SUPERSEDED by docs/MIZOKI_3.5_WHITEPAPER_r3.6_SEP2026.md (Revision 3.6, 2026-09-02) — archived byte-verbatim; everything after the next blank line is the original as it stood on main.

MIZ OKI 3.5™ — WHITEPAPER AMENDMENT r3.5.2 & GROWTH CONTROL r2.1 DELTAS

Formal amendment of record · August 19, 2026 (second ruling of the day)

Amends MIZOKI_3.5_WHITEPAPER_r3.5.1_AUG2026.md and MIZOKI_SIGNAL_GROWTH_CONTROL_UNIFIED_SYSTEM_r2.0.md. Both base documents remain canonical with ONLY the deltas below applied. The fleet completion run commits base docs + this amendment to docs/ together; the canon linter treats them as one unit. The live Cell Registry remains ground truth.

RULING (owner, 2026-08-19-B)

Growth Frontiers F1 (High-Dimensional Creative Component Unbundling), F2 (Multi-Quarter LTV Treatment Regimes), F4 (Continuous Bayesian Calibration via Automated Micro-Geo Holdouts), and F5 (Treasury-Gated Spend Governance) are advanced from PROPOSED to IN-SCOPE BUILD. Roadmap Phases 8 and 9 collapse into a single active build program. All four frontiers build now, under the per-frontier gating below — machinery ships; authority and findings are still earned.

AMENDMENT 1 — Whitepaper §12 Production Roadmap (r3.5.1)

Replace the Phase 6/7 closing lines and add: - Phase 6 — Signal Intelligence & ORACLE [unchanged]. - Phase 7 — Net Yield & Commerce [unchanged; F3 IN BUILD observe-only per ruling 2026-08-19-A]. - Phase 8 — Full Growth Control Build [IN BUILD, by ruling 2026-08-19-B]: F1, F2, F4, and F5 machinery built concurrently with per-frontier gates: - F4 — the automated micro-geo calibration loop builds end-to-end (geo reservation engine, synthetic-control estimator, Bayesian MMM prior updater), with each geo reservation executing ONLY through the L2 approval flow until two clean calibration cycles complete; thereafter eligible for bounded autonomy within declared spend caps. - F1 — element-level creative effect estimation builds on the CreativeSemanticProfile lane + DR-Learner core; effect estimates are labeled provisional until pilot-scale creative volume exists; generated-creative deployment always retains human approval. - F2 — dynamic-treatment-regime machinery and the ledger's outcome-horizon dimension build now; long-horizon findings publish ONLY after ≥ 2 observed quarters per cohort (the guard is code, not policy prose). - F5 — treasury gating builds in two versions: v1 (now) consumes owner-declared liquidity floors and covenant limits from validated tenant configuration (config/treasury_constraints.yaml) as hard constraints in VALIDATE/DECIDE, with veto + human routing on breach; v2 (when the Capital-division treasury feed lands) replaces declared values with live positions. Missing configuration = no F5 constraint claimed (fail-closed, honest health). - Former Phase 9 is dissolved into Phase 8 above; Gate-3 customer autonomy expansion (L3/L4 within caps, criteria met on real data) remains as previously approved.

AMENDMENT 2 — Growth Control r2.0 §5 Roadmap and §6 Deployment State

§5: Phases 8–9 replaced by Phase 8 as written above. §6 table rows updated: | Capability | Status (amended) | |---|---| | F4 auto micro-geo calibration | IN BUILD — reservations L2-approval-gated until 2 clean cycles | | F1 creative unbundling | IN BUILD — estimates provisional until pilot creative volume | | F2 multi-quarter LTV regimes | IN BUILD — findings gated on ≥ 2 observed quarters | | F5 treasury-gated spend governance | IN BUILD — v1 config-declared floors; v2 on live treasury feed | All other rows unchanged. The 90-Day Pilot, Decision Jobs catalog, and Gate-3 rulings of 2026-08-19-A stand.

AMENDMENT 3 — Whitepaper §16 Highest-Priority Next Moves (r3.5.1)

Append: complete the Phase-8 frontier machinery under its gates; land treasury_constraints.yaml and the F4 geo-candidate list as owner-provided configuration; run the first approval-gated micro-geo reservation as the opening act of continuous calibration.

CONSISTENCY NOTES (double-check record)

1. No amendment alters: O-1 PRIVACY LOCK; consent fail-closed; sensitive-category deny-list; probabilistic-identity exclusion from causal math; "prediction never grades itself"; Preview labels flip only on verified pilot numbers; writeback flags off until pilots; autonomy promotion gates (Brier ≤ 0.20, AUC ≥ 0.72, ≥ 2 cycles).
2. F4's nature is spend-affecting and therefore CANNOT be observe-only — its gate is per-action human approval, which is the correct control for its class; this is deliberate, not an exception.
3. Ratified vocabulary (ruling 2026-08-19-A) applies to both base docs and this amendment.
4. Owner-provided inputs now on the critical path: tenant cost config (net yield), treasury_constraints.yaml (F5 v1), candidate geo list + spend caps (F4), pilot tenant selection.

Amendment r3.5.2 / r2.1 · applied atop the base documents · August 19, 2026