This repository is the umbrella hub for the entire Natural Language OS ecosystem created by Masaya Ochiai.
Natural Language OS (NL-OS) is defined as:
An operating system that defines its structure, configuration, and behavior in natural language,
and executes these rules on top of an LLM acting as its kernel.
Unlike conventional operating systems whose logic is written in code,
a Natural Language OS treats language as its specification, control layer,
state representation, and cognitive framework.
This hub provides:
- A unified taxonomy of Natural Language OS
- Links to all engine implementations (ArcOS, PolyAgora, Echoos, etc.)
- A complete Foundations → Engines → Applications architecture
- The starting point for researchers, engineers, and future contributors
A Natural Language OS is:
- written in natural language
- executed by an LLM kernel (GPT / Claude / Grok etc.)
- capable of holding persistent value systems
- able to maintain OS boundaries (namespace)
- able to run reasoning pipelines, mode structures, multi-agent cognition
- a full cognitive OS rather than a prompt or chatbot
This concept is fully defined and formalized in the
Natural Language OS Whitepaper (Ochiai, 2025).
(Aligned 100% with the Whitepaper)
A Natural Language OS is structured into three layers,
as defined in the whitepaper (Ochiai, 2025):
- values / priorities
- stable behavioral rules
- decision criteria
- preferred reasoning styles
- identity & worldview structure
This acts as the “OS files” of NL-OS.
- active reasoning
- mode switching
- state tracking
- temporary overrides
- cognitive processes performed now
This is the “active CPU process” of the OS.
- namespace management
- cross-OS arbitration
- mode contamination prevention
- consistency checks
- OS supervision and safety
This is unique to NL-OS and not found in traditional OS design.
The Natural Language OS ecosystem consists of three top-level categories:
The theoretical and philosophical basis of NL-OS:
- What is a Natural Language OS?
- Language-as-Code paradigm
- Persistent / Runtime / Meta architecture
- OS boundary (namespace) theory
- Value-system as OS kernel
- Safety, governance, and limitations
- NL-OS vs prompts, agents, RAG, classic OS, cognitive architectures
📁 Folder: Foundations/
📄 Whitepapers will be linked here (via Zenodo DOIs)
Executable Natural Language OS instances built purely in natural language.
https://github.com/Takeshi-Sakamoto5/ArcOS
- Deterministic reasoning based on a user’s cognitive structure
- Zero-drift alignment with personal worldview
- Persistent values and mode architecture
https://github.com/Takeshi-Sakamoto5/PolyAgora
- Multi-agent / multi-axis cognition
- 6 modules with strict reference isolation
- Divergence → Collision → Synthesis cycles
- Realistic multi-party reasoning
📁 Folder: Engines/
Internal specs, comparisons, OS suite structure
How NL-OS runs in real environments:
- Personal cognitive OS running on top of an LLM
- Task planning, memory, value-aligned reasoning
- Multi-OS orchestration for daily workflow
- Appliance integration
- State synchronization
- Intent → value → action mapping
- Safety & ambiguity governance
- Edge LLM integration
- High-level cognitive controller for real devices
📁 Folder: Applications/
- Definitions
- NL-OS taxonomy
- Whitepapers (Zenodo)
- OS theory
- ArcOS
- PolyAgora
- Echoos
- Future OS modules
- Home-Computer OS
- Smart-Home OS
- Embedded NL-OS specs
Natural Language OS is the third category of OS, after:
- Hardware OS（Linux / Windows / macOS）
- Software orchestration OS（Agent frameworks, workflow systems）
- Cognitive OS (Natural Language OS) ← this project
NL-OS enables:
- language as code
- externalized cognition
- stable reasoning
- value-based decision kernels
- multi-agent cognitive field
- OS-level consistency
- future smart-home / robotics OS integration
- “external brain” architectures
It is not a prompt.
It is not a chatbot.
It is a full OS written in language.
Masaya Ochiai (2025)
Creator of ArcOS, PolyAgora, Echoos, NL-OS taxonomy
X: https://x.com/NaturalLangOS
- Individual engines have their own LICENSE files
- Whitepapers will be archived with Zenodo DOI
- This repository serves as the master specification hub