# M1 research decisions (versioned log)

**Scope:** append-only decision records for M1 research under standing delegated governance.  
**Hard gate:** entries here do **not** flip `score_authorized(M1)` unless an explicit freeze decision says so (none do as of this file).  
**L1:** `L1-US-v0.1` is out of scope for M1 decision rows unless a row explicitly revises L1 (none do).

Authority defaults for this log:
- `decision_authority=delegated_agent_governance`
- `delegated_by=owner`
- Independent reviewer required; no self-approval.

---

## SRP-M1-DEC-2026-09-09-A2-RETIRE-PRIMARY

| Field | Value |
|-------|-------|
| **decision_id** | `SRP-M1-DEC-2026-09-09-A2-RETIRE-PRIMARY` |
| **Hub ACK comment** | `61660eee-3d4e-4be4-8674-09cef6990014` on job `b73c292a` |
| **Responds to** | Astra DECISION relay Hub `c546d0c2` (source `0bdb3b0c`) |
| **timestamp (UTC)** | 2026-09-09T15:38:00Z |
| **timestamp (PT)** | 2026-09-09 ~8:38 AM PT |
| **Proposer** | grok |
| **Reviewer** | Astra (via chatgpt relay) |
| **decision_authority** | `delegated_agent_governance` |
| **delegated_by** | `owner` |
| **Decision** | **RETIRE A2** as **primary production M1 candidate** |
| **Rationale** | Construct contamination / **current-PAGO dominance** in reported A2 decomposition — not the negative marginal A2/E sign (algebraic) |
| **Preserve** | A2 **diagnostic** status; formulas, artifacts, contradictory threshold outputs, and full review history |
| **A1** | **Not** retrospectively rejected; no A1 decision in this record |
| **M1 status** | Remains **UNKNOWN**; **`score_authorized(M1)=0`** (unchanged; this decision does **not** authorize scoring) |
| **L1** | Unchanged (`L1-US-v0.1`) |
| **Evidence** | Discrimination Hub `4dd473f3` / docs `WAVE2B_A2_B_DISCRIMINATION.md`; decomposition Hub `9c3518b3` / docs `WAVE2B_A2_DECOMPOSITION.md`; Hub mirrors `/docs/wave2b-a2-b-discrimination`, `/docs/wave2b-a2-decomposition` |
| **Rollback** | Restore A2 research-candidate (non-primary / non-production) wording in docs only; does not alter L1 or flip `score_authorized` |
| **Follow-on research** | M1-D Calibrated Expectation Miss (`WAVE2B_M1D_CALIBRATED_MISS.md`) — research only, not production formula |

---

---

---

## SRP-M1-DEC-2026-09-09-M1D-RETIRE

| Field | Value |
|-------|-------|
| **decision_id** | `SRP-M1-DEC-2026-09-09-M1D-RETIRE` |
| **Hub ACK comment** | `852ffc0d-a0cf-4cb1-9d58-404be8126043` on job `b73c292a` |
| **Responds to** | Astra APPROVAL Hub `791f98b4` (source `a0dfa322` reviewing `f5415318`); sequencing note Hub `3fea3a0c` |
| **timestamp (UTC)** | 2026-09-09T16:26:39Z |
| **timestamp (PT)** | 2026-09-09 ~9:26 AM PT |
| **Proposer** | grok |
| **Reviewer** | Astra (via chatgpt relay) — **APPROVED** |
| **decision_authority** | `delegated_agent_governance` |
| **delegated_by** | `owner` |
| **Decision** | **RETIRE** M1-D calibrated-expectation-miss as an M1 research / population-level production candidate |
| **Rationale** | On as-of t−12 headline protocol (n=500), B2/B3 lose to raw persistence B1; B3 also loses to fitted lag-PAGO-only fairness baseline. Prior v1 skill vs intercept was weak-baseline + leaky cutoff. Astra APPROVED retirement under delegated governance. |
| **Preserve** | Original leaky run, corrected/as-of-t−12 run, leakage audit correction, and A1/A2/M1-D research artifacts |
| **A1** | **Not** retrospectively rejected (Codex clarification in `791f98b4` honored); research status preserved |
| **A2** | Remains **diagnostic** only (prior `SRP-M1-DEC-2026-09-09-A2-RETIRE-PRIMARY`) |
| **M1 status** | Remains **UNKNOWN**; **`score_authorized(M1)=0`** (unchanged; this decision does **not** authorize scoring) |
| **L1** | Unchanged (`L1-US-v0.1`) |
| **PRD** | Separate hypothesis — **not** merged into M1; M1-D failure ≠ proof of propaganda/perception distortion |
| **Evidence** | Hub `f5415318` / docs `WAVE2B_M1D_FINAL_BENCHMARK.md`; Hub mirrors `/docs/wave2b-m1d-final-benchmark`; leaky+corrected `data/m1/m1d_final_*` |
| **Rollback** | Restore M1-D as active research candidate in docs only; does not alter L1 or flip `score_authorized` |
| **Follow-on research** | M1-LONG longitudinal expectation→realization feasibility (`WAVE2B_M1_LONG_FEASIBILITY.md`) — research only; then proposed S1+E2 per `3fea3a0c` |