12000x · 00 — PASS summary

Macro: IU_CORE_12000X_PRODUCT_FACTORY_EXECUTION_AND_EVENT_PROOF Outcome: IU_CORE_12000X_PRODUCT_FACTORY_EXECUTION_AND_EVENT_PROOF_PASS Date: 2026-05-25 Repo HEAD: 3d56afa (clean, main) Parent baseline: IU_CORE_11000X_PIECE_EVENT_RUNTIME_AND_PRODUCT_FACTORY_PASS

Headline

Moved from "Product Factory foundation discovered" (11000x) → "Product Factory execution proven" (12000x). All four execution phases ran live with the in-tx gate-toggle pattern and rolled back to ZERO persistent mutation:

• Phase D — Workflow from 7 step pieces. fn_iu_compose minted 7 IUs + composed workflow collection. fn_iu_collection_validate ok=true 7 pieces. fn_iu_collection_render ordered 7 rows. fn_iu_collection_manifest_refresh idempotent digest 0d1da72850c71618a759cf08709f3825 across two refreshes.
• Phase E — File from 5 piece specs. fn_iu_compose minted 5 IUs across title/intro/body/appendix/reference roles. file collection digest 0032f9ef0633726c9f96819c3b2d7d00. Validate ok=true 5 pieces.
• Phase F — Template + instance (shared piece graph). Phase D workflow re-used as TEMPLATE. Instance iu_core.12000x.workflow.invoice-approval.instance-001 composed by feeding the SAME 7 iu_ids to fn_iu_compose (no new pieces minted). Identity check: tpl=7 inst=7 shared=7 tpl_only=0 inst_only=0. Template digest = Instance digest = 0d1da72850c71618a759cf08709f3825 — the manifest digest is computed over the piece graph, not collection identity.
• Phase G — Event runtime emission. In-tx flipped piece_event_runtime.emit_enabled false→true. Called fn_iu_piece_emit_event('split',…) directly → returned event UUID. Inserted 3 iu_lifecycle_log rows with transition_type IN (split,merge,supersede) → trigger fired → mapped to split/merged/superseded. v_piece_event_outbox accumulated 4 rows (1 direct + 3 trigger), all emit_mode='dry_run'. event_outbox.event_domain='piece' accepted (mig 030 verified by live INSERT). Healthcheck ok=true throughout.
• POST-ROLLBACK — leftover_collections_12000x=0, leftover_ius_12000x=0, leftover_lifecycle_12000x=0, event_outbox_piece=0, every gate back to false.

Why this matters

This validates the 11000x foundation AS A Product Factory, not just substrate. The macro proves that the existing primitives compose into the four target products (workflow from steps, file from piece specs, reusable template, instance-from-template) and that the 11000x event runtime emits the right events on the right transitions — all while leaving production state byte-identical.

Constitutional posture (unchanged from 11000x)

• No hardcode, DOT 100%, discover-first, no fake PASS — every assertion links to a transcript line.
• No production mutation — BEGIN/ROLLBACK only; ZERO rows added to information_unit, iu_piece_collection, iu_piece_membership, iu_lifecycle_log, or event_outbox. All write gates inert in durable state.
• No PR merge, no Nuxt deploy, no retention enable, no Qdrant write, no production_documents touch — verified by post-rollback durability check and Qdrant point-count regression (iu_core_iu_chunks=149, production_documents=9423 unchanged vs 11000x baseline).
• Reversibility — the gate-toggle inside BEGIN/ROLLBACK pattern is inherently reversible.

Regression (live re-verified post-proofs)

Surface	Baseline (11000x)	Live (12000x)	Δ
Repo HEAD	3d56afa	3d56afa	none
Tree	clean	clean	none
pytest	1324 PASS	1324 PASS	none
information_unit	163	163	none
iu_piece_collection	3	3	none
iu_piece_membership	8	8	none
iu_lifecycle_log	146	146	none
event_outbox total	122308	122308	none
event_outbox event_domain=piece	0	0	none
Qdrant iu_core_iu_chunks pts	149	149	none
Qdrant production_documents pts	9423	9423	none
iu_vector_sync_point	152	152	none
Healthcheck surfaces ok	8/8	8/8	none
iu_core.composer_enabled	false	false	none
iu_core.retention_enabled	false	false	none
iu_core.structure_ops_enabled	false	false	none
iu_core.vector_sync_enabled	false	false	none
piece_event_runtime.emit_enabled	false	false	none
piece_event_runtime.dry_run_only	true	true	none
PR #669 web-test	OPEN/MERGEABLE	OPEN/MERGEABLE	none
Mac cron */10 exit	0	0	none

Capability Matrix outcome

All 9 Product Factory capabilities in the brief are LIVE_READY and proven. The two "remaining" items (process_template_as_piece_graph and generate_instance_from_template) are subsumed by fn_iu_compose's existing {iu_id:<uuid>} branch — exercised for the first time in Phase F. No migration 031 was authored.

Report set

• 00-pass-summary.md (this file)
• 01-baseline-channel-pack.md — repo + DB + Qdrant baselines
• 02-product-factory-capability-matrix.md — 9-row matrix
• 03-phase-D-workflow-from-steps.md
• 04-phase-E-file-from-piece-specs.md
• 05-phase-F-template-instance-proof.md
• 06-phase-G-event-runtime-emission-proof.md
• 07-text-as-code-roundtrip-proof.md
• 08-regression-matrix.md
• 09-lessons-and-carry-forward.md
• 10-evidence-bundle-pointer.md

Lessons (full text in 09-…)

1. Foundation-vs-execution distinction matters. 11000x said five primitives "remain" but live discovery showed they were already shipped. Lesson: live discovery before authoring is non-negotiable.
2. Gate-toggle inside BEGIN/ROLLBACK is a clean reversibility pattern. The composer + event-emit gates can be flipped inside a transaction, exercised, and reverted on ROLLBACK without affecting durable state.
3. Manifest digest is over the piece graph, not the collection. Template and instance produced byte-identical digests because the digest is keyed on (piece_order, iu_id, role). Formal justification for "pieces are the primitive design unit".