# v0.15.0 Research References and Source Discovery Validation ## Scope This record validates the v0.15.0 Research References & Source Discovery addition and the repository state packaged from `/mnt/data/mi_v0150_work`. The feature adds `/research/references/`, a first-party local-only index derived from outbound HTTP/HTTPS URL tokens already present in the 25 curated Markdown research reports. External destinations were intentionally **not fetched** during rendering or validation; this release validates repository extraction and routing, not external source quality or reachability. ## Focused feature checks - `includes/research-references.php` and `/research/references/` PHP syntax passed. - Independent reconstruction from the 25 curated Markdown bodies produced **1,642 external URL occurrences**, **804 unique normalized external URLs**, and **525 normalized source domains**. - The rendered references page contained exactly **804 reference items** and **525 domain groups**. - Every rendered external reference URL matched the independently reconstructed set. - Every rendered reference had at least one traceable raw URL token in a curated report and every rendered report backlink matched the canonical local report-reader route for a report containing that reference. - Domain grouping matched the independently reconstructed normalized host for every reference. - Internal absolute MachineIntelligences.org URLs were excluded from the outbound reference index. - Normalization remained conservative: Markdown punctuation escapes are removed for usable destinations, hosts are lowercased, default HTTP/HTTPS ports are normalized, while `www`/other subdomains and path/query/fragment identity are preserved. - No PHP runtime external network-client pattern was introduced for the feature; no crawler, database, external search service, or third-party runtime library was added. - `/research/`, `/research/library/`, and `/research/methodology/` all link the canonical `/research/references/` route while the primary navigation definition remains unchanged from v0.14.0. - The references page visibly states that occurrence does not establish reachability, currentness, correctness, authority, endorsement, or support for a claim, and routes readers to Research Methodology for the verification boundary. ## Full public-route and technical checks - All **104 PHP files** passed `php -l`. - `sitemap.xml` contained **84 canonical page URLs**, including `/research/references/`. - All **84 canonical public routes** returned HTTP 200 from the local PHP test server. - All 84 routes had exactly one H1, a non-empty unique title, a non-empty unique meta description, the expected self-referential canonical URL, matching Open Graph URL, and parseable JSON-LD. - Clean-route normalization passed for **167 missing-slash / explicit `index.php` variants** with HTTP 301 to the expected canonical directory URL. - Existing root legacy PHP page endpoints continued to return permanent clean-route redirects. - `robots.txt` retained the canonical sitemap pointer. - **149 unique rendered local target paths** collected from public href/src attributes returned HTTP 200 or the expected permanent canonical redirect. - Rendered public URL attributes, canonicals, Open Graph URLs, and JSON-LD did not expose `.php` public URLs. - The local server log contained no PHP warning, notice, deprecation, parse-error, or fatal-error output during the validation run. ## Research authority preservation - `report-manifest.json` still contains **25 supplied-report entries**. - All **25 curated SHA-256 values** matched the current curated Markdown bytes. - All **25 original source attachment SHA-256 values** were rechecked against the supplied attachments available in `/mnt/data` and matched the manifest. - Every supplied report retained both stable `#curation-boundary` and `#research-body` anchors. - Every one of the 25 curated report files remained byte-identical to the v0.14.0 accepted package. - `docs/long-term-memory/research/report-manifest.json` remained byte-identical to v0.14.0. - `includes/glossary-data.php` remained byte-identical to v0.14.0, preserving the accepted 20-term glossary definitions and registry data. - The reference index is derived presentation/discovery metadata only; it does not modify report text, report hashes, glossary definitions, or structured-data semantics. ## UAIX, documentation, and hygiene checks - The UAIX package version and `SITE_RELEASE` are **0.15.0**. - Required active `.uai` records were present and non-empty. - `.uai/long-term-memory.uai` contains **61 durable pointer records**, including v0.15.0 history and validation pointers plus all 25 supplied-report deep links. - **209 UAIX backtick path references** and **192 repository-local Markdown link tokens** were checked through repository path/link validation with no missing target detected. - The exact UAIX setup/update URL remains unchanged and appears once in each required owner file: `AGENTS.md`, `.uai/readme.human`, and `.uai/agent-pickup.uai`. - Active File Handoff Content and Improvement buckets contain only inert `.keep` placeholders. - The active prohibited organizational-status literal occurred only in its owning `.uai/taboo.uai` rule record; no leakage was found elsewhere in the repository. - The unwanted paternalistic catchphrase scan returned zero findings outside the active rule framework. - Lightweight private-key/token marker scanning returned zero findings. - No `.git` metadata was present in the supplied repository state. ## External checks intentionally not performed - External reference reachability, HTTP status, currentness, authority, publication status, or claim support were **not checked**. That would be a separate fresh-verification task and is explicitly outside the local extraction contract. - Live deployment was not authorized or performed. - Production DNS, TLS, web-server rewrite behavior, PHP/cache/CDN configuration, canonical host behavior, and uptime remain unverified. - Search Console, Bing Webmaster Tools, indexing/ranking, and generative-search citation behavior remain unavailable/unverified. - Production Core Web Vitals / field-performance data remain unavailable. - Independently verifiable cryptographic machine-execution provenance remains unimplemented. ## Result The v0.15.0 Research References & Source Discovery feature is locally validated as a first-party, deterministic, local-only extraction and navigation layer over the unchanged 25-report curated corpus. The release preserves the existing methodology boundary: **a reference appearing in the corpus proves occurrence, not external verification**. ## Root-package validation - The root-deployment archive contains **336 entries** with `index.php`, `.htaccess`, `.uai/`, `docs/`, `includes/`, `assets/`, `robots.txt`, `sitemap.xml`, and `favicon.ico` directly at archive root rather than inside an enclosing project directory. - A clean extraction of the release archive was served separately on a local PHP test server. - The clean extraction re-linted all **104 PHP files**, re-smoked all **84 canonical routes**, rechecked clean redirects, reconstructed all **804 references / 525 domains / 1,642 occurrences**, revalidated the 25 source/curated report hashes, and reran local-link, UAIX, taboo, secret-hygiene, handoff, and authority-preservation checks with zero errors.