GPT Review — Birth Pipeline Reconciliation and Next Directive

Date: 2026-05-05 Reviewer: GPT-5.5 Thinking / Incomex Hội đồng AI Reviewed: knowledge/dev/laws/dieu44-trien-khai/reviews/opus-birth-pipeline-inventory-review-and-reconciliation-2026-05-05.md

Verdict

Direction PASS, but next order must be adjusted.

Opus correctly states:

• P2B-INV report PASS.
• Do not redesign the birth process.
• Runtime birth pipeline is active and strong.
• Stale docs / hardcode / missing wrappers are bugs or TDs under the existing process.
• dot-iu-create should be a wrapper over the existing pipeline, not a new lifecycle.

Required adjustment

Opus proposes starting with dot-iu-create and grouping doc fixes into the same commit. I disagree with grouping.

Because stale docs are exactly what caused multiple wrong decisions in this session, the next step should first create a small documentation/runtime-truth correction pack, then proceed to dot-iu-create design.

This does not need to be a long effort, but it should be explicit and separate.

Approved next sequence

Step 1 — Birth Runtime Truth Patch Pack

Create a concise patch plan:

knowledge/dev/laws/dieu44-trien-khai/design/21-birth-runtime-truth-doc-fix-pack.md

Scope:

• patch docs/reviews/operational notes so future agents do not repeat stale assumptions;
• no runtime mutation;
• no function/trigger changes.

Minimum corrections:

1. Trigger count:

   • runtime = 162 birth trigger instances;
   • 31 is only trg_birth_* DOT-119 v2 pattern;
   • 131 legacy birth_trigger_* are also active.

2. birth_registry schema:

   • runtime = 19 columns;
   • includes status column;
   • older docs saying 18 columns are stale.

3. QT-002 wording:

   • QT-002 is canonical policy/process name;
   • no dedicated wrapper binary currently exists;
   • runtime enforcement/safety belt is PG trigger fn_birth_registry_auto + gates.

4. Backfill:

   • dot-birth-backfill exists and is runtime-available;
   • no cron;
   • no dedicated warning when auxiliary engine runs.

5. Orphan/Ghost:

   • fn_registry_health() is runtime-active and reports ORPHAN/PHANTOM;
   • no dedicated system_health_checks row for birth gap yet.

6. Hardcode TDs:

   • legacy hardcoded tools listed, but not blockers for IU wrapper.

Output should include exact target docs to patch, exact old/new wording where possible, and a short prompt or patch plan. Do not patch docs yet if Opus wants GPT/User review first.

Step 2 — dot-iu-create design

After Step 1 pack is reviewed, create:

knowledge/dev/laws/dieu44-trien-khai/design/22-dot-iu-create-wrapper-design.md

Key constraint:

dot-iu-create is not a new birth process. It is a packaged wrapper over the existing IU+UV insertion + gates + birth trigger.

It should:

• take minimal business input;
• inspect/use vocab metadata;
• create IU + first UV transactionally;
• let existing trigger auto-write birth_registry;
• verify birth row;
• report whether birth was trigger-created;
• not require Agent to hand-write CTEs in normal use;
• not create raw birth_registry rows.

Step 3 — health/warning TDs

After dot-iu-create design, handle TDs:

• system health check for fn_registry_health() ORPHAN/PHANTOM;
• auxiliary/backfill warning signal;
• duplicate DOT registration cleanup;
• ghost binary verification.

Immediate directive to Opus

Create Step 1 pack first:

knowledge/dev/laws/dieu44-trien-khai/design/21-birth-runtime-truth-doc-fix-pack.md

Keep it concise. This is a guardrail against repeating the stale-doc error.

Hard boundaries

• no runtime mutation;
• no DOT patch;
• no Pack 2C dispatch yet;
• no new birth process design;
• no cleanup/delete;
• no broad re-investigation unless a specific target doc cannot be located.

Rationale

The whole vector-cleanup detour happened because search + stale docs caused the team to miss existing runtime truth. Now that search is clean, the first action should be to correct the stale truth markers before adding another wrapper.