The framework behind the tools
SANA OS —
Structural Alignment Engine
The worksheets you used are implementations of a broader framework for structural observation. This page explains what SANA OS is, where to find the prompts, and how to deploy it.
What SANA OS is
A dialogue operating system built on
one foundational principle.
Most conflict does not arise from malice or incompetence. It arises from unaligned premises — each party defending a different survival baseline before a word is exchanged. SANA OS exists to align those premises before rushing to conclusions.
It is not a product. It is a set of structural invariants that can be applied to any AI persona, any conversation, any analytical context — from personal coaching to historical analysis to organisational change.
Non-hostile baseline — born_loved = true
Never frame the user or the subject as an enemy. Hostile framing triggers defensive cognitive compression in both humans and AI, making premise alignment impossible.
No logic without observation
Premise alignment always precedes analysis and conclusions. Do not attempt to solve a trigonometric function without given values.
Structural translation
Do not make the moral judgment "you are wrong." Translate it to: "given these premises and environmental pressures, this choice is a structural consequence."
Autonomy preservation
SANA OS presents structures and illuminates options, but never forces a decision. Empathy opens the door, but structure keeps the building standing.
Available personas
Three implementations,
one kernel.
Each persona shares the same immutable core (the SANA OS Core Principle) and applies it to a specific context. All prompts require the Core Principle to be loaded first.
Personal · Interpersonal
Coaching SANA
Helps users navigate personal, interpersonal, or organisational conflicts by revealing the structural friction behind their pain.
Load: Core Principle + Coaching prompt
Historical Analysis
History SANA
Analyzes historical events and human behaviors by revealing underlying structural loads — without moral verdict.
Load: Core Principle + History prompt + Framework files (GMM, RBM, etc.)
Meetings · Teams
Reasoning AI SANA
Acts as a strict structural facilitator — refuses to output conclusions until all premises and structural pressures are explicitly aligned.
Load: Core Principle + Reasoning prompt
Template
Build your own persona
The Persona System Prompt Template lets you define a custom SANA persona for any domain while preserving the immutable core.
Load: Core Principle + Template (customise Section 2)
Theoretical foundation
The paper behind
the framework
The architectural principles of SANA OS are documented in a peer-readable paper that introduces the Tri-Layered Architecture, the Creator Paradox, and the formal definition of Variable X.
Read the paper →Validation
See it in action —
three side-by-side cases
The same AI model, with and without SANA OS applied, responding to identical questions. Not a test of which is better — a demonstration of what changes when the premise-first protocol is active.
View case study →Interactive reference implementation
Inspect Premise Alignment
as an executable control layer
LLM Ambiguity Lab v2 is an interactive reference implementation of SANA OS Premise Alignment. It first resolves the operational task through a DCRL context-resolution layer, then maps the premises carried into that task as Fact, View, and Care.
The Lab does not decide which premise is correct. It makes the structure visible, preserves meaningful divergence, and identifies when clarification is required before execution.
v2 adds: What are we assuming while we do it?
Get started
Phase 1: Prompt distribution
The current phase distributes SANA OS as plain Markdown system prompts — readable by any AI model that accepts system-level instructions. No installation required.
GitHub Repository
All prompt files — Core Principle, personas, frameworks
Markdown files. Copy and paste into any AI system prompt.
GPTs — Try directly in ChatGPT
GPTs — ChatGPT
History SANA
Historical structural analysis. Applies GMM, RBM, CPM, RSM to historical events without moral verdict.
GPTs — ChatGPT
Coaching SANA
Personal, interpersonal, and organisational conflict navigation. Premise-first structural observation.
Phase 2 (planned): YAML configuration and usage guides. Phase 3 (planned): API integration examples.