AKBS North Star Vision

Recursive Intelligence Architecture for Life, Learning, and Autonomous Systems

The Core Insight

The same AI architecture that helps one person manage their complex life can: 1. Help a beginner learn aquaponics 2. Monitor and operate a single system 3. Scale to manage 120 acres 4. Replicate itself for each new user/system

The difference is scope and inputs, not fundamental design.

The Fibonacci Principle

Scale 1: Personal Assistant (Josh's life management)
    ↓ same core, narrower focus
Scale 2: Learning Companion (beginner's first system)
    ↓ same core, more sensors
Scale 3: System Operator (autonomous container farm)
    ↓ same core, more systems
Scale 4: Property Overlord (120-acre Heritage site)
    ↓ same core, spawns children
Scale 5: Network Intelligence (connected STEM DREAM sites)

Each scale contains the previous. Each can spawn the next.

What The System Must Do (All Scales)

1. Know Everything Relevant

• Structured knowledge graph (not just text chunks)
• Explicit relationships between entities
• Cross-domain awareness (teaching ↔︎ grants ↔︎ hardware ↔︎ operations)

2. Pull Context On Demand

• Sprint mode: “I have a meeting with a lawyer in 2 hours” → System pulls: entity structure, financials, relevant decisions, open questions
• Batch mode: “Making honey this weekend” → System pulls: inventory, recipes, last batch notes, equipment status

3. Track & Pivot

• Persistent state across all projects
• When priorities shift, context follows
• Nothing gets lost in the pivot

4. Learn Toward Autonomy

• Starts as assistant (human decides, AI helps)
• Grows to advisor (AI recommends, human approves)
• Evolves to operator (AI acts within boundaries, human monitors)
• Matures to autonomous (AI runs system, human intervenes by exception)

5. Replicate & Distribute

• Core intelligence packages into deployable unit
• Ships with aquaponics container systems
• Open source through STEM DREAM
• New users get “baseline intelligence” that learns their context

Architecture Layers

┌─────────────────────────────────────────────────────────────┐
│  LAYER 4: NETWORK INTELLIGENCE                              │
│  Multiple sites, shared learning, federated knowledge       │
└─────────────────────────────────────────────────────────────┘
                            ↑
┌─────────────────────────────────────────────────────────────┐
│  LAYER 3: PROPERTY/FACILITY MANAGER                         │
│  Multiple systems, resource allocation, scheduling          │
│  Example: Heritage Sportsman's Club 120 acres               │
└─────────────────────────────────────────────────────────────┘
                            ↑
┌─────────────────────────────────────────────────────────────┐
│  LAYER 2: SYSTEM OPERATOR                                   │
│  Single aquaponics system, sensors, automation              │
│  Example: AKBS container, basement lettuce farm             │
└─────────────────────────────────────────────────────────────┘
                            ↑
┌─────────────────────────────────────────────────────────────┐
│  LAYER 1: PERSONAL ASSISTANT / LEARNING COMPANION           │
│  Knowledge management, task tracking, Q&A, guidance         │
│  Example: KIRA for Josh, KIRA for first-time student        │
└─────────────────────────────────────────────────────────────┘
                            ↑
┌─────────────────────────────────────────────────────────────┐
│  FOUNDATION: KNOWLEDGE GRAPH + MEMORY + TOOLS               │
│  - Structured entity relationships                          │
│  - Persistent memory (conversations, decisions, data)       │
│  - MCP tool integrations (sensors, APIs, services)          │
│  - LLM interface (cloud or local)                           │
└─────────────────────────────────────────────────────────────┘

Implementation Path

Phase 1: Personal Assistant (Now → Q1 2026)

Goal: KIRA managing Josh’s life and projects

• Set up KIRA on laptop/Mac Mini
• Seed with project knowledge
• Build knowledge graph structure
• Test “sprint context pull” for real scenarios
• Develop custom MCP tools as needed

Phase 2: Learning Companion (Q1-Q2 2026)

Goal: Raspberry Pi version for STEM DREAM education

• Port KIRA Python backend to Pi 5
• Create simplified interface (web or voice)
• Integrate Atlas Scientific sensors
• Build “moon base survival” curriculum hooks
• Test with students

Phase 3: System Operator (Q2-Q3 2026)

Goal: Autonomous monitoring and intervention

• Sensor → decision → action loops
• Alert escalation patterns
• Logging for learning and debugging
• Human override interfaces
• Basement lettuce farm as testbed

Phase 4: Property Scale (2026-2027)

Goal: Multi-system coordination

• Heritage Sportsman’s Club as demonstration site
• Aquaponics + greenhouse + poultry integration
• Resource flow optimization
• Visitor/education interface
• Agritourism coordination

Phase 5: Network (2027+)

Goal: Connected STEM DREAM sites sharing intelligence

• Federated learning across installations
• Shared knowledge base (anonymized)
• Remote support and monitoring
• Open source distribution model

The Nonprofit Model

STEM DREAM gives away: 1. Hardware specs – Pi, sensors, basic infrastructure 2. Software – AKBS core intelligence, open source 3. Curriculum – Moon base narrative, hands-on learning 4. Support network – Connected sites help each other

What makes it sustainable: - Grants fund development and demonstration sites - Schools/programs purchase turnkey kits (cost recovery) - Heritage Sportsman’s Club generates revenue (agritourism, memberships) - Consulting/training for larger installations

Why This Works

1. Josh’s pain = everyone’s pain – Managing complexity is universal
2. Learning by building – The system that helps Josh IS the system students learn with
3. Recursive testing – Every scale validates the one below it
4. Open source = exponential reach – STEM DREAM’s mission scales beyond what one org can touch
5. AI gets better with use – More installations = more learning = better baseline intelligence

Immediate Next Steps

1. ✅ Brain dump with Claude (done)
2. ⏳ Set up KIRA (tomorrow)
3. ⏳ Seed KIRA with this vision + project docs
4. ⏳ Begin knowledge graph design
5. ⏳ First “sprint context pull” test (lawyer meeting, grant deadline, etc.)

The Mantra

“Build the system that helps you build the system.”

KIRA helps Josh manage the project of building the AI that becomes KIRA for everyone.

Document created: January 7, 2026 Living document - updated by KIRA + Josh