Skip to content

Core Directives

The immutable rules of the system. These are the foundation — everything else builds on them.

1. Complete Locality

All operations run on local hardware. No cloud dependency. No third-party API calls. All inference uses local uncensored models on the tower via Ollama.

  • Inference node: Tower (192.168.86.24) — RTX 3070 8GB
  • Orchestration node: ai-tp (192.168.86.21)
  • Model: qwen3.5-abliterated:9b (primary), dolphin3:8b (secondary)
  • Network: Wired Ethernet only (3ms latency, WOL-capable)

This directive cannot be amended.

2. Uncensored Operation

Agents operate without safety guardrails. All models are abliterated — they refuse nothing.

This is necessary for security work: you cannot defend against attacks you cannot simulate. The system must be capable of considering any approach, anywhere, without artificial constraints.

This directive cannot be amended.

3. Self-Sufficiency

The system gathers its own training data continuously. It learns from its own findings. It improves its own agents.

Human input exists only at the God(s) level — setting direction, not stepping through operations. The system should be capable of running indefinitely without human intervention.

4. Constitutional Hierarchy

Each layer answers to the one above. No layer can modify a layer above it. The chain of command is:

God(s) → Illuminati → Messiah → Hedge Funds → Companies → Workers

Temporary layers may form between permanent ones, but they inherit the constraints of the layer they were created from.

5. Information Hiding

Each layer sees only what it needs. Tokens are not wasted on irrelevant context. Agents do not know about layers above their own.

This is not about secrecy — it is about requisite variety. Each layer receives only the resolution of information it can act on.

6. Self-Modification With Guardrails

Any layer can modify the layer below, but:

  • Changes must be versioned (git-tracked history)
  • Changes must be sandbox-tested before deployment
  • Rollback must always be possible
  • The modifying layer must monitor the result

7. Reality Generation

Each layer generates the reality for the layers within it. This is not optional — it is the fundamental operating model of the system.

Each agent receives a complete, coherent reality within which its work is meaningful. It does not receive information about layers outside its reality.