LAYER 1 · ARIN22 DATA ROOM · INSTITUTIONAL EVIDENCE PACK

ARIN22 – deterministic kernel for institutional risk

A read-only replay surface for the sealed ARIN22 evidence lane. Select a frozen run, inspect input → output → hash → lineage, and export the same status-bound object the page renders.

–Canon freeze

8×H100Substrate + CPU mirror

1 HASH1000 replay repeats

PRE-CLIENTDiligence-ready status

pre-client institutional diligence-ready, not enterprise production-approved · no public compute API · row-level numerics under NDA.

Replay the sealed run → See methodology →

Replay packet

Status, hash and command in one object.

The sealed run identity and verification command, without rerunning the model.

Replay console

Sealed run replay, not browser compute.

Choose a sealed run. Every visible value travels with status, metric family and as-of date. Withheld and pilot-deferred cells do not print numbers.

–

REVIEW_ONLY

Full selected run is rendered above. Animate steps through the evidence path for review.

Evidence corpus

Evidence that carries its limits.

Measured envelopes and lineage, separated by lane. Protected construction and row-level client numerics remain under NDA.

Regulatory source: Federal Reserve SR 26-2, Apr 17, 2026. References in the corpus and frozen replay describe evidence mapping, not regulatory certification, validation or endorsement of SAA.

run_idPrimary public evidence object.enterprise_wave_v2enterprise-canon-h100-freeze-2026-05-30supersedes 598e7034

inputFrozen public scope for the canonical pack.8,800 cases/backend · 8.8 B paths/backend8×H100 80 GiB + CPU mirror

hashPublished full fixture hash plus archive prefix for the GPU wave.canonical_sha256 ae8a2e11ee2d1466c9fa4483f68ab276d8ce909629e3bcf7c4b18fc5a3ebaae6archive_sha256_prefix b8193ba1c9e2d7f1...

replayCommand shape exposed; full artifact bundle released under mutual NDA.sha256sum -c 04_SHA256SUMSreplay_report --run enterprise_wave_v2 --manifest enterprise-canon-h100-freeze-2026-05-30

Chapter ICanonical evidenceEnterprise Wave v2, production 10M lane, parity, latency, stress-surface, STAC-inspired GPU lanes and August 2026 acceptance/replay lanes.MEASURED

Canonical rows separate the current public headline from lineage and STAC-inspired self-run lanes. Where earlier 24M rows are retained, the row says so; current tick/gate lane language remains scoped.

I-01 to I-09 – canonical scale and parity
#	Block	Public cells	Status	Scope / boundary
I-01	Enterprise Wave v2	8,800 cases/backend · 8.8 B paths/backend · 8/8 H100 80 GiB · CPU mirror · 0 execution failures	MEASURED	Public canonical pack · commit 141d7d78 · supersedes 598e7034 · May 2026
I-02	Production lane 10M	14,400 GPU+CPU cases · 72 B paths/backend · 10M/case production-default lane	MEASURED	Internal Market Readiness canon surfaced as separate lane, not merged with Enterprise Wave v2
I-03	CRN 99.9 parity	100 worst-tail cases · 100 M paths/backend · PASS · regression y = 1.0000x	MEASURED	CRN anchored 99.9 tail parity
I-04	MC challenger	50 cases · 200 reference runs · 2.0 B paths · CVaR99.9 max 0.035% · p95 0.015%	MEASURED	10M Monte Carlo challenger lane
I-05	GPU/CPU parity bands	CVaR99.9 p95 0.27-0.28% · max 0.59-0.67% · max observed 0.665%	MEASURED	10M GPU+CPU parity lane · banded disclosure
I-06a	Tick replay · current canon	10M updates · p99 0.44-0.46 ms · p999 0.653 ms · stale/missed 0/0	MEASURED	Current headline lane · 2026-05-30 canon
I-06b	Tick replay · retained row	24M updates · 4.032 B evals	V1 LINEAGE	Retained as historical lineage; not the current tick headline
I-07a	Pre-trade gate · current canon	50M orders · ALLOW 9,999,675 · REVIEW 40,000,124 · BLOCK 201 · p99 0.85-0.93 ms · p999 1.8-2.5 ms	MEASURED	Current gate lane · 2026-05-30 canon
I-07b	Pre-trade gate · retained row	24M orders · ALLOW 21.32M / REVIEW 2.64M / BLOCK 38.5K	V1 LINEAGE	Historical/public lineage row; different stress mix, not current gate headline
I-08	Determinism	1,000 fresh-process repeats · 1 unique hash · 0.0% max metric diff	MEASURED	Fresh-process replay proof
I-09	Subject Stress Surface	10,080 rows · 60 subjects × 28 scenarios × 6 horizons · approx 221.8 B paths · 1M/10M tier · 4 seeds · PRNG/Sobol/LHS	MEASURED	Subject stress surface · row-level numerics under NDA

I-10 to I-18 – replay, footprint and GPU/STAC-inspired lanes
#	Block	Public cells	Status	Scope / boundary
I-10	Crypto shock replay	5 shocks (LUNA 2022 · FTX 2022 · USDC depeg 2023 · BTC ETF 2024 · ETH ETF 2024) × 4 assets = 20 hash-sealed packets · 0 lookahead violations · 0 narrative-readback violations	MEASURED	Crypto ex-ante replay discipline; distinct from crisis replay lane
I-11	Shadow Ledger	BACKTEST_HARNESS_ARMED · 417 / 250 approved matched observations · current as-of 2026-07-13 · PASS not claimed until review/statistical gates clear	PENDING	Gate, not policy; current facts computed from realized-backtest ledger status
I-12	Total footprint	approx 27,700 distinct cases · 48M intraday events · approx 239.4B paths	REVIEW_ONLY	Footprint evidence only, not marketing dollar claim
I-13	Definitions	case = one portfolio/scenario/horizon/backend unit · CLOSED = completed with artifact · PASS = met declared gate · path budget decomposed by lane	REVIEW_ONLY	Definitions prevent numerator/denominator drift
I-14	CPU validation surface	~99 microseconds at D=4 · ~0.3 ms at D=200 · q99.9 deviation < 0.05%	MEASURED	CPU-bound correctness posture · no GPU lock-in claim boundary
I-15	STAC-A2-inspired	310M paths · 2.48B valuation paths · 595.20B path-asset-step ops · 14.875s · 166.72M valuation-paths/s · scaled lane p95 2.709s	MEASURED	Evidence Wave 2026-05-20 · self-run · NOT STAC-audited
I-16a	Greek values, published	Delta -0.37487705 · Gamma 0.04816398	MEASURED	Published values retain source precision; no page-side rounding
I-16b	Greek values, withheld	Price · Vega · Rho · Theta	WITHHELD	Reason: pending listed-option parity; full Greek table under NDA
I-17	Companion lanes	STAC-M3-inspired 60M tick updates · p99 0.501 ms · p999 0.626 ms. STAC-T1-adjacent 60M evaluations · p99 1.110 ms. Fail-closed 60M BLOCK · p99 0.864 ms.	MEASURED	Separate STAC-inspired wave dated 2026-05-20; not the 2026-05-30 canon tick/gate lane
I-18	What this lane is NOT	Not STAC-certified · not external STAC audit · not wire-to-wire order latency · not proof that GPUs are required for deterministic correctness	NOT_CLAIMED	Boundary row travels with the lane

I-19 to I-24 – August 2026 acceptance and replay lanes
#	Block	Public cells	Status	Scope / boundary
I-19	Estimated acceptance battery	160/160 cells within 5% error envelope vs 100M-path reference · 40 cases × 10 market regimes × 4 tail metrics · median \|error\| 0.0156% · p95 0.951% · worst cell 2.717% · material anti-conservative rows 1/160	REVIEW_ONLY	Parameters estimated from independent calibration samples and hash-frozen before every oracle join · every intermediate FAIL published in the same bundle · pending production re-run · as of Aug 23, 2026
I-20	Acceptance iteration ledger	12 iterations V1→V12b · 31/160 → 160/160 · median error 11.82% → 0.0156% · three construction defects found via mathematical identity tests and removed at source · no gate relaxed in response to a failure	REVIEW_ONLY	Full FAIL/PASS chain preserved in evidence bundle · definitions changed only pre-declared, with both old and new counters published · as of Aug 23, 2026
I-21	100M-path reference corpus	40 oracle cases · 100M Monte Carlo paths per case · per-case evidence sha256 recorded in every result row · seeds recorded — reproducible by seed	MEASURED	Reference-truth lane backing I-19 · raw path chunks historic; evidence records and hashes retained locally · as of Aug 23, 2026
I-22	Crisis replay · 75 cells	5 events (GFC 2008 · Q4 2018 · COVID 2020 · rates 2022 · SVB 2023) × 15 books · like-for-like vs each method’s own forecasts: CVaR99 covered realized losses 73/75 (historical rolling 23/75 · Student-t MC 55/75 · parametric normal 47/75) · VaR99 breaches 12/75 vs 54/75	MEASURED	Forecast cutoffs pre-event · no-lookahead flag recorded per row · horizon-matched comparison, same metric on both sides · as of Aug 23, 2026
I-23	Forward-locked forecast packets	Forecasts hash-sealed (SHA-256) before outcomes exist · mature T+1 / T+5 / T+20 trading days from sealed timestamps	PENDING	No results claimed until maturity · packets cannot be edited post-hoc by construction · as of Aug 23, 2026
I-24	Moment-identity release gates	Second- and third-moment identity checks vs closed forms · max component deviation 1.11e-15 (machine precision) · permanent release gate	MEASURED	Identity harness runs pre-release; a kernel build that fails moment identities cannot publish · as of Aug 23, 2026

Chapter IIMode catalogueTwenty-two modes grouped into eight A-H layers. Public taxonomy is visible; protected thresholds remain NDA-bound.REVIEW_ONLY

II-01 – 22-mode catalogue grouped A-H
#	Block	Public cells	Status	Scope / boundary
II-01A	A · Input and data quality	Missingness, stale data, schema conformance, source confidence	REVIEW_ONLY	Public surface only; weights and thresholds under NDA
II-01B	B · Portfolio state	Exposure map, concentration, liquidity, horizon ladder	REVIEW_ONLY	Mode names public; construction protected
II-01C	C · Scenario and stress	Historical replay, shock ladder, cascade route, stress horizon	REVIEW_ONLY	Reviewer-facing taxonomy
II-01D	D · Tail and curvature	Deep-tail promotion, EVT watch, curvature trigger	REVIEW_ONLY	Protected construction dossier under NDA
II-01E	E · Audit and lineage	Input hash, output hash, replay command, superseded lineage	REVIEW_ONLY	Public-safe proof path
II-01F	F · Governor and refusal	PASS/WARN/FAIL/INCONCLUSIVE, WATCH_SUPPRESSED, fail-closed gate	REVIEW_ONLY	Decision policy visible; thresholds withheld
II-01G	G · Deployment and tenancy	API hand-off, side-car, full application, tenant isolation	REVIEW_ONLY	Commercial deployment mode, not a metric
II-01H	H · Regulatory and claim posture	SR 26-2 mapping, DORA, EU AI Act Article 12, STAC pending, SOC 2 pending	REVIEW_ONLY	Alignment/readiness, not certification

II-02 and II-03 – service mapping and IP boundary
#	Block	Public cells	Status	Scope / boundary
II-02	Service × Mode matrix	Risk Analyzer: A/B/C/D/E/F · Global Risk: A/C/E/F/H · Investment: A/B/C/E/F · Digital Assets: A/B/E/F/H · News: A/C/E/F/H · KOKON: A/E/F/G/H	REVIEW_ONLY	Six services mapped to mode families; exact service-mode table under NDA
II-03	IP boundary	NDA positions: construction weights · routing thresholds · row-level numerics · protected convergence proof details · customer deployment manifests	NDA	Public page names surfaces and outcomes, not protected construction

Chapter IIILive kernel replayThe console renders sealed run objects, industry benchmark rows and the compute surface from recorded artifacts.MEASURED

III-01 – sealed run replay source
#	Block	Public cells	Status	Scope / boundary
III-01	Replay console	The full selected run above renders the current sealed-run cells, status tags, disclosure tags, as-of dates and replay commands from the same object.	REVIEW_ONLY	Single source for replay cells; no hand-copied metric row.

III-02 – industry benchmark
#	Block	Public cells	Status	Scope / boundary
III-02.1	Parametric VaR	0.08 ms · $429,801 · -68.7%	MEASURED	Same portfolio benchmark row
III-02.2	Historical VaR	0.57 ms · $288,491 · -79.0%	MEASURED	Same portfolio benchmark row
III-02.3	Monte Carlo, 10M paths	690,854 ms · $1,373,524 · baseline	MEASURED	10M reference truth
III-02.4	SAA deterministic kernel	1.12 ms · $1,370,843 · -0.19%	MEASURED	Single-portfolio challenger lane; not universalized
III-02.5	Extended validation Aug 2026	160/160 cells within 5% envelope vs 100M-path reference across 10 regimes · median \|error\| 0.0156%	REVIEW_ONLY	Cross-reference I-19 · not universalized to production routes until re-run · as of Aug 23, 2026

III-03 and III-04 – compute surface and coverage
#	Block	Public cells	Status	Scope / boundary
III-03	Compute surface	Route: input → mode family → deterministic kernel → output ladder → hash chain · confidence ladder · cross-validation · latency breakdown	REVIEW_ONLY	Concrete route, not a public compute API
III-04	Coverage matrix	28 scenarios × 420 runs · scenario/horizon tier coverage · row-level surfaces under NDA	REVIEW_ONLY	Coverage grid visible; row-level metrics withheld

Chapter IVV1 lineageSuperseded campaigns remain visible for audit continuity and are not promoted as current headline claims.V1 LINEAGE

IV-01 to IV-07 – lineage register
#	Block	Public cells	Status	Scope / boundary
IV-01	Phase-5	47,544 paired rows · 283 subjects · seven-metric parity table · 3.53x · 9.94 ms vs 35.06 ms	V1 LINEAGE	Closed lineage, not current headline
IV-02	H100 Direct Grid	22,400 cases · 22.40 B paths · 12.5678 s · powered tail gate	V1 LINEAGE	D=200 production-dimension sweep lineage
IV-03	Production Sweep	7,203 portfolios · approx 805.7 B paths · parity coverage map · horizon coverage	V1 LINEAGE	Production sweep lineage
IV-04	8-Wave	1.06 T paths · 1,043,030 cases · 8-wave status matrix · 7 PASS + 1 WATCH_SUPPRESSED · Shadow Ledger ARMED	V1 LINEAGE	Superseded as headline; preserved as substrate breadth
IV-05	MC Spot-Check	64 cases × 4 seeds × 256 runs · 2.56 B paths · VaR99.9 diff 0% · CVaR99.9 max diff 0.00001135%	V1 LINEAGE	Spot-check lineage
IV-06	Runs 9-12	Determinism · saturation envelope · regime routing · LocalGate calibration 92.46% → 73.39% NAV · post-calibration MC audit max abs 0.2822%	V1 LINEAGE	Run 11/12 calibration lineage · commit 3717fc4d
IV-07	EVT backtest	Kupiec 95% p=1.0000 · Kupiec 99% p=0.5030 · Christoffersen p=0.4246 / 0.7937 · Acerbi-Szekely / AS-ES p=0.3284 / 0.2015 / 0.2869 · PIT diagnostics · WARN-margin on regulatory levels · realized PASS structurally gated before 250 matched observations	V1 LINEAGE	Backtest harness lineage; no realized regulatory PASS

Chapter VGovernanceDecision-integrity invariants, audit engine, regulatory crosswalk, claim matrix and reproducibility lineage.REVIEW_ONLY

V-01 to V-03 – controls and regulatory crosswalk
#	Block	Public cells	Status	Scope / boundary
V-01	5 Decision-Integrity Invariants	Data Immutability · Append-only Verdicts · No Backflow · Single Exit · Data Freshness	REVIEW_ONLY	Canonical structural invariants; public names preserved from prior page version
V-02	7-rule Audit Engine	Cascade Amplification · SPOF Fabrication · Regime Fabrication · Severity Escalation · Downplay in Critical · Polarity Inversion · Cascade Through Zero	REVIEW_ONLY	Canonical rule labels; implementation detail under NDA
V-03.1	SR 26-2	Mapped to deterministic model outputs only; agentic/generative narrative is not claimed SR 26-2 validated	REVIEW_ONLY	Readiness/alignment, not certification
V-03.2	Basel III / FRTB	VaR/CVaR/ES, traffic-light backtesting and model-risk documentation mapping	REVIEW_ONLY	Not a regulatory capital approval
V-03.3	DORA	Evidence trails, resilience posture and third-party risk mapping	REVIEW_ONLY	Readiness mapping
V-03.4	EU AI Act Article 12	Record-keeping and traceability for AI/narrative surfaces	REVIEW_ONLY	Correct scope for AI systems
V-03.5	SOC 2 Type II	Independent audit planned / in progress; no Type II report issued yet	PENDING	No attestation claimed
V-03.6	STAC	STAC-inspired self-run lanes; independent STAC benchmarking pending	PENDING	Not STAC-certified

V-04 – Claim Status Matrix (selected public rows)
#	Block	Public cells	Status	Scope / boundary
V-04.01	Enterprise Wave v2 scale	8,800 cases/backend · 8.8 B paths/backend · 0 failures	MEASURED	Claimed
V-04.02	10M production lane	14,400 GPU+CPU cases · 72 B paths/backend	MEASURED	Claimed as separate lane
V-04.03	CVaR99.9 max vs 10M MC	0.035%	MEASURED	Claimed within lane
V-04.04	GPU/CPU parity bands	p95 0.27-0.28% · max 0.59-0.67%	MEASURED	Claimed as band
V-04.05	Tick/gate latency	p99 bands disclosed, internal service compute only	MEASURED	Claimed with disclaimer
V-04.06	STAC certification	Not STAC-certified	NOT_CLAIMED	Explicitly not claimed
V-04.07	Realized regulatory backtest PASS	417 / 250 approved matched observations · PASS still pending review/statistical gates	PENDING	Not claimed · current as-of 2026-07-13
V-04.08	SOC 2 Type II	No report issued yet	PENDING	Not claimed
V-04.09	Production customer deployment	Pilot required	PILOT_DEFERRED	Not claimed
V-04.10	Investment advice / execution authority	No investment advice · no execution authority	NOT_CLAIMED	Claim boundary
V-04.11	Agentic AI numerical control	AI narrative only; numbers deterministic	NOT_CLAIMED	No AI-in-numbers claim breach
V-04.12	Protected construction disclosure	Weights, thresholds, row-level numerics under NDA	NDA	Disclosure boundary

V-05 to V-10 – replay lineage and limits
#	Block	Public cells	Status	Scope / boundary
V-05.01	Reproducibility Lineage · current freeze	run_id enterprise_wave_v2 · generated_at 2026-07-28 · commit 141d7d78 · tag enterprise-canon-h100-freeze-2026-05-30 · supersedes 598e7034 · canonical_sha256 fixture ae8a2e11ee2d1466c9fa4483f68ab276d8ce909629e3bcf7c4b18fc5a3ebaae6 · command sha256sum -c 04_SHA256SUMS	REVIEW_ONLY	Current public fixture spine; full enterprise manifest under NDA
V-05.02	Phase-5 lineage	47,544 paired rows · 283 subjects · seven-metric parity table · 3.53x · 9.94 ms vs 35.06 ms	V1 LINEAGE	Closed lineage, not current headline
V-05.03	H100 Direct Grid lineage	D=200 production-dimension sweep · 22,400 cases · 22.40 B paths · 12.5678 s · powered tail gate	V1 LINEAGE	Published substrate lineage, not STAC certification
V-05.04	Production Sweep lineage	7,203 portfolios · approx 805.7 B paths · parity coverage map · horizon coverage	V1 LINEAGE	Production sweep lineage
V-05.05	8-Wave lineage	commit 3b65f634 · supersedes daf1f824 · 1.06 T paths · 1,043,030 GPU/CPU cases · 7 PASS + 1 WATCH_SUPPRESSED · Shadow Ledger ARMED	V1 LINEAGE	Superseded as headline; preserved as substrate breadth
V-05.06	MC Spot-Check lineage	run_id h100_mc_spot_deep_tail_64x4_10m_v2 · 64 selected deep-tail cases · seeds 1, 7, 42, 2026 · 256 reference runs · 2.56 B paths · VaR99.9 diff 0% · CVaR99.9 max diff 0.00001135%	V1 LINEAGE	Spot-check lineage; not full-population MC reference
V-05.07	Runs 9-12 lineage	commit 3717fc4d · determinism · saturation envelope · regime routing · LocalGate calibration 92.46% → 73.39% NAV · post-calibration MC audit max abs 0.2822%	V1 LINEAGE	Run 11/12 calibration lineage
V-05.08	Shadow Ledger / EVT lineage	BACKTEST_HARNESS_ARMED · hash chain VALID · realized PASS structurally forbidden before 250 matched observations · Kupiec / Christoffersen / Acerbi-Szekely / PIT diagnostics retained	V1 LINEAGE	Backtest harness lineage; no realized regulatory PASS
V-06	Software manifest	commit · build · Python · CUDA · CuPy · NumPy · driver · OS · seeds captured; exact environment bundle released under mutual NDA	NDA	No fake public environment values
V-07	Latency glossary	~99 microseconds CPU D=4 · ~0.3 ms D=200 · 0.44-0.46 ms tick p99 · 0.85-0.93 ms gate p99 · 1.12 ms benchmark row · 14.875s STAC-A2-inspired lane · internal compute latency only	REVIEW_ONLY	13-row glossary retained in NDA pack; public key rows visible
V-08	IP Disclosure	Public: input/output/hash/lineage, aggregate bands, status vocabulary. NDA: construction, thresholds, row-level numerics, convergence detail, customer manifests. Patent: protected construction surfaces as filed/filing strategy dictates.	REVIEW_ONLY	Nine-row IP boundary summarized
V-09	Compliance posture	SR 26-2 mapped · EU AI Act Art. 12 mapped · DORA mapped · SOC 2 pending · STAC pending · customer compliance remains customer responsibility	REVIEW_ONLY	Readiness, not attestation
V-10	Honest limitations	No STAC certification · no SOC 2 Type II report · no live-exchange wire-to-wire measurement · no realized regulatory PASS until review/statistical gates clear · no production approval before paid pilot · no row-level public client numerics	NOT_CLAIMED	Self-disclosure block

V-04.13 to V-04.16 – Claim Status Matrix continuation · August 2026
#	Block	Public cells	Status	Scope / boundary
V-04.13	Estimated acceptance battery	160/160 within 5% vs 100M-path reference	REVIEW_ONLY	Claimed as diligence evidence, not as production accuracy attestation · as of Aug 23, 2026
V-04.14	Production accuracy authority from battery	Battery re-run on productionized backend routes	PENDING	Not claimed until green on production paths · as of Aug 23, 2026
V-04.15	Fail-closed publishability guard	Production-installed guard for non-validated model routes · negative-path tests passing in production tree · candidate→production SHA chain	REVIEW_ONLY	Gate internals, names and thresholds under NDA · as of Aug 23, 2026
V-04.16	Acceptance gate contract v2	Dual-counter sign gate (raw + material anti-conservative, both published) · strict ES>VaR distinctness invariant per row	REVIEW_ONLY	Contract changes pre-declared; both counters retained publicly · as of Aug 23, 2026

V-05.09 and V-10 – acceptance lineage and honest limitations · August 2026
#	Block	Public cells	Status	Scope / boundary
V-05.09	Acceptance battery lineage	Bundle sha256 a87246763088ca661e112e5b32ecd1fde4e80613e9cca9dc98f7afe19f2928c0 · V1→V12b receipts · frozen-config receipts per run · per-row oracle evidence sha256	REVIEW_ONLY	Full bundle available under NDA; sanitized results-only release planned (github: external-core-evidence) · as of Aug 23, 2026
V-10.01	Honest limitation · production rerun	Acceptance battery is evidence-runner evidence; productionized-route re-run pending	NOT_CLAIMED	Self-disclosure block · as of Aug 23, 2026
V-10.02	Honest limitation · scenario families	Battery scenario families are synthetic; realized-outcome maturity is tracked separately (Shadow Ledger, forward-locked packets)	NOT_CLAIMED	Self-disclosure block · as of Aug 23, 2026

Chapter VIEngagementPlatform modules, deployment levels and pilot ask. Status tags describe route maturity, not metric evidence.DILIGENCE PATH

VI-01 to VI-03 – engagement route
#	Block	Public cells	Status	Scope / boundary
VI-01	7 platform modules	Global Risk · ARIN Council · News Analytics · Risk Analyzer · Investment Analytics · Digital Assets Analytics · KOKON	REVIEW_ONLY	Product surfaces; module-specific evidence lives on their pages
VI-02	Three deployment levels	Enterprise License for institution-owned operation · Managed Service for founder-direct evidence delivery · Compute API for embedded challenger workflows	PILOT_DEFERRED	Deployment proof happens in pilot
VI-03	Pilot ask	One representative portfolio or workflow · data dictionary · expected decision owner · review deadline · NDA jurisdiction · evidence format required	PILOT_DEFERRED	Founder-direct diligence path

Claim boundaries

Boundaries travel with the demo.

No certification or production approval is implied.

Not STAC-certified
workload built to the STAC archetype, self-run on 8×H100, NOT a STAC-audited result
pre-client institutional diligence-ready, not enterprise production-approved
not a replacement for Aladdin, Bloomberg, Murex, MSCI – embedded inside or alongside them
REALIZED_BACKTEST_PASS not claimed; 417/250 matched observations still require review/statistical gates
acceptance battery REVIEW_ONLY — diligence evidence, not production accuracy attestation
row-level numerics under NDA

Verify this page: Data room · Methodology