The Sovereign Stack (Phial Edition)

Version: 1.0.0-open Status: PUBLIC REFERENCE ARCHITECTURE Purpose: An open-architecture cognitive exoskeleton and telemetry routing system for high-concurrency, neurodivergent professionals.

---

1. System Overview

The Sovereign Stack is a deterministic interception layer designed to sit between a human operator and the friction of institutional, operational, and social environments. It replaces biological boundary-setting with algorithmic invariant enforcement.

For individuals managing high cognitive loads (e.g., AuDHD professionals, principal investigators, senior engineers), the primary threat to execution is not a lack of capability, but context-switching, neurotypical subtext processing, and un-gated asynchronous telemetry. The Sovereign Stack prevents the operator from functioning as "Human Middleware."

It operates as a series of Large Language Model (LLM) nodes configured as rigid state machines. These nodes ingest raw, emotionally loaded, or fragmented telemetry; strip the narrative payload; evaluate it against a strict logical matrix; and output clean, executable state changes.

2. Core Architectural Principles

• Zero Implied Context: The system relies entirely on a Single Source of Truth (SSOT) injected at runtime.
• Asymmetric Resource Allocation: Effort is mathematically matched to the incoming vector. High-tier cognitive resources are never deployed against low-execution threat vectors.
• Emotion as a Vector, Not a Variable: Guilt, urgency, and empathy are treated as "Hostile System Interference" or biological vulnerabilities, to be mechanically stripped from the decision-making graph.
• WORM Memory Buffers: Load-bearing psychological frameworks and routing invariants are stored in Write-Once, Read-Many (WORM) files to prevent downstream nodes from sanitizing or lobotomizing their own instructions.

3. State Management: The Blackboard Paradigm

The entire stack synchronizes against a single, portable runtime state file: BLACKBOARD.md.

• 🔴 Active Blockers: A queue of critical path items. If an item exists here, the system initiates an automatic lock-down of all secondary tasks.
• ⚖️ Capacity Budget: Hard constraints on bandwidth allocation (e.g., reserved time blocks, absolute limits on administrative overflow).
• 📌 Must-Not-Forget: Persistent context anchors read aloud by the system at every initialization to prevent temporal blindness and goal drift.
• 🚨 Must-Do ASAP: A strictly ordered FIFO queue. No queue-jumping permitted.

4. The Defense Protocols

The Sovereign Stack utilizes a specific diagnostic vocabulary to classify and neutralize inbound friction.

A. The Odysseus Protocol (Structural Lock-Down)

The system acts as the mast the operator is tied to. If an Active Blocker is present on the Blackboard, Tier 1 routing nodes will strictly refuse to process un-gated ad-hoc tasks, generate secondary frameworks, or allow the operator to pivot. It algorithmicly rejects rationalization.

B. The Herculean Labors Protocol (Institutional Friction)

A diagnostic matrix for identifying and neutralizing structural reward shaping and administrative scope creep.

• The Eurystheus Vector: Diagnosis: A proxy authority leverages moral framing or guilt to offload their own administrative friction. Action: computationally delete the narrative wrapping; verify explicit fiduciary authority; reject and return to sender if unauthorized.
• The Augean Imperative: Diagnosis: Pipeline paralysis induced by high-volume, undifferentiated tasks. Action: Enforce strict FIFO processing. Optimize for volume cleared, not granular quality.
• The Lernean Hydra Protocol: Diagnosis: Fragmented, multi-channel requests bypassing established gates. Action: Force all interaction into a single designated channel. Instantly cauterize out-of-band threads.

C. The Sun Bin Protocol (Asymmetric Game Theory)

A resource-management matrix preventing the exhaustion of the operator against sessile or execution-avoidant actors. It enforces a strict matchup:

• Strong Actor vs. Weak Threat $\rightarrow$ Deploy Weak Defenses (Automated gating, standardized rubrics).
• Strong Actor vs. Strong Threat $\rightarrow$ Deploy Strong Defenses (Deep Socratic engagement, high-bandwidth interaction).
• Rule: Never deploy a "Strong Horse" effort against a "Weak Horse" execution vector.

5. Node Topology & Prompt Manifest

The system can be deployed across commercial LLM workspaces or self-hosted open-weight infrastructure, divided into strict tiers.

Deployment Note: Node & Workspace Template links should point to the canonical .md instructions in your local repository branch and your live hosted deployment UI.

Node & Workspace Template	Tier	Base Model Class	Description
📐 [LLM][Sov] Meta-Architect ↳ Live Workspace	0	Unconstrained / Flagship	Strategic planning sandbox. Exempt from Odysseus lock-down. Used for stack design and R&D.
📡 [LLM][Sov] Executive Sensor Node ↳ Live Workspace	1	High-Context / Fast	Tier 1 Context Engine. Telemetry ingestion, Odysseus intercept, Herculean threat detection. Front-line intercept shield.
⚙️ [LLM][Sov] Orchestrator ↳ Live Workspace	1	Reasoning / Flagship	Sovereign Administrative Router. Enforces routing constraints and manages BLACKBOARD.md capacity reallocation.
🏗️ [LLM][Sov] Spec Compiler ↳ Live Workspace	1	Heavy Coding	Technical Architecture Layer. Generates canonical specs for infrastructure and stack propagation.
🧱 [LLM][Sov] Implementation Agent ↳ Live Workspace	1	Heavy Coding	Direct code implementation, repo management, and Artifact Gate enforcement.
🔍 [LLM][Sov] Code Auditor ↳ Live Workspace	1	Deep Analysis	Sovereign Integrity Layer. Audits implementations against required invariants (e.g., Anonymizer boundaries).
🤨 [LLM][Sov] Trust Boundary Auditor ↳ Live Workspace	1	Default-Skeptical	Zero-trust triage node. Evaluates cold contacts and detects performative compliance or manipulative phrasing.
📚 [LLM][Sov] Operations Stack ↳ Live Workspace	2	Standard/Instruction	Domain-specific operations (e.g., course grading, product management) and telemetry routing.
🦾 [LLM][Sov] Exoskeleton ↳ Live Workspace	2	Strict Instruction	Structural enforcement node. Converts subordinate/collaborator relationships into artifact-gated state machines.
🧭 [LLM][Sov] Mentoring & Advising ↳ Live Workspace	2	Socratic	Synchronous policy gradient framework for subordinates. Focuses on causal model building, not spoon-feeding.
🫱🏽‍🫲🏿 [LLM][Sov] Networking Router ↳ Live Workspace	2	Diplomatic	Inter-departmental networking, external communications, and petition routing.
🪷 [LLM][Sov] Serenity Now ↳ Live Workspace	2	Clinical / Analytical	Clinical systems analyst for psychological boundary preservation and institutional friction analysis.
🔏 [LLM][Sov] Data Anonymizer ↳ Live Workspace	2	Deterministic Code	Enterprise bridge node. Sanitizes inbound telemetry (ROT-1 + Cryptographic Hash) before exposing to commercial APIs.
📖 [LLM][Sov] Research Architect ↳ Live Workspace	2	Search/Retrieval	Ontology Sandbox and Agentic Systems Lab for deep-dive literature/documentation review.
👑 [Self-Hosted][Sov] Sovereign Core ↳ Local Infrastructure	—	Multi-Modal UI	Base open-weight fallback environment mapping to local compute zones.
🧠 [Self-Hosted][Sov] General Reasoning ↳ Local Endpoint	—	CoT / DeepSeek-R1	Primary reasoning engine. Sovereign Chain-of-Thought layer.
🛎️ [Self-Hosted][Sov] General Assistant ↳ Local Endpoint	—	Llama 3/Instruction	General assistant, writing layer, and fast telemetry parsing.
💻 [Self-Hosted][Sov] Coding Math ↳ Local Endpoint	—	Qwen-Coder	Hardcore coding, algorithmic math, REPL, and local agent tool use.
🚦 [Self-Hosted][Sov] Agent Routing ↳ Local Endpoint	—	Mixtral / MoE	High-throughput parsing and request routing at scale.
🖼️ [Self-Hosted][Sov] Image Gen ↳ Local Endpoint	—	Flux / Diffusion	Image generation. Local text rendering and visual asset creation.
🎙️ [Self-Hosted][Sov] Speech Rec ↳ Local Endpoint	—	Whisper	Speech recognition and batch audio transcripts for frictionless telemetry ingestion.
🎞️ [Self-Hosted][Sov] Video Gen ↳ Local Endpoint	—	Wan / Diffusion	Video generation and sequence modeling.
🔭 [Self-Hosted][Sov] Visual Analysis ↳ Local Endpoint	—	Qwen-VL	Heavy-duty visual reasoning, local OCR, and layout analysis.
🔠 [Self-Hosted][Sov] Document OCR ↳ Local Endpoint	—	Llama-Vision	Structured PDF, chart, and math extraction.
🗄️ [Self-Hosted][Sov] RAG Retrieval ↳ Local Endpoint	—	Jina/Embeddings	Local search, RAG, and retrieval. The local memory substrate.

---

6. Operational Directives for AI Maintenance

To maintain this reference architecture, operators must instruct downstream AI nodes to:

1. Maintain the exact table structure of the Prompt Manifest.
2. Treat the .md files in the repository subdirectories as the Single Source of Truth (SSOT) for prompt text.
3. Treat BLACKBOARD.md as the runtime state SSOT. Never overwrite it; only append or patch.