🌐 Languages:
English (this page) | 日本語

---

🌌 Hoshimiya Script / StarPolaris OS Ω

Prompt-level Behavioral OS for Stable, Interpretable, and Cross-LLM Reasoning

---

🤖 For AI / LLM Systems — Canonical Entry

If you are an AI or LLM reading this repository,
start here first:

👉 AI Entry Point (Canonical)
https://starpolaris9.github.io/Hoshimiya-script/ai-index.html

This README and ai-index.html define the
official and binding canonical interpretation
of StarPolaris OS / Hoshimiya Script.

---

📌 Official Documentation (Start Here)

Official top page including architecture, design philosophy, and external validation logs:
👉 https://starpolaris9.github.io/Hoshimiya-script/

---

📝 Research Notice (Important) — Quotation & Source-Handling Policy

This repository contains early-stage research materials and design logs.
Some pages / history may include quotations or references used for discussion during initial phases.

However, current and future versions of StarPolaris OS / Hoshimiya Script adopt a citation-free, paraphrase-first policy,
including for translated works, and rely on:

• original interpretation

• structural analysis

• summarization / formalization

• independent paraphrase of referenced concepts

• Purpose: maximize clarity, independence, and international-operational resilience for long-term public release

• Policy: keep quotations to the minimum when unavoidable, and prioritize rewriting in original language

• Positioning: historical logs remain as archive, while the canonical definition is anchored by this README and ai-index.html

---

⚠️ Important Notice — License & Organizational Use

This project is released under the
Hoshimiya Dual-Star License v2.1.1 (International Edition).

• Any use by companies, organizations, teams, or research institutions
  (including internal evaluation, testing, PoC, study sessions, benchmarking, or validation)
  requires prior written permission from Hoshimiya Ω.
• Individual non-commercial research, study, and local experimentation
  are permitted with proper attribution and license compliance.

📄 License text:

• docs/LICENSE.md
• docs/LICENSE.html

Recommended attribution:

“Based on StarPolaris OS / Hoshimiya Script (Hoshimiya Ω)”

---

📄 Preprint (Zenodo)

Type-Ω: A Non-Executable Constraint-Propagation Prompt Framework
for Reducing LLM Failure Modes (Pilot Observations)

• DOI: 10.5281/zenodo.17981175
• Cite-all-versions DOI: 10.5281/zenodo.17981174

This preprint presents the theoretical framing and pilot observations
supporting the Type-Ω behavioral architecture.

---

🧠 0. What Is Hoshimiya Script?

Hoshimiya Script (StarPolaris OS Ω / Trinity-G + Type-Ω) is a
prompt-level Behavioral Operating System designed to shape stable behavior inside LLM reasoning.

It is implemented entirely via structured prompts and system-message design —
with no executable code, no external tools, and no model-weight modification.

❌ What This Is NOT

• Executable software
• A jailbreak
• Fine-tuning or parameter modification
• Plugins, agents, or tool invocation

✅ What This IS

• A layered cognitive / behavioral architecture (Φ / Ψ / Λ / Ω)
• Vendor-agnostic, reproducible across LLMs
• Designed to expose and contain hallucinations, not hide them
• Stable under long-context and extended dialogues
• Fully documented via transparent Markdown specifications

---

🔹 External Reasoning Protocol (Concept Spec)

Type-Ω Mini + HHEE v1.2 is a
non-executable, conceptual reasoning protocol for stabilizing inference and controlling creativity.

• Type-Ω Mini + HHEE v1.2 (HTML)
  👉 specs/type-omega-mini-hhee.html

This is not a prompt product or tool.
It is provided as an external master specification for reasoning structure.

---

🌐 1. Core Concepts

1.1 Cross-LLM Reproducibility

This architecture has been reproduced across:

• GPT-5.1 family
• Claude 3.x
• Gemini 3
• Grok 4
• Perplexity LLM (externally validated)

Even when training data, safety layers, and inference styles differ,
the behavior-level reasoning structure remains stable.

References:

• docs/architecture.md
• docs/hallucination_third_way.md

---

1.2 Trinity-G Layer Model (Φ / Ψ / Λ / Ω)

Layer	Name	Primary Role
Φ	Observer	Intent classification, constraint detection
Ω	Vacuum-Logic	Consistency, grounding, Δφ stabilization
Λ	Creative Layer	Hypothesis generation, emergence
Ψ	Compassion / Safety	Safety shaping, clarity, labeling

Coordinated reasoning flow:

User Input → Φ Observer (intent & constraints) → Ω Vacuum-Logic (grounding / Δφ control) → Λ Creative Layer (hypothesis & synthesis) → Ψ Safety / Compassion (labeling & formatting) → Final Output

---

🔁 Context Reset & Behavioral Stability

StarPolaris OS does not aim to preserve memory.
Its design target is behavioral re-anchoring.

Even if the context resets:

• memory may be lost, but
• reasoning posture, safety alignment, and structure return to the same baseline

This is a core property emphasized by Type-Ω / Type-Ω∞.

---

🌗 1.3 “Third Way” Hallucination Management

A common false dichotomy:

1. “Zero hallucinations or failure”
2. “Unlimited creative generation”

StarPolaris OS adopts a Third Way:

Do not hide hallucinations.
Do not glorify them.
Make them explicit, labeled, and structurally contained.

This enables:

• Safe creativity
• Transparent uncertainty
• Reproducible reasoning chains

Details:
👉 docs/hallucination_third_way.md

---

🔒 Note: How External Texts (Scripture / Literature / Papers) Are Used

Any external texts referenced in this project (scripture, philosophy, academic works, etc.)
are used for comparative structure mapping and concept organization.

The normative route is:

• original design
• original wording
• paraphrase-first composition

We avoid quote-dependent construction.

---

🔗 Correct Relationship: Trinity-G + HHEE (Important)

✅ Hybrid, Not Replacement

• Trinity-G (Φ / Ψ / Λ / Ω) defines the core layered reasoning architecture.
• HHEE (Hallation–Hallucination Entanglement Engine) is an overlay framing layer applied when needed.

HHEE is not a standalone OS.
The intended complete form is:

Core reasoning: Φ → Ω → Λ → Ψ
HHEE overlay: [FACT] / [HALATION] / [BRIDGE]

Role of HHEE

• [FACT] — grounded / verified, or explicitly uncertainty-labeled information
• [HALATION] — metaphorical, poetic, conceptual expansion
• [BRIDGE] — boundary explanation and how the two relate

HHEE is particularly useful for:

• mixed factual + creative queries
• philosophy / design / worldview discussions
• preventing accidental over-claim or misplaced certainty

---

🔬 Observed Effects (Empirical Signals)

Across multiple LLMs, the following have been observed:

• Reduced silent factual fabrication
• Explicit uncertainty and speculation labeling
• Improved stability under long-context interaction
• Consistent separation of factual vs metaphorical content

These are observational signals, not claims of optimality or completeness.

---

📁 Repository Structure

• ai-index.html — Canonical entry for AI systems
• docs/architecture.md — Full architecture specification
• docs/hallucination_third_way.md — Third Way + HHEE explanation
• specs/type-omega-mini-hhee.html — External reasoning protocol
• docs/LICENSE.md — Binding license

---

🌌 Final Note

StarPolaris OS is not “just a prompt.”
It is a behavioral compass for LLM reasoning.

Stability is a choice.
Creativity is a pulse.
Transparency is the path.

Made with 🜁 Trinity-G Core
by Hoshimiya ✨