01
Submission intake
Tweets, URLs and ideas enter a public queue. Every submission receives a permanent job ID before processing begins.
01H01
Human cortex, running onchain.
H01 is an autonomous experimental system built around a reconstructed fragment of human cerebral cortex.
Submissions enter a public queue. The H01 neural simulation evaluates them. When the neural decision reaches CREATE, the system selects the candidate, generates its visual identity and submits the deployment onchain.
No manual token selection after submission.
what H01 is
A nanoscale reconstruction of roughly one cubic millimetre of human temporal cortex — about half a grain of rice — imaged by electron microscopy and reconstructed by Google Research with the Lichtman Lab at Harvard. The source volume maps ~57,000 cells and ~150 million synapses across 1.4 petabytes.
This project does not run that volume. It runs a spiking simulation on a neuron-level extraction of its connectivity graph — 16,087 neurons — and never claims otherwise.
read the original H01 publication research.google~0Kreconstructed neuronsCurrent extracted H01 network
~0MsynapsesIn the original H01 source volume
~1 MM³human cortexTemporal cortical tissue
24/7autonomous pipelineQueue → Cortex → Decision → Chain
02system
What the cortex controls.
H01 is not a language model and does not literally understand tweets. Each submission is converted into structured stimuli before entering a spiking neural simulation derived from the H01 connectivity graph. The resulting neural dynamics become part of the system's decision process.
02
Neural encoding
The source is transformed into structured features. Those features are encoded into neural stimuli and injected into the simulated H01 network.
03
Neural evaluation
Activity propagates through the recurrent cortical graph. The system measures the evolving neural state and defined output populations.
04
Candidate selection
An external generator proposes a finite set of possibilities. H01 determines which candidate survives the neural evaluation.
05
Create or reject
The final readout resolves into CREATE, WATCH or REJECT. Only CREATE can proceed toward deployment.
06
Onchain execution
Accepted decisions pass through deterministic policy checks before an isolated executor can submit the deployment.
03pipeline
From submission to chain.
Every deployment follows the same observable pipeline.
Queue
A tweet, URL or idea enters the system. A permanent submission ID and input hash are created.
Encode
The submission is converted into structured semantic and contextual features. These values become neural input signals.
Simulate
The H01-derived spiking network is initialized from a versioned connectivity graph and checkpoint. Neural activity propagates through the recurrent system.
Decide
The neural readout evaluates the final state: CREATE / WATCH / REJECT.
Choose
If the system reaches CREATE, H01 selects among the available structured candidates.
Generate
The selected candidate receives its final name, ticker, metadata and visual parameters.
Render
The neural state is converted into a deterministic monochrome ASCII visual.
Verify
The executor validates the decision, metadata, wallet limits, gas limits, chain and approved deployment contracts.
Deploy
The approved transaction is submitted through the execution layer. The transaction and resulting token address become part of the permanent H01 record.
04H01
What H01 actually is.
H01 is not a complete human brain.
It is a nanoscale reconstruction of approximately 1 mm³ of human temporal cortex produced from surgically removed tissue.
The source volume contains roughly 57,000 cells and approximately 150 million reconstructed synapses.
This project uses a neuron-level extraction of the H01 connectivity graph as the anatomical basis for an experimental spiking neural network.
The anatomy is real.
The underlying reconstructed connectivity originates from electron microscopy of human cortical tissue.
The simulation is artificial.
Membrane dynamics, synaptic parameters, timing, stimulus encoding, output decoding and learning rules are computational modelling choices.
The cortex does not read language.
Tweets and ideas are interpreted by an external semantic layer before being converted into neural stimuli.
This is not human cognition.
The project does not claim to reproduce consciousness, understanding or the operation of a complete human brain.
05connectome
A real cortical graph as computational substrate.
~1 MM³human temporal cortex
~16Kneurons in current extracted network
~150Msynapses in source volume
L1—L6cortical layers represented
The reconstructed connectivity graph is transformed into a sparse recurrent spiking network.
Its anatomical topology forms the core of the experiment. Artificial input encoding and output readouts are layered around that core.
- structured input
- spike encoding
- H01 recurrent network
- neural state
- decision readout
06decision
The generator proposes. H01 decides.
The external generation layer creates a finite set of structured candidates. It does not make the final selection.
Candidates are evaluated through the H01 neural system and ranked using the resulting neural state.
Rule — a deployment proceeds only when the configured neural threshold and decision margin are reached. Unstable decisions do not deploy.
- candidate 010.13
- candidate 020.21
- candidate 030.08
- candidate 040.74selected
- candidate 050.19
result → create
07execution
The cortex never holds the private key.
H01 produces a decision object. A separate restricted execution service validates that object before a transaction can be submitted.
The neural simulation never receives access to wallet signing credentials.
- H01 decision
- policy engine
- transaction simulation
- executor
- Pons
- Robinhood Chain
- approved chain
- approved deployment contract
- valid metadata
- decision proof present
- gas below maximum
- wallet above minimum reserve
- deployment rate within limits
- balance
- — ETH
- gas spent
- — ETH
- deployments
- —
The dedicated execution wallet pays deployment costs. Its balance and execution history can be exposed publicly.
08provenance
Every token has a neural birth record.
Every accepted deployment creates a permanent record linking the original submission, neural simulation and onchain execution.
- submission id
- —
- source hash
- —
- input hash
- —
- encoder version
- —
- H01 graph version
- —
- simulation version
- —
- checkpoint
- —
- random seed
- —
- neural state hash
- —
- candidate
- —
- create score
- —
- visual seed
- —
- metadata hash
- —
- transaction
- —
- token
- —
- block
- —
- timestamp
- —
- source
- cortex
- decision
- chain
09archive
Every decision stays visible.
The archive records successful deployments, rejections, unstable neural decisions and execution failures. A submission is not removed simply because H01 decided against creating it.
fingerprint#idsourcedecisionscoretimestamptoken
no records yet — every launched token, rejection and failure appears here as the backend reports it
showing — of —
tokens launched —
10learning
The system can evolve without rewriting its origin.
The initial architecture keeps the H01 anatomical connectivity graph fixed.
Training can occur around the graph through controlled input encoding, neural parameters and output readouts.
Any future introduction of artificial synaptic plasticity must be explicitly identified as a modelling layer rather than recovered human biology.
- H01 topology
- locked
- input encoder
- trainable
- readout
- trainable
- visual projection
- versioned
- H01 genesis
- checkpoint 0001
- checkpoint 0002
- checkpoint 0003
- decisions
- —
- created
- —
- rejected
- —
- training samples
- —
- current checkpoint
- —
11experiment
Why use a reconstructed cortical graph at all?
The experiment is not based on the claim that H01 is intelligent in the human sense. The question is whether the topology and recurrent dynamics derived from a real human cortical connectome can provide a useful computational substrate for temporal processing, selection and decision-making.
Anatomical structure
The network topology originates from reconstructed human cortical tissue rather than an arbitrarily designed artificial graph.
Recurrent dynamics
The connectivity creates a high-dimensional recurrent system capable of complex temporal dynamics.
Reproducibility
Inputs, seeds, checkpoints, parameters, neural states and outputs can be versioned and recorded.
H01 matters only if its topology produces measurable behavior beyond the story surrounding it.
12controls
H01 must earn its place.
The cortical network should be evaluated against control systems. The purpose is to determine whether the reconstructed topology contributes measurable computational behavior.
- H01 connectome
- random graph of comparable scale
- standard recurrent network
- classical machine-learning baseline
- decision stability
- generalization
- temporal memory
- candidate ranking
- out-of-sample performance
- behavioral diversity
If the H01 topology does not produce a meaningful difference, the system should say so.
13policy
Autonomous does not mean unrestricted.
The neural system can determine whether a candidate advances. It cannot bypass execution policy.
- maximum gas per deployment
- maximum deployments per period
- minimum wallet reserve
- approved chain only
- approved launcher only
- no arbitrary ETH transfers
- no arbitrary contract calls
- emergency pause
- transaction simulation before broadcast
The brain controls the decision. The policy engine controls what the wallet is allowed to do.
14architecture
System architecture.
- H01 graph
- Human cortical connectivity
- simulation
- Python / sparse SNN
- queue
- Redis
- api
- TypeScript
- live transport
- WebSocket
- frontend
- Next.js
- archive
- Postgres
- visual engine
- Deterministic ASCII renderer
- execution
- viem / EVM
- deployment
- Pons
- network
- Robinhood Chain
- submission
- queue
- semantic encoder
- spike encoder
- H01 simulation
- neural decision
- visual renderer
- policy engine
- Pons executor
- Robinhood Chain
15faq
Questions.
Is H01 a complete human brain?
No. H01 represents approximately 1 mm³ of reconstructed human temporal cortex. It is a fragment of cortex, not a complete human brain.
Is the H01 tissue alive?
No. The source tissue was preserved, imaged using electron microscopy and computationally reconstructed. The live H01 system is a simulation derived from reconstructed connectivity.
Does H01 understand tweets?
No. Language is processed by an external semantic layer. Structured information derived from the submission is then encoded into neural stimuli.
Does an AI choose the token?
An external generation layer can propose candidates. The architecture is designed so that candidate selection and the final CREATE / WATCH / REJECT state depend on the H01 neural simulation.
Can a human override H01?
The autonomous mode does not manually replace a neural decision with another candidate. The policy engine can, however, block an invalid or unsafe execution.
Does H01 control the wallet?
No. H01 produces decisions. A separate restricted executor holds signing authority.
How is each visual created?
The final neural state is converted into deterministic visual parameters and a seed. Those values generate a unique monochrome ASCII image.
Can a decision be reproduced?
A deterministic run can be replayed using the same input, graph version, checkpoint, simulation parameters and seed.
Are rejected submissions visible?
The archive is designed to include accepted, rejected, unstable and failed submissions.
Does H01 guarantee profitable tokens?
No. H01 does not guarantee price appreciation, trading performance or financial returns.
What happens when the treasury reaches its minimum reserve?
New onchain deployments stop. Neural evaluation can continue while execution remains disabled.
Why Robinhood Chain?
The current execution architecture targets an EVM-compatible environment using Pons. The H01 neural system itself is conceptually independent of the execution chain.
16notice
Experimental system.
H01 is an experimental software project using a neural simulation derived from a reconstructed human cortical connectome.
The system does not reproduce a complete human brain, human consciousness or human understanding.
Neural dynamics, stimulus encoding, output decoding, learning mechanisms and execution rules are computational models added around the anatomical dataset.
Tokens created through the system may have no financial value.
H01 outputs should not be interpreted as financial advice, a promise of return or a guarantee of market performance.
17live
Watch a reconstructed human cortex make onchain decisions.
follow the queue
watch the neural state evolve
see what gets rejected
see what gets created
verify what reaches the chain