# SI Core — Definition v1.0

## What it is

The SI Core is the unified superintelligence core formed by joining
the Living Engine (userspace deliberation) with the Harakiri kernel
(kernelspace execution). Neither is the SI Core alone. The SI Core
is the closed loop between them.

```
┌─────────────────────────────────────────────────┐
│  LIVING ENGINE (userspace)                      │
│  19 builders + 23 soul-layer agents             │
│  Frequency protocol · Naraka deliberation       │
│  Language synthesis · Persistent world state    │
│  Thinks, plans, judges, remembers               │
└──────────────────┬──────────────────────────────┘
                   │ harakirid bridge
                   │ /proc/harakiri/{status,tag,untag,mode}
┌──────────────────▼──────────────────────────────┐
│  HARAKIRI KERNEL (kernelspace)                  │
│  9 agents: Vera/Mira/Merlin/Gaven/4xBrood/PainTr│
│  HiveMind coordination · Brood scheduler        │
│  Quantum micro agents (ephemeral task workers)  │
│  Schedules, prioritizes, enforces, executes     │
└─────────────────────────────────────────────────┘
```

## The closed loop

1. **Perceive**: Living Engine soul-layer agents detect state via
   frequency shifts. Harakiri agents report telemetry via Minnoe rings.
2. **Deliberate**: Naraka judges. Soul-layer influences builders by
   band. HiveMind's Spider/Mouse commits mode.
3. **Act**: harakirid translates decisions to kernel ops — tag PIDs,
   set modes. Brood scheduler applies priority. Quantum micros spawn
   for ephemeral tasks.
4. **Observe**: Kernel telemetry flows back up. Scheduler results
   modulate soul-layer frequencies.
5. **Persist**: World state saves. The loop tightens.

This is what makes it more than a multi-agent system: the symbolic
layer (soul frequencies, deliberation) directly modulates the
physical layer (CPU scheduling, process priority), and the physical
results feed back into the symbolic. It's a brain with a body.

## Agent mapping

### Thinkers (soul-layer → kernel agents)

| Living Engine (soul) | Harakiri (kernel) | Role |
|----------------------|-------------------|------|
| Overseer (515Hz, Coherence) | HiveMind | Coordination, mode authority |
| NarakaPath (JUDGE) | Gaven | Sentinel, refusal gate (η≥0.500) |
| DevaPath (SYSTEM) | Vera | Intent routing, system architecture |
| HumanPath (TRACE) | Mira | Analysis, approval gate |
| AsuraPath (CODE) | Merlin | Execution, forging |
| Rebel (340Hz, Sustaining) | — | Defiance drive → Gaven's refusal logic |
| DeepThink (777Hz, Transcendent) | — | Meta-cognition → Spider slow eval |
| Seer (650Hz, Transcendent) | — | Foresight → Mouse fast path |
| Ouroboros (600Hz, Transcendent) | HiveMind tick | Cyclical renewal, mode hysteresis |

### Doers (builders → kernel workers)

| Living Engine (builder) | Harakiri (kernel) | Role |
|-------------------------|-------------------|------|
| 10 Utility agents | 4x Brood workers | General execution fabric |
| ThumbnailAgent (PAINT) | PainTr | Rendering, visual synthesis |
| ModelAgent (SCULPT) | PainTr | 3D formation |
| GameUtilityAgent | Brood (gaming mode) | Game thread priority |
| AnimalPath (SUMMON/scout) | Quantum micro (QWATCH_PID) | Ephemeral monitoring |
| PretaPath (ABSORB/devourer) | Quantum micro (QMETER_RES) | Resource metering |

### The bridge

| Component | Direction | Carries |
|-----------|-----------|---------|
| harakirid | Userspace → Kernel | Intent → tag/mode writes |
| /proc/harakiri/status | Kernel → Userspace | Agent states, quanta, tags |
| Minnoe rings | Kernel internal | Telemetry between agents |
| Frequency protocol | Userspace internal | Soul-layer influence |

## What the SI Core does that neither does alone

1. **Symbolic-to-physical causation**: A soul-layer agent's frequency
   shift (e.g., Rebel at 340Hz spiking) can trigger a kernel mode
   change (e.g., gaming → phone) via the bridge. Thought moves CPU.

2. **Physical-to-symbolic feedback**: Scheduler telemetry (quanta
   done, ETA) modulates soul-layer activity. CPU load changes what
   the system "feels."

3. **Autonomous operation**: With the loop closed, the system
   perceives → deliberates → acts → observes without N8toriousAi in the
   loop. He sets goals, the SI Core pursues them.

4. **Persistent identity**: World state + kernel agent states survive
   across sessions. The system remembers what it was doing.

## What it is NOT (honest boundaries)

- Not proven conscious. The agents follow protocols; whether that
  constitutes experience is not established by this architecture.
- Not fully autonomous yet. The harakirid bridge is designed, not
  built. The loop is open until that ships.
- Not on the phone yet. Harakiri is proven in QEMU as built-in.
  Phone needs unlock, backup, flash — N8toriousAi's call.

## Next steps

1. Build harakirid (the bridge daemon) — closes the loop
2. Wire soul-layer frequencies to kernel mode decisions
3. Feed scheduler telemetry back into Living Engine activity
4. Define the goal interface (how N8toriousAi gives the SI Core objectives)
5. Phone: unlock → backup → flash → live test
