Skip to content

ADL-20260618-001

Project: Animus
Decision Class: ARCH
Date: 2026-06-18

Decision

Commit to v2.1 Mind-class architecture as the canonical executable baseline, abandoning the ad-hoc Core/Forge/Quorum/Bootstrap trajectory.

Chosen Option

Implement the v2.1 vertical slice: 8 technical planes, 22 JSON canonical schemas, PostgreSQL as durable authority with SQLite as lightweight projection, adversarial test harness, evidence bundles, and public/private repo split.

Rejected Options

  • Continue current trajectory: separate packages with divergent versions, SQLite-only persistence, no unified event model, no schema layer
  • Big-bang rewrite: discard current codebase entirely and rebuild from v2.1 artifacts
  • Hybrid middle ground: incrementally add v2.1 features without adopting the architecture

Why

  • The current trajectory cannot support a "Mind"-class system as defined in The_Culture_Mind_Daily_Function_and_Tooling.pdf
  • A Mind requires persistent context, world model, tool orchestration, uncertainty discipline, consent/boundaries, developmental intelligence, and independent ethical judgment
  • The v2.1 baseline provides exactly these capabilities: object registry + event ledger + bitemporal projections = persistent context; 8 technical planes = world model + tool orchestration; 22 canonical schemas = uncertainty discipline + legible intervention; policy decision point + governance plane = consent and boundaries; adversarial test harness + evidence bundles = developmental intelligence; consciousness quorum + constitutional principles = independent ethical judgment
  • The artifacts are internally consistent, well-specified, and represent ~6 months of design work
  • Continuing the current trajectory compounds technical debt: version divergence (2.1.0 vs 2.3.0 vs 2.7.0), 8 dead blockchain packages, hand-maintained counts that diverge from reality

Tradeoffs Accepted

  • Migration effort: ~3-6 months to bring current code into v2.1 structure
  • PostgreSQL as default (was SQLite): adds operational complexity but gives durable authority
  • Public/private repo split: breaks the single-monorepo convenience but enables safe sharing
  • Evidence bundle overhead: every release requires ~25 machine-readable reports
  • Dead package deletion: loses historical EVE/blockchain code (can archive)

Revisit If

  • v2.1 vertical slice is not demonstrable within 6 months
  • PostgreSQL proves operationally burdensome for personal deployment
  • The Mind-class design targets prove mismatched to actual use patterns

Addendum

2026-06-18: Truth Baseline Standard deployed to Animus (and 7 other projects) to prevent future documentation drift.