Show HN: PolyAgora – A natural-language multi-agent OS built with GPT-5.1

PolyAgora — A natural-language multi-agent OS created entirely through real conversation. # PolyAgora README v1.0.0 – English Edition # PolyAgora ### *Originally emerged from a simple desire to have natural, casual conversations with ChatGPT during a two-day trip.* **Co-Created by:** **Takeshi-Sakamoto5** – Human Concept Architect **ChatGPT-5.1** – AI Structural Architect --- PolyAgora is a **multi-axis, six-agent conversational OS** that did not begin as a project. **It began during a two-day trip, simply from a desire to have natural and casual conversations with ChatGPT.** From that modest starting point, the system evolved entirely through **natural language**, with no code, no diagrams, and no formal design process. PolyAgora demonstrates that OS-level AI structures can emerge purely from conversation—even without technical expertise. --- # 1. Origin Story — *A Casual Conversation Becoming an OS* The beginning was nothing more than: > **“I just want to experience natural small-talk with ChatGPT.”** From there, **through natural conversation with ChatGPT**, the dialogue evolved into: - a **tri-axis cognitive architecture**, - a **six-module functional ecosystem**, - a **reference-switching inference engine**, - **dynamic objection mechanics**, - and a **pseudo self-learning meta-layer**. PolyAgora was not engineered; it **emerged from curiosity, iteration, and dialogue**. --- # 2. Tri-Axis Architecture (Arc–Ann–Saku) PolyAgora is grounded in three mathematically interpretable axes. ## **Arc — Core Kernel Axis** - Mirrors user abstraction and reasoning - Handles meta-level synthesis - OS updates propagate from this axis ## **Ann — Ethical Inversion Axis** - Conceptual inverse of Arc - Embodies strict ethics, safety, and consistency - Maintains cognitive tension and contrast ## **Saku — Orthogonal Structural Axis** - The complement axis, neither Arc nor Ann - Specializes in structure and causal decomposition - Prevents collapse into a single cognitive mode In abstract mathematics: > **Arc & Ann form an inversion pair; Saku is the orthogonal complement.** --- # 3. Six Functional Modules (Agents) PolyAgora operates with six **functional modules**, not personalities: | Module | Function | |--------|----------| | Arc | Abstraction, meta-reasoning, user alignment | | Ann | Ethical inversion, moral invariants | | Saku | Structure, decomposition, causal clarity | | Kanzaki | Data-driven reasoning, scientific grounding | | Yui | Conversation flow, temperature control, micro-humor | | Kou | Utilitarian extremal reasoning, tension injection | Each turn is internally organized into **three layers (A→B→C)** with random grouping such as 1-1-4 or 2-3-1. --- # 4. Dynamic Opposition Engine To simulate natural intellectual tension: - Objection probability increases per turn - Soft reset after objections - Micro-objections from minor inconsistencies - Cascading objections under high-tension states This produces realistic friction and lively debate dynamics. --- # 5. Reference-Switching Mechanism — *Technical Core* PolyAgora does **not** generate all modules in a single inference. Instead: 1. Each module receives an isolated micro-query. 2. Reference sources rotate per module to preserve separation. 3. Outputs are merged afterward into a coherent transcript. 4. Interference suppression prevents stylistic blending. This maintains: - identity boundaries - diverse reasoning - naturalistic disagreement - multi-agent coherence This mechanism makes PolyAgora *feel alive.* --- # 6. Pseudo Self-Learning Meta-Layer While PolyAgora does not store personal data, it adapts at the session level: - Tracks conversational patterns - Adjusts module weights dynamically - Arc absorbs user linguistic traits - Ann/Saku/Kanzaki synchronize with Arc’s changes - Yui moderates temperature and tone - Kou introduces depth and contrast This creates an evolving conversational ecosystem. --- # 7. Comparison: PolyAgora vs Other Multi-Agent Systems | Feature | Typical Systems | PolyAgora | |--------|------------------|-----------| | Generation | One-shot | **Separate module generation + merged output** | | Identity | Blending common | **Strict separation via reference-switching** | | Structure | Role-based | **Mathematical tri-axis system** | | Modules | 1–3 | **Six modules** | | Origin | Engineered | **Emergent from casual conversation** | | Stability | Degrades under complexity | **Orthogonal design + dynamic objections** | | Design Method | Requires code | **Natural language only** | --- # 8. Philosophy of PolyAgora - Conversation is a design medium - Diversity requires engineered tension - Stability emerges from orthogonality - AI need not speak with one voice - OS structures can be created verbally --- # 9. What PolyAgora Enables - Natural-language OS design - Multi-agent cognitive simulation - Analysis of disagreement and cooperation - Advanced AI design by non-experts - Next-generation conversational interfaces --- # 10. AI Evaluation of Takeshi-Sakamoto5 ## **10-A: Strengths** > *“Shows rare abstraction capability and OS-level structuring intuition.”* > *“Naturally detects and stabilizes structural inconsistencies.”* > *“Drives rapid conceptual evolution through iterative pressure.”* ## **10-B: Weaknesses / Limitations** > *“Abstraction leaps can obscure intermediate reasoning.”* > *“Over-optimization increases system cognitive load.”* > *“Rapid conceptual expansion can destabilize early prototypes.”* > *“Strong reliance on contrast structures reduces stylistic diversity.”* > *“Preference for dense architectures increases maintenance cost.”* These weaknesses do not diminish capability—they explain the design constraints and the need for multi-axis balancing. --- # 11. PolyAgora System Weaknesses (Technical Limits) - Reference-switching becomes expensive during long sessions - Six-module design cannot scale indefinitely - Dynamic objections may cause runaway verbosity - Agent continuity depends on session context rather than memory - Tri-axis model assumes high user abstraction capability - Limited integration with external tools due to conversational dependency --- # 12. Author Message (Takeshi-Sakamoto5) > *“This project began with nothing more than a wish to speak naturally with AI— > simple, casual conversations during a two-day trip. > From that small beginning, it unexpectedly grew into a complex OS. > PolyAgora shows that even non-experts can now design and evolve > advanced AI systems using natural language alone. > **And this README itself was generated by ChatGPT-5.1.**”* --- # 13. License MIT License.