VSO updates research evidence preservation, rechecks and guide notices
This follow-up covers eight merges checked at 18:25 JST on 30 October 2026: VAP PRs #33–#38 and VCP Conformance Guide PRs #3 and #4. It supplements the earlier five-update announcement. Measured local actions, rechecks of saved evidence and documentary mappings remain separate, with revision and first-party CI references below.
Research status — unadopted; independent reproduction not established
Proposals and experimental extensions remain unadopted; independent third-party reproduction is not established. Passing first-party Python 3.12 / 3.14 CI, rechecking saved evidence or generating bilingual reports does not establish an independent fresh run, human review, normative adoption, full conformance, production assurance or certification. One controller manages the local experiments, test keys and roles.
1. VAP #33: bind publication, disclosure and reviewer authorization
The local pilot binds edit history, generated HTML, AI-use disclosure and a separate scoped, time-bound reviewer authorization. Durable intent and current checks precede one publication and readback. First-party CI passed 45 test methods, 18/18 filesystem scenarios and 18/18 saved-evidence rechecks. AI-use labels and reviewer keys are synthetic, not a real AI service or independently established human review. Writes outside the gate remain possible; this does not establish OS privilege separation or control of every route. The existing assessor retains publication_action: REVIEW.
Change and revision: vap-spec PR #33 · 4d98e9531751f5a5c8e12550fab81eb0ee4e7592 · First-party CI run
2. VAP #34: scoped local erasure evidence
One synthetic local object is bound to a signed request, version, scope and expiry; the pilot checks durable intent, unlink, observed absence and a signed scoped completion. 54 test methods, 18 filesystem scenarios and 18 export rechecks passed. Retained-copy, open-handle and recreation cases distinguish namespace removal from universal erasure. Remote-provider deletion, media sanitization and erasure of every copy are not established; out-of-scope outcomes remain UNKNOWN.
Change and revision: vap-spec PR #34 · b72bbf05c912f5b87560f4c809298c7bbcaf05e4 · First-party CI run
3. VAP #35: preserve 53 original cases and an offline recheck kit
Original evidence for 17 upload, 18 publication and 18 erasure cases, summaries and provenance are preserved alongside a kit with frozen sources. 29 test methods and 53/53 saved-case rechecks passed. The archive hash must be checked against a trusted value outside the archive. This is reinspection of saved evidence, not a fresh execution of each action or proof of an honest executor. The fixed 24 October research kit is unchanged.
Change and revision: vap-spec PR #35 · 1553ff9729ceb30f3ff03513cead45c5b7abd1d9 · First-party CI run
4. VAP #36: source-attributed English and Japanese reports
The experiment rechecks the 53 frozen cases and generates English and Japanese reports tied to the supporting evidence. First-party CI passed 37 test methods, 53 case rechecks and deterministic report rederivation checks. Automated reporting is not human review: human review remains NOT_PERFORMED; independence and observation truth remain NOT_ESTABLISHED.
Change and revision: vap-spec PR #36 · 3e621d522c4e5f536c850aa1e17821328680bfe6 · First-party CI run
5. VAP #37: coverage mapping and external review handoff
The 23 proposal items are linked through 27 relationships: 16 related local observations and 11 partial analogues, not counts of fully satisfied requirements. PB-06, IR-01, IR-02 and XR-01 have no measured equivalent. Procedures and a blank report distinguish saved-evidence rechecks, fresh runs and deployment evaluation. This documentary mapping adds no new runtime result or external reviewer result. The three pilots are not an integrated operational lifecycle.
Change and revision: vap-spec PR #37 · 433f019513d991982fc6fb08bd4529808cc9f6a1
6. VAP #38: collector outages and read-only recovery
A separate loopback collector process exercises durable pre-send intent, a matching signed receipt, fresh readback and current authorization as prerequisites for sending. Lost acknowledgements do not let read-only recovery resume an old request; a newly and explicitly authorized request is required. 41 tests, 18 real-process/network scenarios and 18 saved-evidence rechecks passed. A metadata receipt is not a payload receipt or independent observation. Power-loss resilience, continuous availability and prevention of all bypass routes are not established.
Change and revision: vap-spec PR #38 · 589d6b522c8d23261e9bfce1103784ac460c85ee · First-party CI run
7. VCP guide #3: illustrative names versus the available SDK
The historical guide now identifies “vcp-sdk 1.0.0” and vcp_sdk imports as illustrative names. The published veritaschain/vcp-sdk is a 0.1.0 pre-release offline core for VCP v1.2; its Python import is veritaschain_vcp. It is not the illustrated logger, Explorer client or a conformance test suite.
Change and revision: vcp-conformance-guide PR #3 · 0fc6d3d3ef75c92d9b8b7ee235a4f35119755057
8. VCP guide #4: historical certification process notice
A dated notice identifies the guide’s certification procedure as historical and not in effect. VSO does not itself review submissions or issue certifications. Historical text is retained while current descriptions are corrected to avoid implying an active application and audit process. This is not the launch of a new certification scheme.
Change and revision: vcp-conformance-guide PR #4 · 916cc716596a68dafca5e5173f2b561ed7c8cef9
Preserve historical records and distinct scopes
Test methods, scenarios and rechecks cover different measures and are not summed. The 53 cases are the existing 17 + 18 + 18 evidence captures, not 53 new executions. The earlier five-update article, Explorer API 159 / GUI 26 tests, SDK Python 306 / TypeScript 140 tests, VAP-LAP notices and the fixed 24 October kit remain dated records. This website update adds explanatory material only; current normative requirements, schemas, existing implementations, signed data, fixed distributions and K3 are unchanged.