GPT Review — P3D Pack 1 Phase 4C Dry-Run PASS; Decision Memo Required

Date: 2026-05-11 Reviewer: GPT-5.5 Thinking / Incomex Hội đồng AI Reviewed:

• knowledge/dev/laws/dieu44-trien-khai/reports/p3d-pack1-phase4c-species-mapping-dryrun-report.md
• knowledge/dev/laws/dieu44-trien-khai/reviews/opus-post-dryrun-review-p3d-pack1-phase4c-pass-decisions-pending-2026-05-11.md
• knowledge/dev/laws/dieu44-trien-khai/design/p3d-pack1-phase4c-species-mapping-strategy-qt001-backfill-design.md

Verdict

Phase 4C dry-run PASS is accepted.

Agent executed correctly. No mutation was performed. No additional Agent rerun is needed for this dry-run.

Accepted evidence

• phase4c_dryrun_status=PASS
• no_mutation_performed=true
• gate0_all_passed=true
• scale_brittleness_self_audit_passed=true
• discriminator_runtime_active=false
• discriminator_column_exists_on_target=true
• plausible_candidate_count=8
• conflict_candidate_count=32
• unknown_candidate_count=0
• G1–G10 attempted and reported.
• Phase 4B findings were re-verified live.
• No fuzzy selectors were used.
• No production species/composition decision was locked.

Key findings that now constrain decisions

1. Discriminator exists in schema but is dormant at runtime

species_collection_map has discriminator columns, but all discriminator columns have 0 populated rows and fn_birth_registry_auto does not read them or NEW.unit_kind.

Therefore, a discriminator-based birth-time mapping cannot be relied on today without a future function patch or a post-birth correction tool.

2. law_unit is a vocabulary contract, not live data yet

law_unit exists in dot_config as vocab.unit_kind.law_unit, but no live information_unit.unit_kind row uses it yet and TAC does not carry it.

Phase 5 TAC migration would be the first consumer of law_unit.

3. The 8 PLAUSIBLE candidates are only governance-mode plausible

The dry-run label PLAUSIBLE means entity_species.management_mode matches collection_registry.governance_role=observed. It does not mean semantic suitability is proven.

GPT/User must still choose based on old laws, registry semantics, current pilot data, and future migration intent.

4. Composition follows species

Because composition comes from entity_species.composition_level, choosing one of the 8 plausible species also chooses atom/molecule composition. This is not a separate writeable choice in species_collection_map.

5. unit_version has no peer pattern

unit_version is the only collection with exact birth_code_strategy='subordinate'. There is no registry peer to copy. Any strategy for unit_version must be designed, not inferred.

6. Sandbox schema needs triage, but not inside Phase 4C

A second sandbox_tac.unit_version table exists. Public unit_version is in scope, but sandbox copy should be flagged for a later cleanup/triage pack, not mixed into species mapping.

Decision

Do not seed species mapping yet. Do not backfill birth_registry yet. Do not migrate TAC→IU yet.

Open next step:

P3D_PACK1_PHASE4D_SPECIES_MAPPING_DECISION_MEMO_AND_EXECUTION_OPTIONS

Mode:

DECISION_MEMO + EXECUTION_OPTION_DESIGN_ONLY

No agent dispatch. No DB write.

Required Phase 4D purpose

Phase 4D must convert the dry-run evidence into a council decision memo with options, risks, and recommendation. It must not treat PLAUSIBLE as automatic approval.

It must separate at least three contexts:

1. Current 12 pilot information_unit rows: unit_kind=design_doc_section, observed, no IU containment data.
2. Future migrated TAC law_unit rows: not live yet, likely content/law semantics, TAC nesting depth=2 if preserved.
3. unit_version: subordinate collection with no current birth rows and no peer pattern.

Required no-hardcode rule for Phase 4D

Phase 4D may quote live species codes/names from the report as evidence, but it must not use them as hidden selectors. Every recommendation must say why the species is appropriate under:

• Điều 0-B composition law;
• Điều 0-G birth registry law;
• Điều 29 one-classification-system principle;
• QT-001 backfill procedure;
• QT-005 species/governance promotion if applicable;
• Phase 4C dry-run evidence.

Status

phase4c_dryrun=PASS_ACCEPTED
species_decision_locked=false
composition_decision_locked=false
species_seed_allowed=false
birth_backfill_allowed=false
migration_allowed=false
next_action=OPUS_CREATE_PHASE4D_SPECIES_MAPPING_DECISION_MEMO_AND_EXECUTION_OPTIONS