04 replaced by zero. BRIEF Close adjustment; QUOTES | §4.2, Table 3a |
| Q05 Last traded hour close | Latest path-tenor observations | Fixed 15:00 close | Hour 16 is not a published fixing. BRIEF Close adjustment | §4.2, Table 3a |
| Q06 Forward-start expiration loading | Simple rule without forward-start model | Depend on start and length | 1Y starting in 1Y treated like 2Y. BRIEF Close adjustment | §4.2, Table 3a |
| S01 Four-calendar-day tenor matching | Absorb weekend/holiday adjustments | Exact-date matching; recompute adjusted maturity | Unusual near-tenor stub can match. BRIEF Print selection | §4.1, §5.2 |
| S02 New York execution day | Align close rather than UTC report date | Report UTC day | Evening print still in as-of date. BRIEF Print selection; DERIVE ts_ny | §4.1, §5.2 |
| S03 Adjusted IQR ≤2 bp | Wide spread signals less reliable median given tight liquid tenors | No dispersion screen | No tenor excluded by dispersion this date; 9M/25Y fail counts. BRIEF Print selection; QUOTES | §4.1, §5.2 |
| S04 Keep off-market prints | Avoid tuning a removal rule; median limits influence | Drop beyond 5 bp | Thin tenor more exposed; removal maximum quote movement 0.124 bp. BRIEF Print selection | §4.1, §5.2 |
| S05 A004 only | Instruments accrue Act/360 | Convert other codes | Excludes 47 of 4,258 rather than converting; CFTC definition unverified externally. BRIEF Print selection; QUOTES | §4.1, §5.2, §9 |
| T01 Location of B bump spillover | Neighboring-node formula transmits 5Y bump through 4Y into 3Y–4Y interval | No distinct alternative stated; explanatory mechanism | Largest outside-window response just after 3Y; conditional on this bump. BRIEF Tests; HW-CODE node formula | §5.1 |
| T02 5Y +1 bp window 4Y–6Y; 4.5Y cash flow | Mid-curve, 76 prints; developer attributes bump/cashflow examples to HW papers | All-input bump/stability norm | One forward bump and one cash flow; 10Y has 99 prints. BRIEF Tests; QUOTES | §5.1, §5.3, §7.4 |
| T03 Close-adjusted outside-input median absolute residual ≤1 bp | Tolerance exceeds observed adjusted IQRs | Maturity-specific tolerance | Judgment, not derived; common adjustment. BRIEF Tests | §5.3, §7.3 |
| T04 Late raw-print row | Reduce adjustment dependence | Last hour only; last three hours | Only 49 observations; base curves still adjusted. BRIEF Tests; VERIFY LATE_HOUR | §7.3 |
| T05 Paired sign test by bucket | Compare closeness on same prints without residual-size model | Paired t-test; Wilcoxon | Counts direction, not magnitude; multiplicity and repeated expirations. BRIEF Tests | §5.3, §7.3 |
| T06 Worst monthly-date USD 100m PV | Express economic disagreement in currency units | Actual portfolio PV difference | Single cash flow at worst date; no portfolio supplied. BRIEF Tests; RESULTS DF difference | §5.1, §5.3 |
| T07 Leave-one-tenor-out | Test influence of own prints on fitted path | Cross-validation across hours | Path tenors only, not fitted short loadings. BRIEF Tests; QUOTES leave-out | §4.2, §5.3 |
| T08 Numerical check tolerances | Rounding, printed precision, midpoint approximation, near-node continuity and daily-grid approximation set scales | Tighter/looser | Developer choices, not institutionally validated; specific units in Table 6. BRIEF Tests; CHECKS | §5.3, §7.1 |
| T09 p.102 capitalisation factors | Developer reads factor/forward as consistent and yield as a misprint | Printed yield would fail | Primary page image interpretation unchecked; report as developer account. BRIEF Tests; VERIFY paper_examples | §7.1 |
| T10 Payment lag zero, quarterly convention sensitivities | Two conventions developer considered incompletely fixed by prints | Roll/calendar sensitivities | Other convention variants not measured. BRIEF Tests; RESULTS convention_sensitivities | §7.4 |

## Record 3 — Citation audit

The audit below checks material relationships against the specified inputs, rather than treating a citation's presence as support. “Supported” means the particular asserted relationship is supported within the specified source scope. It does not mean the underlying market, fictional institution or external paper has been independently validated. “Unchecked” is reserved for claims whose original external source was not read; those claims appear in the document only as attributed developer statements. “Unavailable” applies to required actual institutional or original raw inputs not supplied. No material claim is marked supported merely because a matching number occurs somewhere.

### A. Model, data, choices and numerical relationships

| Audit ID / document locator | Material assertion checked | Source locator inspected and actual check | Status / scope |
|---|---|---|---|
| A01 Front matter, §1 | One curve discounts USD and projects compounded SOFR; B is developer preference | BRIEF Model and choice 1; HW-CODE Ois/ois_par_rate matches telescoping relation | Supported as implemented scope and developer preference |
| A02 Front matter, §5.1 | USD 227,458.20 is worst single-payment disagreement, not profit | RESULTS max_abs ×100,000,000 =227458.19731463946; GRID maximum-date row has A>B; checked arithmetic in checks.json | Supported; JSON precision used, CSV alone insufficient for cents |
| A03 §§1–2 | Fictional sandbox, no approved use, no real portfolio | INST Inventory rows and opening disclaimer read; USD100m explicitly unit exposure | Supported only as fictional scenario and absence in supplied evidence |
| A04 §2 | Monthly end 2056-09-18 differs from 30Y node 2056-09-26 | GRID last row, NODES 30Y, BOOT monthly construction | Supported; avoids conflating grids |
| A05 §3.1 | Deposit and payment-lag OIS relations, both day-count bases | HW-CODE Deposit/Ois/ois_par_rate and BOOT T/ql_helpers/hw_instruments read | Supported implemented equations; source-paper equivalence only attributed |
| A06 §3.1 | Known fixing 2026-09-17 3.85%, two deposits to spot | FIXING named fields; BRIEF convention table; BOOT deposit helper values | Supported frozen input identity; not new source-of-record fetch |
| A07 §3.1 | 1M node zero rates differ despite both fitting input | NODES 1M has 3.9407/3.9477%; par formula pay factor cancels for one period | Supported; not a claim that rounded DFs are equal |
| A08 §3.2 | A piecewise-flat forward and B neighbor-dependent sectors/integral | HW-CODE LogLinearDiscount, MonotoneConvex, _sector/_g/_G inspected directly | Supported as code account, not blanket theorem |
| A09 §3.2 | Solver tolerance and loud residual guard differ from house criterion | HW-CODE root call: tol 1e-14, maxfev20000, guard1e-11; HOUSE .01bp | Supported; dimensions kept separate |
| A10 §4.1 Tables 1–2 | Data categories, 23,563/4,258/663/529 counts and 523 used prints | QUOTES filter paragraphs; table used counts and excluded 4+2; DERIVE eligibility | Supported retained derivation account; raw archive not checked |
| A11 §4.1 | Upfront rate-rounding denominators are 1,661 and 2,390, distinct from spot population | QUOTES paragraph: total4051, packages included, 26%/3%; document preserves denominator | Supported; no false filtering denominator |
| A12 §4.1 | Other day-count code count47, outliers20, removal sensitivity.124bp | QUOTES code counts sum to4258; A0044211; outlier paragraph | Supported retained results; primary CFTC mapping separately unchecked |
| A13 §4.1 | Floor10 removes stated eight tenors, no dispersion exclusions | QUOTES count/IQR columns read across22 tenors; DERIVE eligibility both conditions | Supported; tenors9M/25Y fail count, not spread |
| A14 §4.2 | Adjustment sign, median fit and common long-tenor path | DERIVE close_adj = fixed − loading×path; QuantReg q=.5; QUOTES negative path to0 | Supported; “about7bp rise” not treated as monotonic |
| A15 §4.2/Table3a | Fitted/interpolated/clipped loading distinctions and limits | QUOTES loading rows; DERIVE loading_for, min-print/hour conditions | Supported; forward-start rule uses time to expiration |
| A16 §4.2 | IQR narrowing, raw 4Y inversion, .165bp leave-out maximum | QUOTES full tenor and leave-out tables; 100×(4.6206−4.5612)=5.94bp | Supported rounded description; in-sample caveat retained |
| A17 §5 Table3 | Forward maximum jumps59.71/1.68, five-day59.73/5.34, counts18/0, variation421.3/365.0 | RESULTS forward_smoothness/forward_total_variation; BOOT formula differences and grid to last pillar | Supported; absolute successive changes distinct from variation |
| A18 §5.1 | Continuous B still has flat stretches and steep three-day ramp | RESULTS forward_after_3Y_node, sector counts; HW-CODE sectors ii/iii; retained figure visually inspected | Supported; continuity not equated to universal smoothness |
| A19 §5.1 | 5Y outside-window magnitude and absolute PV sensitivity share | BOOT comparisons use d<4Y or d>6Y; inside inclusive; ladder denominator sums absolute changes; RESULTS locality/ladder fields | Supported; excludes netting and wrong interval endpoints |
| A20 §5.1 | 5Y76 prints not most traded; 10Y99 | QUOTES rows compared, rather than reading single count | Supported ranking; wording corrected against possible false “most traded” claim |
| A21 §5.1 | B long-end forward4.6127→3.9957; average4.2013; ~20.6bp/67bp gaps | RESULTS table; BOOT slice days[i20:i_last] excludes30Y node; QUOTES30Y4.6637 | Supported rounded arithmetic and exact population; no mixing with discreteAct365F forward |
| A22 §5.2 | QuantLib paper-setting failed; default blend differs; no-collar same on snapshot | RESULTS interpolation_variants states failed at5th/3M, maxDF8.01175e-5, jump20.556; collar zero nodes | Supported retained outcomes; failure not attributed to all B solvers |
| A23 §5.2–5.3 | Reasons/alternatives/limits for34 choices | BRIEF seven numbered choices and all Smaller choices bullets compared with Record2/document locators | Supported attribution; no alternative invented for T01 |
| A24 §6 Table4 | Historical execution times and versions | All5 environment JSON receipts read, script hashes compared to fetched code; run_info versions agree | Supported receipt statements; not current runs or deployment |
| A25 §6 | Raw reproduction unavailable from selective bundle | WE exclusions; original-raw-input-hashes; PROVENANCE inventory read | Supported scope limitation; raw acquisition not attempted |
| A26 Table5 | All22 instruments satisfy .01bp, including max errors | Current code checked every A/B error column;22 rows; max3.86e-9/2.755e-10 from roundedCSV; RESULTS full precision agrees | Supported current arithmetic, historical model calculation only |
| A27 Table6 | Historical paper, interval, positivity, continuity and roundtrip results | CHECKS all named entries read against verify.py check definitions and tolerances | Supported retained tests; primary paper text remains unchecked |
| A28 §7.2 Table7 | A comparison max/mean and exact business-day bounds | All7,482 CSV row differences recomputed; max1.9251267247000214e-13, mean1.9170748626330293e-14; first/last dates checked | Supported independent arithmetic of retained series; no model reproduction |
| A29 §7.2 | Common schedules limit independence; B interpolation uses same nodes | IMPL limits and summarize_implementations; VERIFY two_implementations | Supported scope; no independent institutional validation claim |
| A30 §7.3/Table8 | Signed residual order,983 cohorts523/140/320, heldout460 | RESIDUALS groups counted; VERIFY out_of_sample code and CHECKS summaries | Supported; input-tenor prints excluded from heldout |
| A31 §7.3 | Historical median upper-middle rule, A.459942/B.404662, both<1bp | VERIFY med_a/med_b code; frozen-row arithmetic reproduces B JSON0.4046620120479877 and A0.45994244105472537 | Supported; ordinary medians.458535/.404222 explicitly separate |
| A32 Table8 | Signed medians/quartiles differ from conventional interpolated quantiles | VERIFY stats index int(p*(n−1)+.5) inspected; table copied from historical verify.md | Supported historical convention; no silent substitution |
| A33 Table9 | Paired counts/p-values/absolute-difference medians and cohorts | CHECKS paired rows; VERIFY boundary lo<years≤hi, tolerance1e-9bp and upper-middle medians | Supported retained comparison; “under1Y” code endpoint explained |
| A34 §7.3 | 30/11 over41 versus16/6 over22 dates, p.004 vs.052; exploratory6M–1Y | CHECKS short, per-expiration, raw and exploratory rows inspected; BRIEF calls exploratory | Supported; no independence or universal superiority inference |
| A35 §7.3 | B269/A189 with2 ties and aggregate+.007bp | CHECKS all-heldout n460 counts, difference.0074033; arithmetic460−269−189=2 | Supported; count superiority not equated with economic materiality |
| A36 §7.3 | Late49 raw signed medians−.76/−.47, limited adjustment | CHECKS late row; VERIFY raw conversion and LATE_HOUR15; QUOTES hour15−.2/hour16zero | Supported; base curves still adjustment-dependent |
| A37 Table10 | Six distance bins, medians/IQR, exact distance from OIS pillars | Recomputed absolute calendar-distance bins and inclusive quartiles from RESIDUALS/NODES; all match DISTANCE JSON within1e-12 | Supported current arithmetic; no financial model rerun |
| A38 §7.3 | B median worse in nearest forward-start bin | Table10 forward0–30: B.598>A.578 despite aggregate B lower | Supported relational check; distance not independent validation/risk |
| A39 §7.4 Table11 | Convention scenario maxima versus baseline, annual and daycount exceed A/B difference | RESULTS convention_sensitivities fields; BOOT variants inspect scaling365/360; compare maxima with.00227458 | Supported scenario comparison; does not identify correct convention |
| A40 Figures1–4 | Axes/units, historical plots and data relationship | All4PNG visually inspected; BOOT plotting calls and data CSV headers read; source files byte-matched and copied | Supported; first figure excludes time-zero point in plotting, per code |
| A41 §7.4 | 4.5Y cashflow PV difference sign and final-node rounding | RESULTS off_node A−B=−269.4566; final jumpA≈8e-10bp/B0; 40Y DF/extrapolation fields | Supported; rounded0 not asserted exact |
| A42 §8/§9 | Every major limitation has proposed control or explicit absent dedicated control | BRIEF limits compared with INST C01–C11 and documented event-step absence | Supported as simulated proposals; no operational evidence |
| A43 §10 | Fictional roles, inventory/tier, findings, pending committee/unsigned record | INST inventory, HER-CR-001, F01–F05 and Approval sections read | Supported as supplied fiction only |
| A44 §11 | Versions/customization/shared dependencies/MIT distinction | Receipt package fields, imports/calls and WE license paragraph; LICENSE read | Supported historical identity and licensing-scope description |
| A45 §12 | Complete SOFR notice and nonaffiliation retained | Exact source-register notice compared with document subsection | Supported verbatim provider notice; not an independent legal analysis |

### B. Unchecked external source claims and unavailable evidence

| Item | Original source/locator from supplied register | Status and document treatment |
|---|---|---|
| HW2008 recommendation, comparison advice, positivity theorem and cubic-spline critique | §6/6.1, pp.72–79, Table1 | Unchecked external contents: primary PDF not retrieved. Document identifies code behavior directly and attributes developer rationale to BRIEF; no paper title treated as proof |
| HW2006 par-equation provenance, examples/misprint, extrapolation and hedge/stability motivation | pp.92,102,107,109,124–126 | Unchecked external contents. Retained tests and developer interpretation reported with source scope; alleged paper misprint not independently established |
| ARRC compounding/calendar/payment-delay description | pp.5,17,31,35 | Unchecked external contents. Implemented conventions supported by bundle code/brief; no external convention conformity certified |
| CFTC A004=Actual360 historical definition | Historical specification AppendixC p.55 | Unchecked external contents. Described as brief attribution; data-code counts supported separately |
| DTCC guide's UTC report-day account | Guidev2.8 p.7 | Unchecked external contents. Developer timing rationale attributed; actual code converts timestamps to New York |
| CME planned transition and futures-convexity example | August2020 §§1,1.3,2.1 | Unchecked external contents, historical-plan attribution only; no completed-transition/current-market claim |
| AB2013 multi-curve background | Abstract-level registered reference | Unchecked external contents; no implementation claim derived from title |
| QuantLib defaults and endpoint-only positivity | Version1.43 external source class/update | Unchecked external original source. Bundle helper arguments/code and retained comparison claims examined; default/endpoint rationale attributed to BRIEF/CHECKS |
| Original raw DTCC/NYFed inputs, raw extracted trade tables | RAW-HASHES and public origin links | Unavailable in selective bundle; not fetched/reconstructed. Prevents complete offline raw-data reproduction |
| Licensed benchmark and actual institution records | INST required-real-world-gaps | Unavailable:14 document markers preserve actual slots; simulated dossier cannot fill them |
| Broader dates, negative rates, stress regimes, full risk-system validation | BRIEF limitations; INST F02/C07/C08 | Unavailable; current documentary/arithmetic scope does not create evidence |

No unresolved material contradiction is concealed. The nearest-index median convention explains the small difference between historical median displays and ordinary two-middle medians. The convenient kernel alias and selected pin have identical bytes and release. The source brief's corrected “approximately0.40bp” aligns with retained0.404662bp; the document does not repeat an unsupported0.41bp result. Numerical equality from rounded display values is not asserted.

## Record 4 — Readiness report

**Exact document SHA-256:** `16d7918bc07263bef7f90b9ed2aa9f407ff01a3a28da4c3085c081f87ddca2bf`  
**Document:** `document.md`, demo-v1, 2026-10-12; **as-of:** 2026-09-18.  
**Selected inputs:** revision `01ae3bdac6f61ecbabcd72e0116ebb2860e93157`; dossier HER-scenario-1; kernel/Quant release 0.4.6-rc.1. `input-manifest.json` records every retained input origin, final URL, acquisition time and SHA-256. `selected-dependencies.json` declares six agreement hashes; each matched. Public bundle files used here matched `SHA256SUMS` where declared.

**Judgment:** The commissioned documentation scope is fulfilled at executor readiness level: a substantive twelve-section document, every selected requirement mapped, all 34 developer choice entries retained, a performed citation audit, exact artifact identification and an honest result record. The document is ready for a separate assessment of these exact bytes. This is not a statement that actual institutional evidence is complete, that the model is validated, or that any real use is authorized.

Findings and remaining work:

1. **Material evidence limit:** One historical positive-rate snapshot and selected developer tests do not establish general validity, negative-rate behavior, stressed performance, long-end truth or comprehensive hedging stability. A separately authorized, retained multi-date/stress/risk campaign and independent assessment remain needed for broader reliance.
2. **Benchmark and input limit:** Derived public execution estimates share an intraday adjustment with outside-input comparisons. No licensed independent benchmark or real feed lineage/tolerance exists in supplied evidence. Original raw inputs are omitted; full raw-data reproduction is not possible from this selective offline bundle.
3. **Implementation limit:** A's7,482-point implementation agreement is strong arithmetic agreement conditional on shared inputs/dates. It does not remove common calendar/convention defects. Separate schedule reconstruction remains needed.
4. **External-source limit:** Original papers and external versioned code were not retrieved; their claims remain expressly attributed to the brief or historical check report, with unchecked status in Record3. A later assessment may obtain and verify the primary sources, particularly the paper-example interpretation and collar/convention assertions. The document does not present these as independently established facts.
5. **Institutional limit:** All dossier facts and controls are fictional. The 14 marked actual evidence gaps remain substantive absences: inventory/tier; internal feed/entitlements; exposure/use inventory; real roles/authorization; prior validation; approval conditions; operating monitoring; production conventions; benchmark/tolerance; close/stale policy; extrapolation policy; risk bump methodology; implementation ticket; signatures. Their disclosure is complete; their supply is not.
6. **Assessment limit:** This readiness report was prepared by the executor that wrote the document. Independent assessment and any institutional authorization are separate and pending. The fictional F-05 record is not retroactively closed on behalf of its named roles.

**Checks actually performed now:** Pinned/bundle byte verification; all 22 A/B error-column comparisons with 0.01 bp; monthly-grid row/maximum/date inspection and full-precision USD conversion; all 7,482 retained implementation-difference calculations plus mean/max; held-out cohort counts and upper-middle median check; all six distance-bin summaries recomputed from frozen rows with inclusive quartiles; source-function inspection; visual inspection of four frozen figures; requirement/choice/claim mapping and artifact-structure checks. These are document readiness and arithmetic checks, not financial-model reruns. Machine-readable arithmetic evidence is in `checks.json`; the current checking script imports no financial-model modules.

**No new source/model run:** Original `derive_quotes.py`, `bootstrap.py`, `hagan_west.py`, `verify.py` and other supplied scripts were inspected only. Historical stage receipts were not relabeled as current execution. No actual institutional approval, deployment, independent review, trading, publication, package installation or source amendment occurred.
