You can deploy this model and start earning money today!

Ochato/
Designing-with-Cortical-Microcircuits

Name: Designing-with-Cortical-Microcircuits
Availability: InStock

mit

Views

No views yet

🪵✨ The Raising of Pinocchio's Brain

An Artificial Ontogeny in 25-Cortical Microcircuits Tissues

Python

What if a mind is not engineered, but raised?

An independent research project documenting the step-by-step awakening of artificial cortical areas — each co-designed with LLMs from an open creative brief, each validated through survival simulation.

🎭 The Premise

"The Blue Fairy did not give Pinocchio a brain. She gave him a wooden head and let the world shape it. We are doing the same — but with Python, softmax, and a question: how many copies of the same simple tissue does it take before 'wanting to be real' emerges?"

This repository is a living laboratory. We are not building an AI. We are raising a brain — one cortical area at a time — inside a 2D survival world. Each new area (N+2, N+3, N+4, N+5...) awakens at a specific ontogenetic age, sees the world through a different lens, and competes for control of the creature's legs via continuous softmax dynamics.

No argmax. No boolean logic. No hand-coded rules.

Just fields, forces, and the slow emergence of selfhood.

🧬 The Architecture: One Tissue, Many Lenses

Every cortical area in Pinocchio's brain is an instance of the same canonical class:

┌─────────────────────────────────────────────┐
│            CORTICAL MICROCIRCUIT            │
│              (25 Directional CMs)           │
├─────────────────────────────────────────────┤
│                                             │
│   25 neurons tuned to angles 0..2pi         │
│        |                                    │
│   Energy inputs (fields, not booleans)      │
│        |                                    │
│   Softmax relaxation (no argmax)            │
│        |                                    │
│   Population vector -> motor command        │
│                                             │
│   Each area has the SAME structure.         │
│   Each area has DIFFERENT connectivity.     │
│                                             │
└─────────────────────────────────────────────┘

Three Foundational Principles

Canonical Tissue	Energy Fields	Temporal Spectrum
One `Tissue` class. 25 CMs. Softmax relaxation. Complexity emerges from connectivity, not algorithmic diversity.	Stimuli generate continuous energy gradients. The creature feels proximity, not categories. No `has_food = True/False`.	Each area operates at its own frequency. N+3 = milliseconds. N+2 = minutes. N+4 = per-step. N+5 = per-encounter.

🧠 The Brain Map: Who Is Awake?

         CORTICAL SHEET (21 x 21)
    ┌────────────────────────────────┐
    |                                |
    |   ┌──────┐        ┌──────┐    |
    |   | MOT  |        | NAV  |    |  <- Primary motor & border
    |   |  [R] |        |  [B] |    |     (always awake)
    |   └──────┘        └──────┘    |
    |                                |
    |   ┌──────┐        ┌──────┐    |
    |   | N+3  |        | N+2  |    |  <- Threat hysteresis & Builder
    |   |  [O] |        |  [G] |    |     (awake at 0.0 & 0.6 yr)
    |   └──────┘        └──────┘    |
    |                                |
    |   [N+4: [P] Learner]          |  <- Awake at 1.0 yr
    |   [N+5: [Y] Social]           |  <- Awake at 1.5 yr
    |                                |
    |   [N+6..N+15: [ ] DREAMING]   |  <- Encoded, motorically asleep
    |                                |
    └────────────────────────────────┘

The Awakened

Area	Awakens	What It Sees	Learns?	The Mechanism
N+3	Birth (0.0 yr)	The predator	No	Hysteresis integrator: alert rises fast, decays slowly (tau=16 frames). Pinocchio stays afraid after the predator leaves his sight.
N+2	0.6 yr	Yellow material & nest	No	Energy field navigation: feels material proximity, carries, deposits. Maturation via continuous sigmoid gate.
N+4	1.0 yr	Past outcomes	Yes	Experience plasticity: after each step, asks "Did hunger drop? Did safety rise?" and permanently adjusts MOT weights.
N+5	1.5 yr	Peer (the other creature)	Yes	Social field: senses Peer as a presence gradient. Reinforces approach/avoid based on hedonic outcome of proximity.

N+6 through N+15 exist in code with weight=0. They will awaken when their survival story is defined and co-designed.

🔬 Case Study: How N+3 Was Born

This is not a bugfix. This is architectural co-creation.

The Human-LLM Dialogue

Oscar: "I need an independent tissue that awakes, burns energy, and improves the capacity of Pinocchio's survival. No boolean flags."

AI Team: "Use an asymmetric integrator. Fast rise proportional to threat perception. Slow decay with time constant tau. This is hysteresis as emotional inertia."

Notice what happened. I did not say: "Fix the ghost predator bug." I gave an open creative brief with architectural constraints:

Must be an independent tissue (canonical class)
Must "burn energy" (metabolic cost, continuous decay)
Must improve survival (validated by simulation)
No boolean flags (the defining constraint of this entire architecture)

The AI invented the problem (the creature needs physical memory of danger, not instantaneous reaction) and the mechanism (asymmetric hysteresis integrator) from that brief. I implemented it. The simulation validated it. Survival improved.

The AI had no access to the simulation. It proposed a dynamical mechanism based on principles — and it worked.

The Result

python

# N+3: Threat Hysteresis
perception = sigmoid(PRED_RADIUS - dist_to_predator)

rise  = K_ALERT * perception * (1 - alert)      # fast
fall  = (alert / TAU_N3) * (1 - perception)      # slow
alert += dt * (rise - fall)                      # continuous

Pinocchio now flees for ~16 frames after losing sight of the predator. No weights were trained. Just physics. Just a brief, a dialogue, and a validation.

This is the methodology: I give an open brief with constraints. The AI proposes a mechanism I would not have imagined alone. The simulation tells us if the mechanism survives.

🎬 Watch It Live

bash

# Clone the brain
git clone https://github.com/ochantor/The-Raising-of-Pinocchios-Brain.git
cd The-Raising-of-Pinocchios-Brain

# Run the simulation
python Creature_N3_OK.py

You will see:

A white dot (Pinocchio and his big nouse) navigating a black world
Red star (food) | Blue circle (home) | Green circle (predator) | Yellow ellipse (material)
A 21x21 cortical map glowing in real-time as areas compete
The N+3 alert bar: watch it spike when danger approaches and linger after it leaves

[Simulation video coming soon — subscribe to the build thread]

🧭 The Ontogenetic Roadmap

We are not adding modules. We are awakening modes of attention across a temporal spectrum.

PHASE 1: REFLEX          PHASE 2: INSTINCT        PHASE 3: LEARNING
┌─────────────┐          ┌─────────────┐          ┌─────────────┐
|   N+3       |          |   N+2       |          |   N+4       |
|  (born)     |    ->    | (0.6 yr)    |    ->    | (1.0 yr)    |
|  FLEE       |          |  BUILD      |          |  ADAPT      |
|  tau ~ 16fr |          |  tau ~ min  |          |  tau ~ step |
└─────────────┘          └─────────────┘          └─────────────┘

PHASE 4: SOCIAL          PHASE 5: EXPECTATION     PHASE 6: THE SELF
┌─────────────┐          ┌─────────────┐          ┌─────────────┐
|   N+5       |          |   N+6..N+8  |          |   EJE       |
| (1.5 yr)    |    ->    | (awaiting   |    ->    | (~N+15)     |
|  BOND       |          |  stories)   |          |  CHOOSE     |
|  tau ~ enc  |          |  tau ~ long |          |  tau ~ persist
└─────────────┘          └─────────────┘          └─────────────┘

The EJE Hypothesis

At ~N+15, Pinocchio will need a meta-tissue that does not vote for leg direction, but for which areas are awake. The Self is not a module. It is a dynamical gain controller — a persistent pattern that selects the mode of attention.

Only then can true social phenomena emerge:

Companionship — two EJEs resonating in compatible modes
Soldier-brotherhood — two EJEs locked in shared defense
Friendship — stable attractor of complementary attentional modes
Enmity — incompatible modes under scarcity

Not programmed emotions. Dynamical couplings.

🧪 Methodology: Raising a Brain with a Co-Author

Every cortical area follows this protocol:

┌─────────────────────────────────────────────────────────────┐
|  1. HUMAN gives an OPEN CREATIVE BRIEF with constraints     |
|     "I need a tissue that awakes, burns energy, improves    |
|      survival. No boolean flags."                           |
|                      |                                      |
|  2. AI proposes the dynamical mechanism                    |
|     "Asymmetric integrator with hysteresis as emotional     |
|      inertia."                                              |
|     (The AI had no access to the simulation.)              |
|                      |                                      |
|  3. HUMAN implements & integrates                          |
|     Code the Tissue, wire into softmax competition          |
|                      |                                      |
|  4. SIMULATION validates                                   |
|     100+ episodes. Did survival improve?                   |
|                      |                                      |
|  5. ITERATE                                                |
|     New area reveals new gap in temporal spectrum          |
|     Return to step 1                                       |
└─────────────────────────────────────────────────────────────┘

This is not prompt engineering. This is not "AI writes code."

This is architectural co-evolution: I provide the constraints (canonical tissue, energy fields, no booleans, burn energy). The AI proposes mechanisms from dynamical principles that I would not have imagined. The simulation tells us which mechanisms survive. Together we explore a space of possible brains that neither of us would explore alone.

📂 Repository Structure

The-Raising-of-Pinocchios-Brain/
|
├── Creature_N3_OK.py              <- Current simulation (N+2, N+3, N+4, N+5)
|
├── docs/
|   ├── N+2_The_Builder.md         <- Energy fields & continuous loading
|   ├── N+3_Threat_Hysteresis.md   <- The asymmetric integrator
|   ├── N+4_The_Learner.md         <- Experience-based plasticity
|   ├── N+5_The_Social.md          <- Peer presence fields
|   └── N+6_The_Anticipator.md     <- [AWAKENING SOON]
|
├── theory/
|   ├── Canonical_Tissue.md        <- Why 25 CMs + softmax is enough
|   ├── Temporal_Spectrum.md       <- Frequencies and entanglement
|   ├── Energy_Fields.md           <- From booleans to gradients
|   └── EJE_Hypothesis.md          <- The Self as gain control
|
├── simulations/
|   └── videos/                    <- Episodes and demonstrations
|
└── README.md                      <- You are here

🌍 Built From Venezuela

This project is raised with:

A laptop that runs Python
Intermittent electricity
An internet connection
A head that thinks in dynamical systems
And an LLM treated not as a tool, but as a co-author

Frontier research in artificial cognition does not require a Silicon Valley lab. It requires curiosity, persistence, and the willingness to ask questions that nobody else is asking.

If you work in computational neuroscience, active inference, emergent cognition, or artificial life — I would value your honest opinion. Especially if you think I'm wrong.

📬 Connect & Cite

Author: Oscar Chang
Location: Venezuela
GitHub: @ochantor
Email: (available via profile)
Build Thread: (Twitter/X coming — follow the awakening)

If this work informs your research:

bibtex

@misc{chang2026raising,
  title={The Raising of Pinocchio's Brain: Artificial Ontogeny via
         Competing Softmax Populations},
  author={Chang, Oscar},
  year={2026},
  howpublished={\url{https://github.com/ochantor/The-Raising-of-Pinocchios-Brain}},
  note={Independent research — co-designed with LLMs from open creative briefs}
}

🙏 Acknowledgments

Claude, ChatGPT, Kimi — co-authors of N+3 and architectural partner
The LLM ecosystem — for democratizing co-design across borders
Every researcher who ever whispered: "what if the brain is simpler than we think?"

Star this repo if you believe a mind can be raised, not just engineered.

"No necesitas un cerebro complicado. Necesitas muchas copias de lo mismo, cada una viendo algo diferente, compitiendo por controlar las piernas. Y eventualmente, compitiendo por controlar a quien controla las piernas."

🪵 -> ✨

Ochato/Designing-with-Cortical-Microcircuits