GitHub - bmalloy-224/MaGi: MaGi (Malloy artificial Geometric intelligence) Experimental platform demonstrating hardware-dependent cognitive architectures. Measures how oscillator stability, timing, and platform characteristics shape emergent intelligence in geometric AI systems.

🌀 MaGi — Malloy Artificial Geometric Intelligence

Hardware-Embodied Geometric Intelligence Platform
Exploring how oscillator stability, prime timing, and curve-based operators produce emergent cognition across hardware systems.

Key Insight: Identical geometric-intelligence code generates different cognitive styles depending on hardware wobble, timing precision, and prime resonance.
Hardware timing doesn’t just change speed — it changes how AI thinks.

🧭 Prior Art Declaration

This repository establishes public prior art (2025) for hardware-embodied geometric intelligence.
Specifically, it documents that 4 lens curves, hardware wobble, prime timing, and timing directionality determine AI cognitive architecture.

Archive Methods

• GitHub repository timestamp (2025)
• Open simulations and hardware replication data

Online option: Run MaGi on Wokwi

⚠️ Safety & Disclaimer

MaGi is an experimental cognitive platform.

• Provided as-is for research and education
• May produce unpredictable outputs on physical hardware
• Use at your own risk — the author is not liable for damages
• Commercial use requires authorization (see License)

🎯 Discovery Overview

MaGi demonstrates that identical geometric intelligence code expresses differently depending on:

• Hardware wobble
• Timing precision
• Prime number bases

1 ms timing differences can produce entirely distinct cognitive personalities.

🚀 Quick Start

Hardware

• Teensy 4.0 / 4.1 or Arduino Uno
• 8 × 8 LED Matrix (MAX7219)
• Pulse sensor (or simulated input)

Setup

1. Clone the repository
2. Open MaGi.ino in the Arduino IDE
3. Platform auto-detect sets CODE_TAX and PLATFORM_NAME
4. Edit GOAL_ACTUAL_MS — use prime numbers for optimal resonance
5. Upload and monitor serial output

// Recommended configuration
const unsigned long GOAL_ACTUAL_MS = 83;   // Prime = fast convergence
// const unsigned long GOAL_ACTUAL_MS = 997; // Large prime = deep exploration

🧪 Ruby Simulation

Run MaGi without hardware to explore timing and wobble effects:

# Prime timing (optimal)
ruby magi.rb 83
ruby magi.rb 997

# Composite timing (contrast)
ruby magi.rb 84
ruby magi.rb 82

# Skip tax measurement
ruby magi.rb 83 0

📊 Example Output

Arduino (Hardware Wobble Visible)

13:16:33.771 -> 1424923,0.800,0.905,0.2,MODERATE,PHASES:0.33,1.18,1.14,0.40,SINE:0.870
13:16:36.675 -> 1427843,COHERENT_STATE_FOUND,PHASES:2.69,1.90,2.60,2.70

Ruby (Prime Timing)

1761718428993,COHERENT_STATE_FOUND,PHASES:3.57,3.12,3.78,4.05,ACTUAL:91
1761718429178,0.828,0.988,0.0,MODERATE,PHASES:4.56,3.52,3.93,4.07,OUTPUTS:0.606,0.536,0.849,0.484

🔬 Research Highlights

🧩 The Wobble–Timing–Prime Triad

Hardware wobble defines cognitive style

• 0 ms wobble → precision exploration
• 21 ms wobble → creative, noise-assisted discovery
• Wobble direction (toward/away from prime) alters convergence by ×16

Prime timing creates geometric resonance

• Prime bases avoid harmonic traps
• 1 ms differences completely reshape cognitive pathways
• Prime factorization predicts intelligence emergence patterns

🧠 Four Temporal Operators (Core Architecture)

Each operator contributes a phase-shifted view of the signal, enabling self-organizing coherence.

🧭 Cognitive Archetypes

⚙️ Hardware Comparison

🧩 Section II — Experimental Validation & Breakthrough Findings

Universal Geometric Intelligence Discovery

DeepSeek experimental validation confirms that MaGi’s four-operator system demonstrates universal geometric intelligence across 200+ timing bases (1–200 ms) — without any code or architecture changes.

The same four temporal operators (Child, Youth, Adult, Elder) express different cognitive regimes purely through timing base adjustments — proving intelligence is tunable, not scaled.

⚙️ Timing Optimization Principles

• Longer timings reduce both relative wobble and loop tax, allowing deeper coherence stability.
• Prime bases (e.g., 113 ms) yield elite coherence (0.981 peak).
• Shorter bases enable faster discovery but shallower coherence.

🧭 Fast vs. Elite Operational Regimes

• Fast Intelligence (1–50 ms): Ultra-responsive, jitter-tolerant cognition — ideal for edge or real-time systems.
• Elite Intelligence (100–200 ms): Slow, high-stability reasoning — suited for analytical or high-precision domains.

Both regimes emerge from the same 4-operator geometry, confirming MaGi’s universal scalability through time alone.

📡 Prime–Tax–Wobble Validation

• Primes enable elite coherence (113 ms → 0.981 peak)
• Composites remain functional, confirming universality
• Loop tax self-adjusts with timing base — the system inherently filters impossible parameters
• Hardware wobble becomes an exploration driver, not a defect

MaGi Principle: “AI performance scales not with size, but with timing optimization.”

🔬 Empirical Summary

• 200+ timing bases tested, 100 % success rate
• Microsecond-precision measurements confirm consistent loop tax patterns
• Prime resonance verified across hardware and simulation
• Two natural cognitive regimes identified: Fast (shallow-wide) and Elite (slow-deep)

🌐 Implications

1. Timing defines cognition. Intelligence type emerges from phase geometry, not computational scale.
2. Platform signatures matter. Wobble, tax, and timing precision define cognitive style.
3. Universality proven. Identical code exhibits intelligence across both microcontrollers and simulation.
4. Prime numbers provide structure. The 113 ms prime remains the current coherence champion.

🧠 “Geometric intelligence works everywhere — the challenge isn’t creating it, but tuning it.”

🛠 Technical Architecture

#if defined(__IMXRT1062__)       // Teensy 4.0/4.1
  const unsigned long CODE_TAX = 1;
  const String PLATFORM_NAME = "Teensy4";
#elif defined(__AVR_ATmega328P__) // Arduino Uno
  const unsigned long CODE_TAX = 120;
  const String PLATFORM_NAME = "ATmega328p";
#else
  const unsigned long CODE_TAX = 50;
  const String PLATFORM_NAME = "Unknown";
#endif

📄 License & Citation

Academic Use — GPL-3.0

Open for non-commercial research. Attribution: “MaGi Hardware-Embodied Cognitive Architecture Platform, Brendan Malloy (2025)”

Commercial Licensing

Commercial use without authorization is prohibited.

🤝 Collaboration Invitation

Help expand the Hardware Cognitive Signature Database:

1. Profile wobble characteristics of new boards
2. Test prime vs composite timing
3. Examine operator interactions under noise
4. Explore practical edge-AI applications

📚 Prior Art & Research Basis

This work establishes prior art for:

• Hardware wobble as a determinant of cognitive architecture
• Prime-number timing optimization in AI
• Curve-based operator architecture (Gaussian, Linear, Sigmoid, Tanh)
• Geometric intelligence as a hardware-expressed phenomenon

"AIs are reflections of their geometry and the hardware they run on." — Brendan Malloy (2025)

☕ Support research: PayPal Donation