EVIDENCE CRITIC 3 — INITIAL AUDIT, 2 NOVEMBER 2026

Scope and result

The frozen English and Russian publication manuscripts, 155 atomic rules, all 1,200 query-to-rule mappings, source provenance and main empirical claims were examined. Every atomic rule was compared semantically with its retained analyst evidence and its country, service, customer, nationality, cohort and date predicates. All 329 V2 claim statements and limitations were read for scope and provenance. This is not a claim that all 329 expanded V2 claims received a new independent primary-clause verification.

The union of original support URLs for the 155 rules and 1,200 queries contains 104 exact URLs. All 104 were attempted: 92 yielded relevant primary text, 11 remained blocked or content shells, and one was a readable disclosures directory without operative eligibility text. Two of those URLs support rules without query mappings; the query-only set contains 102 URLs, with 90 relevant-text reads, 11 blocked, and one directory. Regional checks were performed by regional_source_recheck_v3; PEDSL, Bybit Global and Gemini closure readings were supplied by critic_legal_v3 and attributed in source-coverage.json. These are internal AI-assisted checks, not independent human validation or journal peer review.

All rule support texts were freshly readable to their relevant scope except the Bitfinex derivatives contract path. The per-rule decisions are 50 supported at the stated scope, 104 qualified, and one unverifiable. The qualified group includes genuinely conditional statements, historical or future windows, operator ambiguity and five predicate corrections. These counts must not be called 154 reconstructed historical documents or 1,200 independently validated customer entitlements. A URL reading does not establish every proposition ever attached to that URL.

Material corrections

1. Gemini dates: both original manuscripts group Australian withdrawal-only status with the UK/EEA at 5 March 2026. The fresh closure FAQ distinguishes Australian crypto-withdrawal-only from 1 March and UK/EEA withdrawal-only from 5 March; all cited regions share the announced 6 April closure. Australia is outside the fixed query frame, so this corrects prose rather than the matrix.

2. Five regional predicate corrections affect 16 distinct cells while preserving their outer statuses. The exact field-level recommendations are in sources/regional-recheck/patch-recommendations.json and copied into audit.json. They concern Bitpanda UK's default seven-day exit window and product/leverage extensions; OKX's restriction specifically on noncompliant asset-referenced/e-money tokens; OKX's explicit eligible new staking subscription route; Bitvavo's provider-initiated termination trigger; and Binance.US active staking versus Soft-Staking withdrawal mechanics. These are substantive scope improvements, not typographical changes.

3. Eight HTX links attach A10-R07 to onboarding or withdrawal even though that rule explicitly lists only spot, derivatives, fiat and staking. Remove or label these links as context-only. The eight cells are A10-C-htx-global-{JP,AE,CA,IN}-{onboarding,withdrawal}. They all remain unknown. No affirmative service status should be manufactured to repair a metadata link.

4. The manuscripts' statements that external publication had not occurred are stale after the 1 November public release. Version date, first posting, subsequent revision and peer-review status should be explicit. The absence of a DOI must not be inferred from absence of journal publication.

[Non-scientific correspondence about manuscript preparation omitted from this public export. The manuscript author note describes the research tools and author responsibility.]

Mapping diagnostics resolved

The scope scanner initially flags 33 nonempty-scope mismatches. Twenty-five are legitimate cross-residence applications of citizenship predicates: Kraken staking, Bybit EU nationality and Bitvavo US Person exclusions. A residence-only query leaves citizenship unspecified. These are not country-conversion errors. The other eight are the HTX service-list mismatch described above.

There are also 37 empty-dimension context links. Each is individually listed and adjudicated in audit.json. All target unknown cells: Binance GB six; Bybit Global AE six; Bitget India two; HTX four nonexcluded countries times four named services sixteen; MEXC France three and Russia four. Empty services or jurisdictions record unassigned scope, not universal applicability. These links legitimately retain the reason an affirmative answer could not be established. No status or link deletion is required; they should be treated as context evidence, not operative entailments. Twelve rules have no mapped query cells, predominantly identity, historical or future context; their existence is not a present-access claim.

Provenance and archive checks

All 635 public source observations have available authored evidence notes and matching registered hashes: 384 V2 observations, 223 V1 integrated observations, and 28 additional V1 review observations. The combined registry contains 421 exact URLs. The hash checks authenticate saved notes, not original remote pages.

Six unique remote-capture paths are registered in V2 native metadata. None is present inside the public reproducibility bundle. All six are available in the original local enrichment-v2 workspace; four are substantive selective captures (ESMA CASP CSV, field-description CSV, Binance PDF and Japanese FSA PDF). The two Bitstamp HTML captures are explicitly anti-automation shells. Counting those six as six archived operative documents would be false. The distinction is now explicit for every source record in source-coverage.json.

Every Markdown citation was parsed with balanced parentheses so the Saskatchewan Solana PDF URL was not truncated. The English manuscript has 175 linked citations across 88 exact URLs; Russian has 355 across 167. All evidence URLs match the source registry or recorded alternate reading URLs. Only the Holder.io affiliation link and the ESMA reuse FAQ are outside those evidence registries. Registered citation linkage and existence do not prove substantive support; source-coverage.json retains separate fresh, saved-prior, blocked and directory-only states. Expanded legal cases are addressed by critics/legal/caserecheck.json and authorities.json, and literature by the root's separate review.

Fresh global source findings

The substantive global clauses examined support the conservative operator/product/predicate distinctions. Coinbase UK Appendix 4 section 2.5 expressly limits cryptoasset futures, PERPs and options to professional clients. Coinbase India section 3.1 requires residence under both tax and exchange-control laws; actual residence alone does not satisfy both tests. Crypto.com section 18.2(h) supplies the residence/tax/relevant-connection link to notified restricted jurisdictions; an isolated users-from list is not sufficient by itself. Gate section 2.5 is location-based, while section 2.9 retains a broader beneficiary-residence ambiguity and section 2.10 expressly excludes US natural-person residence. HTX's agreement initially failed text extraction but was retrieved through ordinary certificate-validated HTTPS; it expressly defines its country exclusion as residence OR physical location and preserves service-dependent operators and conditional exit controls.

Historical and future timing remains material: Binance Japan's 2023 withdrawal-only notice, Bitget's August 2024 payment-rail statement, Bitfinex Ontario's 2022 exit deadline, Bitget Japan's 1 November close-only date, MEXC's assigned November-to-January batches and Bitvavo's 12 November future terms must not be collapsed into one current date. Fresh November 2 reads do not reconstruct unsaved October 30 web states. Query coding remains a frozen documentary experiment unless an explicit versioned overlay changes it.

Quantitative and interpretive boundaries

The final frozen matrix recomputes to 12 available, 273 unavailable, 187 restricted, 718 unknown and 10 inaccessible cells. The frame is 20 surfaces by 10 countries by six services. These are documentary response counts, not actual success rates, country prevalence estimates or independent observations. The V2 metadata's 384 observations, 334 exact URLs and 301 declared substantive readings are historical acquisition declarations. They do not mean 301 complete contracts or 301 independent fresh validations. The 118 regulatory observations and 102 instrument groups are not counts of licences. A licence, register listing, issuer assertion, instrument group, technical endpoint and customer entitlement remain different objects.

The original manuscript already correctly warns that V2 changes only two Kraken cells and does not comprehensively re-adjudicate the other 1,198. Preserve that statement. The later V3 audit can report exactly what it additionally checked, without erasing the history of the earlier experiment. Unknown and inaccessible cases remain explicit. No live registration, trade, deposit, withdrawal or authenticated API operation was attempted.

Outstanding verification

Eleven original query-support URLs remain blocked or content shells; the sole disclosures directory supplies links rather than operative terms. The original Bitfinex derivatives path remains the relevant unresolved atomic-rule evidence. Wider expanded V2 source statements and historical remote completeness are not certified 100 percent. The final revised manuscript and overlay still require a fresh independent consistency review once root supplies them. Canonical data and manuscripts were not edited by this critic.
