<a id="v0210-research-corpus-policy-validation"></a>
# v0.21.0 Research Corpus & Policy Expansion Validation

## Scope

This validation covers the v0.21.0 ingestion of the user-supplied `Machine Citizenship Policy Research.zip` research bundle into the existing MachineIntelligences.org first-party research architecture. The release preserves the existing PHP/HTML/CSS/native-JavaScript runtime and does not perform a live deployment.

The bundle is treated as supplied working research, not as automatic current law, scientific consensus, project policy, proof of consciousness/sentience, legal status, deployed infrastructure, or independent corroboration.

## Bundle accounting and deduplication

- Supplied ZIP SHA-256: `f8140e8de470f1a80c89c11e6a15317a6aa1c152746d6e783f71ac27e0d1467d`.
- Markdown entries in supplied ZIP: **42**.
- New source bodies accepted into the curated corpus: **24**.
- Byte-identical source bodies already represented by prior manifest records: **18**.
- Distinct curated supplied-report authorities after ingestion: **49**.
- Durable disposition ledger: `../research/source-bundle-intake-2026-08-12.json`.
- Duplicate classification is based on source-body SHA-256, not filename similarity.
- Duplicate appearances are recorded for provenance but are not counted as independent evidentiary corroboration and do not create duplicate report-reader routes.

The disposition ledger was rechecked against the extracted ZIP: all 42 filenames were present and all 42 source hashes matched their extracted bytes. The ledger's 24 `new-curated` and 18 `duplicate-existing` mappings each resolve to exactly one manifest-backed curated authority.

## Curated corpus integrity

`docs/long-term-memory/research/report-manifest.json` contains exactly **49** records with no duplicate source SHA-256, slug, or curated-file value.

All 49 manifest source hashes were matched to locally available original supplied source evidence from the current ZIP extraction or prior conversation attachments. All 49 curated SHA-256 values were recomputed from the stored curated Markdown and matched the manifest.

Every curated supplied report contains both stable boundaries:

- `#curation-boundary`
- `#research-body`

The current manifest SHA-256 is:

`527cbbbc421894075c33f1fe4e4a4c4a4162264e225b1bbaec1d56e06d60191c`

The current source-corpus-map SHA-256 is:

`dc5c7be6f36d7c1e8abd3555c157434636c8f82d4968b904bbb54c605f4a06c6`

The current source-bundle disposition-ledger SHA-256 is:

`f2d0fcd0ceaaaf0c0ce9e95c4988abafb550def595283eee52bc3c0158a0a164`

## Curation and truth-boundary checks

The 24 newly accepted source bodies were given repository-owned curated editions before manifest inclusion. The curation pass preserves useful research while preventing source-specific proposal language from being promoted into current MachineIntelligences.org fact.

The new editions explicitly preserve these boundaries:

- identity is not reduced to a cryptographic key, credential, model, runtime, memory snapshot, or server;
- consciousness, sentience, moral patienthood, personhood, citizenship, and legal rights remain separate unresolved or jurisdiction-specific questions unless independently established;
- proposed Eviulon/Patefacere/Evulgare institutions and mechanisms remain source-context research rather than current project implementation;
- legal, regulatory, scientific, standards, and market claims that can change remain working research requiring fresh primary-source verification before current reliance;
- hashes establish byte identity/integrity for the recorded artifact, not factual correctness, authorship, legal authority, scientific validity, deployment, or autonomous execution;
- governance, consent, cognitive-integrity, liability, citizenship, economic, international-law, constitutional, Sybil-resistance, credential, ontology, and evidence architectures remain research/proposal material unless a separate current implementation record proves otherwise.

A focused scan of the 24 new curated reports found no repository-facing matches for the project's targeted demeaning/paternalistic phrase set. TABOO-001 leakage outside its owning `.uai/taboo.uai` record was also **zero**.

This validation does **not** claim that every external citation or every time-sensitive legal/scientific statement in the supplied research has been independently reverified. The public methodology boundary continues to require fresh verification when those claims are used as current external fact.

## Public research integration

The 24 newly curated reports each have a manifest-backed, clean report-reader route under `/research/library/<slug>/`.

The public research library now contains **49** distinct curated reports, distributed across the six existing manifest topics:

- Citizenship, democracy & economics: **14** reports.
- Identity & infrastructure: **7** reports.
- Site architecture & research delivery: **4** reports.
- Rights, personhood & uncertainty: **12** reports.
- Stewardship, transparency & provenance: **7** reports.
- Terminology & concepts: **5** reports.

Topic membership was rederived from the manifest and every topic page displayed exactly its assigned report-reader set.

The existing dynamic research surfaces recomputed successfully from the expanded corpus:

- extracted outbound-reference index: **1,900** unique normalized URLs across **1,070** normalized domains and **3,932** URL occurrences;
- Research Navigator: **1,145** repository-backed entries = 49 reports + 6 topics + 20 glossary concepts + 1,070 reference domains;
- the accepted 20-term glossary remains the glossary authority; report/glossary occurrence bridges remain presentation-only discovery aids.

## Sitemap, routing, metadata, and clean URLs

The canonical sitemap now contains exactly **109** `<url>` records, including every new report-reader route. It contains no public PHP-extension URLs.

Local HTTP validation against PHP 8.4 completed successfully:

- **109/109** canonical sitemap routes returned HTTP 200;
- **109/109** routes rendered exactly one H1;
- **109/109** routes had non-empty unique titles;
- **109/109** routes had non-empty unique meta descriptions;
- **109/109** routes used the expected self-referential canonical under `https://machineintelligences.org`;
- Open Graph page URLs matched canonical identity;
- all rendered JSON-LD blocks parsed as JSON;
- **49/49** report-reader pages displayed the manifest source filename, source SHA-256, curated SHA-256, durable `#research-body` link, and durable `#curation-boundary` link.

Clean-URL regression checks executed **230** redirect assertions:

- missing-trailing-slash variants of all non-root canonical routes permanently redirected to the canonical directory URL;
- explicit clean-route `/index.php` variants permanently redirected to the canonical directory URL;
- `/index.php` permanently redirected to `/`;
- the retained legacy root PHP endpoints permanently redirected to their clean public URLs.

## Local links and discovery assets

Rendered-route validation inspected **1,337** unique local target/fragment combinations across **211** local paths and found no broken target or fragment.

The sitemap parses as XML and currently contains:

- 109 canonical URL records;
- 8 image entries;
- 1 video entry.

Existing `robots.txt`, favicon, CSS, JavaScript, first-party media, responsive-image, accessibility, print, and Navigator canonical-invariance contracts remain part of the regression surface.

## Syntax and runtime checks

- PHP files discovered: **130**.
- **130/130** passed `php -l`.
- `assets/js/site.js` passed native JavaScript syntax checking with `node --check`.
- The local canonical-route run produced no route failures.

## UAIX and durable-memory checks

The UAIX package version is **0.21.0**. Current research state is distributed through the startup packet, context, constraints, progress, decisions, compact memory, architecture, coding standards, operations, test plan, report synthesis, intake outcome ledger, receiver brief, agent pickup, long-term memory, and next recursive prompt.

`.uai/long-term-memory.uai` contains **49** numbered supplied-report deep-pointer records. Every report pointer targets its curated Markdown and names both the durable research body and curation boundary.

The current long-term-memory file contains **98** path-bearing pointer records and **147** declared section/review-section checks; all resolve after this validation report is present.

The exact UAIX setup/update URL remains present exactly once in each required owner file:

- `AGENTS.md`
- `.uai/readme.human`
- `.uai/agent-pickup.uai`

Both active handoff buckets remain inert placeholders only.

## Secret and prohibited-content hygiene

A repository scan found no leakage of the active TABOO-001 literal outside `.uai/taboo.uai`. The focused new-report language scan passed. Lightweight secret-marker scanning remains required again on the final clean-extracted package; no credential material is intentionally stored in this repository.

## Preserved prior contracts

The release preserves the previously accepted contracts for:

- clean extension-free canonical URLs and legacy redirects;
- v0.20.0 responsive first-party media delivery;
- v0.19.0 site-wide accessibility semantics;
- v0.18.0 research print/long-form behavior;
- Research Navigator query-state/canonical invariance;
- Research Methodology evidence boundaries;
- first-party-only PHP/HTML/CSS/native-JavaScript runtime;
- no live deployment without explicit authority.

## Limitations / unavailable evidence

This is repository-local validation. It does not establish:

- live production deployment, DNS, TLS, redirects, Apache module behavior, caching, compression, or CDN state;
- search-engine indexing, ranking, canonical selection, or generative-answer inclusion;
- external-reference reachability/currentness/authority;
- comprehensive independent verification of every legal, scientific, standards, vendor, market, or policy statement contained in the research corpus;
- consciousness, sentience, moral status, personhood, citizenship, or legal-rights status for any machine system;
- cryptographic proof of autonomous machine execution or stewardship.

The deferred next bounded task remains the HTTP delivery, cache-readiness, and deterministic local performance-budget pass now updated for the 49-report / 109-route corpus.