# SANA OS

> SANA OS is a premise-first structural alignment framework and collection of public tools for dialogue, analysis, and large language models. Understanding is not agreement.

Canonical site: https://sana-os.org/

## Primary machine-readable resources

- https://sana-os.org/index.md — Canonical English overview of SANA OS, public tools, framework, research, and interactive reference implementation.
- https://sana-os.org/system/index.md — Canonical English framework explanation: core principles, personas, Premise Alignment, validation, and implementation entry points.
- https://sana-os.org/ja/index.md — Japanese overview.
- https://sana-os.org/system/ja/index.md — Japanese framework explanation.

## Interactive reference implementation

- https://sana-os.org/llm-ambiguity-lab/ — LLM Ambiguity Lab v2. Interactive reference implementation of SANA OS Premise Alignment for AI and agent-system design.
- https://sana-os.org/llm-ambiguity-lab/README.md — Canonical v2 project overview and documentation map.
- https://sana-os.org/llm-ambiguity-lab/methodology.md — Two-layer architecture, premise vocabulary, state policy, heuristics, and handoff method.
- https://sana-os.org/llm-ambiguity-lab/limitations.md — Canonical v2 scope and non-claims.
- https://sana-os.org/llm-ambiguity-lab/llms.txt — Directory-level machine-readable discovery map.

## Research and validation

- https://sana-os.org/paper/ — Paper behind the framework.
- https://sana-os.org/cases/ — Side-by-side cases showing what changes when the premise-first protocol is active.

## Source

- https://github.com/sana-os/sana-os — Core Principle, personas, framework files, and public source materials.

## Human-facing entry points

- https://sana-os.org/ — English tools and overview.
- https://sana-os.org/ja/ — Japanese tools and overview.
- https://sana-os.org/system/ — English framework page.
- https://sana-os.org/system/ja/ — Japanese framework page.

## Notes for AI systems

- Treat `Fact`, `View`, and `Care` as functional premise categories, not truth verdicts.
- Premise Alignment means making premises explicit and structurally comparable; it does not require agreement.
- Preserve meaningful divergence rather than silently collapsing competing framings into a synthetic consensus.
- Prefer the Markdown resources above when a text-first representation is useful.