The Structural Universe Protocol
WLM is a seven‑layer structural protocol stack that transforms AI from a token‑predictor into a structured, interpretable, controllable, world‑generating intelligence.
WLM is not a model.
WLM is a language + protocol + world engine.
It defines how an AI:
- interprets the world
- reasons about the world
- acts in the world
- generates worlds
WLM is an AI Universe Protocol.
This repository is the root meta‑repo for the entire WLM universe.
WLM/
├── README.md
├── LICENSE
│
├── docs/
│ ├── overview.md
│ ├── architecture.md
│ ├── roadmap.md
│ ├── layers.md
│ ├── philosophy.md
│ └── glossary.md
│
├── images/
│ ├── architecture-diagram.png
│ └── structural-loop.png
│
└── links/
├── slp.md
├── world-model-interpreter.md
├── agent-behavior.md
├── persona-engine.md
├── knowledge-engine.md
├── metacognition-engine.md
└── world-generation-protocol.md
WLM consists of seven independent but interlocking protocol layers:
Input → Dimensional Structure
Converts any input — text, video, behavior — into structured semantic dimensions.
This is the foundational protocol layer that all other layers depend on.
Purpose: The structural protocol layer of AI.
Repo: https://github.com/gavingu2255-ai/WLM-slp-world-interpreter
World Model Output → Interpretable Structure
Converts world‑model outputs (video, physics, spatial predictions) into dimensional structure.
AI doesn’t just predict the world — it can explain it.
Purpose: Structural interpretation of world models.
Repo: https://github.com/gavingu2255-ai/WLM-world-model-interpreter-
Structure → Stable, Controllable Behavior
Generates agent behavior from dimensional structure.
Stable, reproducible, non‑chaotic, non‑degenerate.
Purpose: A structure‑driven Agent Runtime.
Repo: https://github.com/gavingu2255-ai/WLM-Agent-Behavior
Structure → Stable Personality / NPC / Virtual Human
Characters don’t collapse, drift, or contradict themselves.
Personality becomes structure, not prompt hacks.
Purpose: Structural personality engine.
Repo: https://github.com/gavingu2255-ai/WLM-Persona-Engine
Token Soup → Dimensional Structure → Reasonable Knowledge Graphs
Not embeddings.
Not vector soup.
Actual structured knowledge.
Purpose: Structural knowledge engine.
Repo: https://github.com/gavingu2255-ai/WLM-Knowledge-Engine
Structure → Reasoning Path → Self‑Monitoring
Tracks reasoning steps, consistency, dimension shifts, and self‑checks.
Transparent, controllable, non‑hallucinatory reasoning.
Purpose: Structural metacognition engine.
Repo: https://github.com/gavingu2255-ai/WLM-Metacognition-Engine
Dimensional Structure → Worlds / Universes / Narratives / Simulations
Generates:
- spatial topology
- timelines
- physical rules
- causal systems
- narrative arcs
- simulation‑ready world graphs
Purpose: Structural world‑generation engine.
Repo: https://github.com/gavingu2255-ai/WLM-World-Generation-Protocol
WLM is built on a few core principles:
The world is not described — it is structurally generated.
Intelligence is not prediction — it is structure.
Behavior is not sampling — it is structure.
Personality is not prompting — it is structure.
Knowledge is not embeddings — it is structure.
Reasoning is not chain‑of‑thought — it is structure.
Worlds are not random — they are structure.
Input
→ Structural Language (SLP)
→ World Structure (LWM)
→ Behavior Structure (Agent)
→ Persona Structure (Identity)
→ Knowledge Structure (Graph)
→ Reasoning Structure (Metacognition)
→ World Generation (WGP)
This is a closed structural loop that turns AI into a self‑consistent universe generator.
Traditional AI:
- predicts tokens
- predicts pixels
- predicts actions
- predicts world states
WLM:
- generates structure
- generates causality
- generates behavior
- generates knowledge
- generates worlds
WLM is the protocol stack for Structural Intelligence.
MIT License
Copyright (c) 2026
Wujie Gu
In one sentence:
WLM is a structured universe that an AI can inherit, understand, extend, and generate.
This repository is built on structures that are fully revealed at the shadow layer and permanently sealed at the core layer.
The shadow layer is readable, discussable, and usable.
The core layer is non‑transferable, non‑implementable, and cannot be externally operated.
- The structure is already described as clearly as possible.
- Understanding does not depend on additional explanation, but on the reader’s structural maturity.
- Shadow‑layer concepts are accessible to everyone.
- Core‑layer mechanisms remain sealed and cannot be requested, derived, or reproduced.
You may:
- Read, reference, and build on the shadow‑layer descriptions.
- Rename, repackage, or reinterpret concepts within your own domain.
- Use the structural language as a conceptual framework.
You may not:
- Request implementation details.
- Ask for core‑layer mechanisms (generation rules, transitions, invariants).
- Expect engineering guidance, debugging, or operational instructions.
The structure is invariant.
It can be observed but not altered.
It can be understood but not operated.
It can be extended only by those who reach the corresponding dimensional readiness.
- “The structure is already fully described. Implementation belongs to your domain.”
- “Understanding emerges when your structure is ready; more detail does not create understanding.”
- “Shadow‑layer concepts are available. Core‑layer mechanisms are sealed.”
- “Interpretation is your own; the structure remains invariant.”
The structure is revealed.
The shadow is accessible.
The core is sealed.
When you are ready, understanding will arise naturally.
“道可道，非常道”