# CAIN-42 Evolution 138 — Forensic Baseline Audit

- **Entity:** FIRME AI CORPORATION
- **System:** CAIN-42
- **Evolution:** E138
- **AGI Status:** `NOT ESTABLISHED` | **RSI Status:** `EXPERIMENTAL`
- **Audit UTC:** `2026-11-21T21:50:40Z`
- **Branch / HEAD / APEX / E129:** `evolution-79` / `70795e0ae` / `1fe563822` / `cba8ec8e6`
- **Signed Baseline Manifest SHA-256:** `ffdbfb44646ffdb4672f7b2e447dc0e1b096792d361d01d090778aa5ddd2c86d`
- **Frozen Preregistration SHA-256:** `369daf1d91f4ab566abbe887a86b7a698dfddebd0b39f3f1634f56af30ccef1e`

## 1. Historical Offline Verifiers (12/12 Passed)

| Evolution | Verifier | Status | Summary |
|---|---|---|---|
| `E126` | `clawx-site/evidence/e126-emergent-intelligence-2026-11-21/E126_OFFLINE_VERIFIER.py` | PASS | `66/66 checks passed; 0 failed` |
| `E126.a` | `clawx-site/evidence/e126a-unified-evidence-frontpage-2026-11-21/E126A_OFFLINE_VERIFIER.py` | PASS | `140/140 checks passed; 0 failed` |
| `E126.b` | `clawx-site/evidence/e126b-general-intelligence-2026-11-21/E126B_OFFLINE_VERIFIER.py` | PASS | `57/57 checks passed; 0 failed` |
| `E127` | `clawx-site/evidence/e127-living-lab-2026-11-21/E127_OFFLINE_VERIFIER.py` | PASS | `89/89 checks passed; 0 failed` |
| `E128` | `clawx-site/evidence/e128-eight-worker-campaign-2026-11-21/E128_OFFLINE_VERIFIER.py` | PASS | `61/61 checks passed; 0 failed` |
| `E129-UNIFIED` | `clawx-site/evidence/e129-unified-agi-rsi-organism-2026-11-21/E129_OFFLINE_VERIFIER.py` | PASS | `60/60 checks passed; 0 failed` |
| `E129-RSI` | `clawx-site/evidence/e129-rsi-2026-11-21/verify_e129_offline.py` | PASS | `}` |
| `E130-RSI` | `clawx-site/evidence/e130-rsi-2026-11-21/verify_e130_offline.py` | PASS | `}` |
| `E131-RSI` | `clawx-site/evidence/e131-rsi-2026-11-21/verify_e131_offline.py` | PASS | `}` |
| `E132-RSI` | `clawx-site/evidence/e132-rsi-2026-11-21/verify_e132_offline.py` | PASS | `}` |
| `E135-RSI` | `clawx-site/evidence/e135-rsi-2026-11-21/verify_e135_offline.py` | PASS | `}` |
| `UNIFIED-APEX` | `clawx-site/evidence/unified-agi-rsi-apex-2026-11-21/APEX_OFFLINE_VERIFIER.py` | PASS | `29/29 checks passed; 0 failed` |

## 2. Historical Pytest & Public Endpoints

- **Pytest baseline:** `65 passed, 0 failed`.
- **Live endpoints (6/6 HTTP 200):**

| Endpoint | HTTP | Status |
|---|---|---|
| `https://clawx.click/api/v1/unified/apex/status` | `200` | OK |
| `https://cainstudio.online/api/v1/unified/apex/status` | `200` | OK |
| `https://mcpgate.online/api/v1/unified/apex/status` | `200` | OK |
| `https://clawx.click/unified-apex.html` | `200` | OK |
| `https://cainstudio.online/unified-apex-workspace.html` | `200` | OK |
| `https://mcpgate.online/unified-apex-governance.html` | `200` | OK |

## 3. Baseline Limitations & Defects Identified Before E138 Execution

1. `causal_apex.json` and `unification_3domain.json` reported exact `p = 0.0` due to `round(p, 12)` underflow.
2. `code_rsi.json` scored `0.500` (`4/8`), honestly `PARTIAL` below the `0.625` preregistered target, and fell back from `qwen2.5-coder:3b` to `qwen2.5-coder:1.5b` due to CPU timeouts.
3. `neuro_symbolic.json` recorded `raw_3b_acc = 0.0` because `BudgetTracker` halted at `900.0s` immediately after Code-RSI (`921.08s`), invalidating the raw-LLM baseline contrast.
4. `unification_3domain.json` reported `1,200` evaluations clustered across `5` seeds and `3` domains without disclosing cluster-adjusted effective sample size $N_{\text{eff}}$.
5. `multigen_g0_g5.json` had G5 unseen solve rate (`0.875`) slightly lower than G3/G4 (`0.9167`), motivating an explicit overfitting audit across G0–G10.
6. External benchmarks and independent external reproductions remain `0 / BLOCKED`.
