# E128 Forensic Baseline Audit

**Entity:** FIRME AI CORPORATION · **System:** CAIN-42 · **Evolution:** E128
**Generated:** 2026-11-21T17:20:55Z · **AGI Status at Start:** NOT ESTABLISHED
**Baseline Status:** PASS · **Manifest SHA-256:** `5ad71ce24bd68628ac6893a6165c4d7eba8c6e53c2120c53ade8f13c7f6bb486`

## 1. Verified Immutable Git Baselines

| Ref | Verified | Full Commit |
|---|---|---|
| `head` (`HEAD`) | PASS | `7404d4ac933d58f2dc1f345bcff069cd7bbf29ab` |
| `required_baseline_e127` (`c85cd409f`) | PASS | `c85cd409fa7a7164af2363cb22ed90fdc247987e` |
| `implementation_e127` (`47455dd77`) | PASS | `47455dd77b7c673122ccea6891f31cbb178c0109` |
| `previous_intelligence_e126b` (`b9b722b20`) | PASS | `b9b722b20082d24036840ec4ca342c1cc9d10c11` |

## 2. Offline Verifiers & Regression Baseline

| Suite | Result | Summary |
|---|---|---|
| E127 Offline Verifier | PASS | 84/84 checks passed; 0 failed |
| E126B Offline Verifier | PASS | 57/57 checks passed; 0 failed |
| E126A Offline Verifier | PASS | 140/140 checks passed; 0 failed |
| E126 Offline Verifier | PASS | 66/66 checks passed; 0 failed |
| Pytest E126–E127 Regression | PASS | ============================== 32 passed in 5.20s ============================== |

## 3. Hardware, Compute & Model-Provider Configuration

- **Host:** `Linux-7.0.0-43-generic-x86_64-with-glibc2.43`, `8` vCPUs, `30.33 GiB` RAM, `48.28 GiB` free disk, **GPU:** `False`
- **Python:** `3.14.4` (scipy `1.18.1`, numpy `2.5.3`, fastapi `0.136.3`, cryptography `50.0.2`, playwright `None`)
- **Local Ollama:** `available=True` at `http://172.17.0.94:11434` ($0.00/token) with models: `qwen2.5-coder:3b` (3.1B Q4_K_M, `f72c60cabf6237b0`), `qwen2.5-coder:1.5b` (1.5B Q4_K_M, `d7372fd828518a4d`), `qwen2.5:0.5b` (494.03M Q4_K_M, `a8b0c51577010a27`), `qwen2.5:1.5b` (1.5B Q4_K_M, `65ec06548149b04c`)
- **Paid Cloud Models:** `ANTHROPIC_API_KEY` present in `.env` but **explicitly NOT USED** (paid campaign spend budget = `$0.00 USD`).

## 4. Live Websites, APIs & Hugging Face Space Audit

| URL | HTTP | E127 Strip | AGI: NOT ESTABLISHED | Legacy Proto-VAGI Mentions |
|---|---:|---:|---:|---:|
| `https://clawx.click/` | 200 | True | True | 10 |
| `https://cainstudio.online/` | 200 | True | True | 9 |
| `https://mcpgate.online/` | 200 | True | True | 10 |
| `https://clawx.click/api/v1/research/health` | 404 | - | - | - |
| `https://cainstudio.online/api/v1/research/health` | 404 | - | - | - |
| `https://mcpgate.online/api/v1/research/health` | 404 | - | - | - |

- **Hugging Face Space (`integrations/huggingface-space/README.md`):** `STALE_NOT_UPDATED_SINCE_E123` (mentions E123=True, E127=True).

## 5. Recorded Limitations, Unresolved Gates & Contradictory Public Claims

- Live /api/v1/research/* endpoints return HTTP 404 on clawx.click, cainstudio.online, and mcpgate.online because cain-platform-gateway.service has been running since 2026-11-21T04:32:34Z (prior to E127 commit at 16:25:35Z) and has not been restarted.
- Live homepages display the E127 'AGI: NOT ESTABLISHED' strip at the top while still containing legacy 'Proto-VAGI' sections (21-24 occurrences) and older 'AGI Status: NOT ACHIEVED' wording lower down the page.
- Hugging Face Space (integrations/huggingface-space/README.md) still reflects E123/Proto-VAGI and has not been updated or pushed for E127.
- E127 workers were deterministic artifact-auditing scripts (0 LLM inference calls), running 60 tasks over 177.87s on a single-host SQLite queue with peak 5/8 workers online.
- E126.b scientific gaps remain open: causal active selection FAIL (paired 25/25/30 vs random), novel-environment H1 missed (0.895 vs retrieval 0.860, p=0.210), continual learning tied retrieval/episodic (0.6667), adaptation retention FAIL (0.375 -> 0.0), metacognition abstention 0.0, external benchmarks and human review BLOCKED.
- Host is CPU-only (8 vCPUs, 30 GB RAM, no GPU, ~49 GB free disk). No headless browser (chromium/playwright) is installed on this host.
- ANTHROPIC_API_KEY is present in .env but NOT authorized for paid campaign spend; E128 paid budget is strictly $0.00 USD and uses only local free Ollama models (ollama/ollama:0.12.3).
