UAI Memory Organization Report

Status

Current implementation report for IARPG-OPS-2 2.0.6-wip. The split-memory migration is complete at the repository and package level. Stable release promotion remains blocked by native Firefox, native assistive-technology speech output, and actual target-host evidence.

Purpose

Document the inspected baseline, additive memory migration, report consolidation, conflict resolution, deep-link normalization, validation, blockers, and recommended continuation action.

Scope

This report covers the IARPG repository supplied as the 2.0.5-wip baseline, the 38 Markdown reports supplied with the handoff, all existing hot UAI records and mirrors, canonical reports, source manifests, public discovery files, historical prompts and release notes, and the resulting 2.0.6-wip package. It does not independently fact-check every external research claim inside the report archive.

Executive Summary

The repository now uses a split-memory architecture. .uai/ remains the active, date-free, first-load continuity layer. Detailed reports, architecture, decisions, research, implementation history, and superseded material live under docs/long-term-memory/. All 183 pre-existing canonical reports were moved from docs/reports/ to docs/long-term-memory/reports/; this implementation report raises the canonical total to 184. Every canonical report now links back to applicable hot UAI records, and every durable authoritative document has a semantic pointer-ledger record with a path, useful deep link, routing summary, authority, review state, review evidence, truth boundary, checksum, supersession fields, and related hot memory.

The 38 Markdown reports supplied with this handoff were not duplicated: each already matched an existing preserved source file by SHA-256 and remains incorporated through the source manifest, canonical report archive, search, and topical UAI routing. Evidence: provided report intake audit.

Evidence Reviewed

Files Inspected

• AGENTS.md and the documented startup read order.
• All 90 baseline .uai/*.uai files and all matching docs/uai/ mirrors.
• Root totem.uai, taboo.uai, release notes, next prompts, validation files, deployment notes, public discovery records, sitemap, search index, and standards documentation.
• docs/manifest.json, all 183 baseline canonical reports, docs/source-files/manifest.json, and all 255 source-file records.
• The 38 supplied Markdown reports, compared to source-manifest records by SHA-256.
• Report readers, docs routes, Apache rewrite rules, bootstrap metadata, manifests, checksums, and package boundaries.
• Git availability. No .git directory was supplied, so history inspection was not possible.

Authority Used

Current implementation and executed tests outranked derived manifests. Active hidden UAI records outranked stale root compatibility copies. Historical prompts and release notes were treated as implementation history, not current authority. Full conflict detail: memory authority and conflict audit.

Findings

Files Created

• Required hot records: .uai/index.uai, .uai/context.uai, .uai/constraints.uai, .uai/memory.uai, and .uai/long-term-memory.uai.
• Required directories: .uai/archives/, .uai/exports/, and the complete docs/long-term-memory/ hierarchy.
• Durable architecture, ADR, research audits, source map, implementation-history routing, human index, and this implementation report.
• scripts/validatememorylinks.py for offline path, anchor, checksum, backlink, mirror, placement, timestamp, and secret-pattern checks.
• 2.0.6-wip release notes, validation evidence, and the standalone 2.0.7-wip continuation prompt.

Files Updated

• Core and topical .uai records were normalized to date-free current state, repository-relative links, and current release metadata.
• .uai/progress.uai, .uai/decisions.uai, .uai/architecture.uai, .uai/memory.uai, .uai/test-plan.uai, and .uai/next-recursive-prompt.uai now contain direct durable evidence links.
• Every public docs/uai/ mirror was regenerated byte-for-byte from its canonical hidden record.
• All canonical reports received Memory References, Related Durable Documents, and Supersession Status sections.
• Report manifest, source mappings, search index, sitemap, public discovery files, docs copy, bootstrap version, checksums, deployment notes, release notes, and validation files were updated.

Files Moved

• 183 active repository-owned canonical reports moved from docs/reports/ to docs/long-term-memory/reports/.
• The 75 KB source-to-report map body moved from .uai/source-map.uai into docs/long-term-memory/research/source-to-report-map.md; the hot record now contains only current inventory and routing.

Files Intentionally Left in Place

• docs/source-files/ remains at its existing public path because its 255 Markdown files are immutable user-supplied provenance artifacts, not repository-authored canonical report bodies. Moving or appending backlinks would break byte identity and externally published source URLs. This exception is indexed and checksum-verified.
• Root and docs/project-prompts/ historical release notes and next prompts remain because existing package and handoff conventions reference them. Their current-authority boundary is documented in release and handoff history.
• Root totem.uai and taboo.uai remain only as compatibility pointers to the canonical .uai/ records.
• Generated validation JSON, schemas, fixtures, and public standards records remain in their established tool-controlled paths.

Stale or Conflicting Memory Discovered

• Root totem and taboo files described 2.0.4-wip while hidden canonical records described 2.0.5-wip.
• Public discovery counts advertised 147 reports, 219 sources, and 84 UAI records although current manifests held 183 reports, 255 sources, and 90 UAI records.
• The search index omitted 36 international source records and six international UAI records.
• All hot UAI records carried timestamps; many used public-site URLs instead of repository-relative evidence links.
• progress.uai contained a release-history section and source-map.uai contained durable catalog body content.
• Canonical reports lacked backlinks to hot memory.

How Conflicts Were Resolved

• Verified files and machine manifests established the baseline inventory.
• Stale derived discovery and search artifacts were regenerated instead of treated as authority.
• Root UAI duplicates were demoted to compatibility pointers; .uai/ is the sole active memory home.
• Historical release material was retained but explicitly routed as implementation history.
• Report research was preserved without being promoted to verified current truth; each pointer record states its review evidence and truth boundary.

Deep Links Added

• Hot core records now link to specific headings in ADRs, architecture, audits, the implementation report, international fairness reports, and source intake evidence.
• Every canonical report links back to applicable topical UAI records, .uai/memory.uai, and .uai/long-term-memory.uai.
• Durable architecture, decisions, research, implementation history, README, and index files link back to the hot record that summarizes them.

Broken Links Corrected

• Repository-local UAI evidence links no longer depend on https://iarpg.com/ being available.
• Canonical raw report download URLs now use /docs/long-term-memory/reports/.
• The legacy /docs/reports/ path redirects to the new body location and contains no duplicate reports.
• Source-manifest canonical section links were regenerated from the current report headings.

Decisions or Recommendations

• Keep the split-memory decision active until explicitly superseded by a later ADR.
• Use .uai/index.uai as the first startup record and the pointer ledger only for durable routing.
• Keep source evidence immutable and separate from canonical report bodies.
• Run the local memory validator after every report, UAI, or durable-document change.
• The single best next action is to create a balanced five-region batch of dedicated country profiles and then run an independent cross-country fairness audit.

Risks and Limitations

• Native Firefox, native assistive-technology speech output, and actual target-host behavior were not available for direct validation.
• Managed Chromium policy blocked local HTTP navigation. The current browser-local fixture and review harnesses were therefore executed in Chromium against server-rendered HTML with the package's current JavaScript injected locally; the full navigated critical-runtime, theme, and viewport suites remain open.
• The local PHP server does not reproduce Apache .htaccess redirects, so compatibility routing was verified statically; it still requires target-host smoke evidence.
• Many supplied research reports remain user-provided or generated research and were not independently fact-checked. Pointer-ledger truth boundaries prevent them from overriding implementation or tests.
• No Git metadata was supplied, so file history could not be preserved with git mv or inspected for prior authority decisions.

Validation Performed

The final 2.0.6-wip candidate was checked locally without network dependencies:

• python3 scripts/validatememorylinks.py passed with 95 hot UAI records, 184 canonical reports, 193 durable Markdown documents, 2,420 resolved local links, 2,036 validated heading anchors, 193 pointer-ledger records, and 95 byte-identical UAI mirrors.
• php -l passed for all 85 PHP files; node --check passed for all 5 JavaScript files.
• Python JSON parsing passed for all 264 JSON files; the sitemap parsed as XML with 335 routes.
• Draft 2020-12 schema validation passed for all 12 current root and well-known machine records that declare a local schema. Historical versioned fixtures remain preserved rather than rewritten to the current schema constant.
• The current Chromium browser-local fixture harness passed 18 of 18 checks, including package components, component digests, deterministic replay, evidence identity, Level D mission structure, migration field coverage, debrief separation, and OPS-1 code preservation.
• The current Chromium review-lifecycle harness passed 8 of 8 checks, including package integrity, explicit merge-conflict discovery, UTC replay order, evidence-custody coverage, seven migration regression classes, and deterministic batch input digests.
• All 20 IARPG-CJSON-1 parity vectors passed in current Chromium: 11 passing vectors produced the published canonical form and SHA-256, and 9 invalid vectors were rejected.
• A local PHP server returned HTTP 200 for 27 of 27 representative public, global, report, UAI, standards, mission, tool, manifest, search, and sitemap routes.
• Repository scans confirmed 255 immutable source files, 184 canonical report bodies under the required directory, 0 canonical report bodies under the legacy directory, 699 search records, 335 sitemap routes, and 0 nested archives.
• Documentation checksums were regenerated and verified with sha256sum -c docs/checksums.sha256.
• The flat deployment ZIP was generated from the document root with hidden files, passed unzip -t, contained no enclosing project directory or nested archive, and matched its external SHA-256 sidecar.
• Common secret-pattern checks in the memory validator found no copied credentials, tokens, API keys, cookies, or private-key bodies in hot or durable memory.

Machine-readable results and the remaining not-tested promotion gates are recorded in VALIDATION.json, VALIDATION.txt, and release-promotion-gates.json.

Memory References

• Hot memory index
• Current progress
• Active decisions
• Memory operating model
• Architecture
• Pointer ledger

Related Durable Documents

• Split-memory architecture
• ADR-001
• Memory authority and conflict audit
• Provided report intake audit
• Long-term memory index

Supersession Status

Current implementation report for 2.0.6-wip. A later report must identify stable ID ltm-report-uai-memory-organization and name this file in supersedes before replacing it.