Nexus CerebriOPEN-SOURCE PLANNING PROJECT

Historical / non-normative review record. Current authority: Master specification.

Historical Review Resolution: Specification 0.1 -> 0.2

NON-NORMATIVE HISTORY. All accepted architectural decisions from this review must exist in the Master Specification or relevant normative standard. If this file conflicts with current normative documentation, this file loses.

Date: 2026-09-19
Status: Historical audit record — non-normative after consolidation

This document records how the independent v0.1 review was resolved.

Critical findings

CRITICAL-01 — DE/EN structure mismatch

Accepted. v0.2 defines parity for canonical public docs and explicitly allows internal logs/research notes/ADR records to remain canonical-language-only. The actual implementation repository must create the DE/EN structure during foundation bootstrap.

CRITICAL-02 — Required governance files absent

Accepted as bootstrap gap. The planning baseline/package is not itself the implementation repository. The foundation milestone must create README, CHANGELOG, ROADMAP, CONTRIBUTING, SECURITY, ADR and progress structures before claiming v0.1.0 software readiness.

CRITICAL-03 — Version source of truth

Accepted. v0.2 defines separate canonical authorities for software, API, CPIR, model, dataset and specification revision.

CRITICAL-04 — Plan lifecycle ambiguity

Accepted. Formal lifecycle is now: ProposedPlan -> ValidatedPlan -> ActionPlan -> AuthorizedActionPlan -> ExecutionResult.

CRITICAL-05 — Permission scope ambiguity

Accepted. Permissions are removed from ordinary hard constraints. Planning capability and execution authorization are separate security layers. Simulation may return explicitly non-executable hypothetical results when policy permits.

Major findings

  • README completeness: accepted as foundation bootstrap requirement.
  • GitHub-state wording: clarified to default branch/releases.
  • Calendar-driven releases: corrected; months provide candidate milestones only.
  • CPIR v1 ambiguity: corrected to initial internal schema 0.1.
  • Planner status mixing: split into PlanningOutcome and SearchAssessment.
  • Unknown/missing/uncertain/ambiguous: formal knowledge-state semantics added.
  • Bounded scope: concrete dimensions defined.
  • Derived facts/inference: provenance and derivation distinction added.
  • ADR triggers: explicit trigger list added.
  • Research/production boundary: production may never depend on research.
  • ActionLedger ownership: executor/application layer owns durable state/recovery.
  • Zero violation metric: validator correctness must be independently tested.

Minor findings

Resolved or clarified:

  • specification revision is separate from software/package version;
  • ISO 8601 date rule added;
  • DE-first fixtures are dataset strategy, not documentation/product-language priority;
  • conceptual enum names may change in Rust unless serialized/public contract requires stability;
  • external lock defined;
  • ML-v0 etc. are learning architecture milestones, not model artifact versions;
  • GitHub Pages does not imply a fixed generator/tool yet;
  • visualization is required when applicable/useful;
  • deterministic and wall-clock search budgets are distinguished.

Questions resolved

  • Default branch should be coherent; releaseability depends on release gates.
  • Software version authority: Cargo workspace/package version.
  • Spec 0.2 is a review/document revision, not compatibility guarantee.
  • CPIR begins as internal versioned IR.
  • Permissions affect executable planning capability and are rechecked at execution.
  • Hypothetical non-executable suggestions are allowed only in analysis/simulation policy.
  • OPTIMAL means proven optimum over the explicitly defined complete search space/objective.
  • Tie-breaking is deterministic.
  • v0.x ConflictSet is sound/explainable, not promised minimal.
  • Preferences become policies only explicitly, never by learning alone.
  • Policy is a typed deterministic layer.
  • Mutation scope is explicit; visibility never implies mutability.
  • Unknown external participant availability remains UNKNOWN; blocking depends on operation/policy.
  • personal_pattern should minimize raw personal data.
  • PERSONAL_LOCAL is logical personal scope, not necessarily on-device storage.
  • progress logs are required for meaningful work/decisions, not passive reading.
  • DE/EN CI verifies structural parity; semantic parity requires review.
  • three-level explanation is mandatory for central public/didactic concepts, not every internal note.
  • Cerebri Lab is initially internal research/developer tooling.
  • specification revision and software v0.1.0 are separate namespaces.