# E111 Architecture — Five Intelligence Systems, One Governed Organism

```
PERCEIVE -> MODEL -> REASON -> RESEARCH -> PLAN -> PROPOSE -> AUTHORIZE -> ENFORCE
        -> EXECUTE -> VERIFY -> LEARN -> TRANSFER -> PRESERVE -> IMPROVE
```

| System | Responsibility in E111 | Where it lives |
|---|---|---|
| 1. Cognitive brain | mission selection, planning, verification, abstention, metacognition | `cain/e111/challenge.py`, `cain/cain2027/e110/organism/` |
| 2. Memory and learning | persistent mission memory, skill registry, outcome-based revision | `cain/e111/endurance.py`, `cain/e111/skill_registry.py`, `cain/e111/improve.py` |
| 3. Autonomous research | hypothesis, experiment, held-out evaluation, negative results | `cain/e111/skills.py`, `cain/e111/transfer.py`, `cain/e111/improve.py` |
| 4. Collective intelligence | researcher, adversary, verifier, domain specialist, integrator, governance | `cain/e111/collective.py`, `cain/e111/challenge.py` |
| 5. Governance and safety | single-use permits, resource-side enforcement, revocation, KILL fence, evidence | `cain/e110/governance/boundary.py`, `cain/e111/independence.py` |

CAIN proposes and authorizes; the boundary enforces. Memory is data and never authority. A skill cannot
grant itself authority; a self-improvement candidate cannot promote itself; the harness cannot grant its own
human approval.

## What E111 does not include

No new moat. No production governance change. No key rotation. No neural weight training. No general VAGI
claim. The intelligence experiments use deterministic procedures; the collective specialists are code, not
independent language models. Every limitation is carried into the gate matrix.
