Skip to content

Animus — Document Canon & Status Index

Last updated: 2026-07-16

This repo accumulated multiple whitepapers, roadmaps, and design specs written at different times. Some describe shipped code; some describe aspirational designs that were never built; a few make claims that do not survive contact with the source. This index is the single authoritative pointer to what is canonical, what is historical, and what is aspirational. When two documents disagree, this index wins.

How to read a status

  • CANONICAL — current and code-grounded. Cite these.
  • HISTORICAL — accurate when written, now partly stale. Useful for context, not for current claims.
  • SUPERSEDED — replaced by a canonical doc; retained for history only. Do not cite.
  • ASPIRATIONAL-SPEC — a design or plan, not a description of shipped code. Valid as intent; not evidence of capability.

Whitepapers

Document Status Notes
whitepapers/ANIMUS_WHITEPAPER_2026-06.md CANONICAL Code-grounded, evidence-cited, honest about maturity. The paper to cite or hand to a reader.
WHITEPAPER.md SUPERSEDED (was v2.0, Feb 2026) Built around a fictional "production deployment". See below.
whitepapers/animus-whitepaper.md SUPERSEDED Byte-identical duplicate of WHITEPAPER.md.
whitepaper.pdf SUPERSEDED Rendered from the old v2.0. Regenerate from the 2026-06 paper before any external use.

Roadmaps

Document Status Notes
PERSONAL_ROADMAP.md CANONICAL Most current (2026-05-15). Single-user doctrine + resist-productization checklist. The load-bearing roadmap.
ROADMAP.md HISTORICAL Phased plan (Phase 0–6); phases 0–4 largely done, 5–6 aspirational.
ROADMAP_quorum_v2.md ASPIRATIONAL-SPEC 5-week Quorum v2 plan; Week-1 EventLog shipped, weeks 2–5 are specs only.
ROADMAP_research_assistant.md ASPIRATIONAL-SPEC RA-0 locked, RA-1+ to be written.
ROADMAP_TO_10.md CANONICAL The remediation roadmap: closes every audit gap to 10/10, dependency-ordered, with a completeness matrix. Cost-enforcement keystone (A1) first.
TODO_NEXT.md CANONICAL Active next-work tracker; the audit work now lives in ROADMAP_TO_10.md.
reviews/qwen-security-review-2026-06-02.md HISTORICAL (reference) Local adversarial security review (qwen2.5-14b fallback; the uncensored 35B would not load). ~70% restated known gaps; 3 net-new items folded into ROADMAP_TO_10.md Phase E (E11–E13). Model output, triaged.
reviews/animus-10-10-review-2026-06-03.md CANONICAL (verification) 19-agent adversarial audit of the post-Session-6 stack. Found Sessions 1-6 over-claimed 10/10 on D1/D3/D4/D5; 7 confirmed HIGH issues → ROADMAP_TO_10.md Phase C0. The authoritative "is it 10/10" answer: not yet.
WHITEPAPER_COMPARISON_2026-06.md CANONICAL Current code vs the 2026-06 whitepaper gap list — what Sessions 1-6 closed (✅), partially closed (◐), and left open (✗).

Known claims that are NOT backed by code (do not repeat as fact)

These appear across older docs. They are design targets or fiction, not shipped capability:

  1. Media Engine — "~480 videos/month, 3 YouTube channels, 8 languages". No implementing code in this repo (only vendored googleapiclient stubs). Appears in: WHITEPAPER.md §6, OLLAMA_SETUP.md, DEVELOPER_TOOLS.md, animus-build-spec.md. Status: never built. The actually-exercised workloads are developer/fleet-ops (code-review, fleet-triage, security-audit, bounty-watcher) plus the eval harness as a regression gate.
  2. Marketing Engine — 5-platform autonomous posting. Design only. SECURITY_LAYER.md, BROWSER_AUTOMATION.md, DEVELOPER_TOOLS.md reference it as future context. Status: specced, not built.
  3. Encryption at rest (AES-256) / Ed25519-signed memory. Persistence-layer encryption via gocryptfs vault shipped (PR #67, A5): the memory store directory can be mounted inside an encrypted filesystem. Individual memory-record signing remains unimplemented. Status: partial — at-rest encryption operational, record-level signing future work (see THREAT_MODEL.md).
  4. Developer-Tools / Arete-Tools revenue-tier ecosystem. DEVELOPER_TOOLS.md describes a 9-tool suite with $5–49/mo tiers. Status: superseded by the PERSONAL_ROADMAP.md anti-productization stance (no SSO/RBAC/billing).
  5. HOT/WARM/COLD tiered memory + lossless compaction. Fully specced in ANIMUS_MEMORY_GAPS.md; zero implementing code.
  6. Quorum v2 active-inference resolver / liveness watchdog / coupling dashboard. The StabilityScorer Protocol + DefaultStabilityScorer shipped (B7). Active-inference resolver, liveness watchdog, and coupling dashboard remain specs-only.

Design / spec docs (valid as intent, not as capability)

ANIMUS_MEMORY_GAPS.md, SECURITY_LAYER.md, BROWSER_AUTOMATION.md, DEVELOPER_TOOLS.md, WORKFLOW_EVOLUTION_CONSTRAINTS.md, rework/*, specs/* — all ASPIRATIONAL-SPEC unless a feature is confirmed in the canonical whitepaper's evidence index.

CITIZEN_INTEGRATION_CONTRACT.md is CANONICAL as a bridge document: it maps constitutional rights to code but does not claim unimplemented features are operational. Every entry is labeled OPERATIONAL, PARTIAL, or NOT YET IMPLEMENTED.

Grounded references (accurate)

THREAT_MODEL.md, CONSTITUTIONAL_PRINCIPLES.md, CONSCIOUSNESS_QUORUM_BRIDGE.md, CITIZEN_INTEGRATION_CONTRACT.md, and each package's CLAUDE.md are CANONICAL for their subsystem. The constitutional-principles doc-to-code annotation style (every principle points at its enforcing module) is the grounding standard the rest aspires to.