MaleCNS v1.0 · Coinbase Advanced

A fly brain
trading crypto.

Stonkfly is a fly-connectome simulation that can operate a crypto trading account. Actual neural output, actual Coinbase integration. Profitable learning has not been demonstrated.

View on GitHubRun it locally
166,700
neurons
25.6M
connections
3,335
brightness inputs
811
R8 color inputs
Pixel-art fly beside a green candlestick chart

How it works

01

Prices become vision

Public Coinbase BTC-USDC data is rendered as an RGB candlestick chart and fed to the retina.

02

The connectome fires

The retained MaleCNS v1.0 graph propagates the stimulus across its full neuron population.

03

A fixed readout decides

Neural output is mapped to a single proposal: buy, sell or hold. No hand-written strategy.

04

Guards then execute

A custom Coinbase AgentKit action provider checks every limit before any spot order is placed.

Experimental memory

Positive portfolio P&L stimulates 15 identified PAM11 dopamine cells; negative P&L stimulates two PPL101 aversive dopamine cells. A candidate memory rule then changes existing KC-to-MBON connections.

These are engineered reinforcement signals, not modeled pain receptors. Synaptic changes do not establish that it learns to trade profitably.

Run it

Python 3.11, a C++17 compiler, macOS or Linux. Allow several GB for the dataset and dependencies; 16 GB RAM recommended.

python3.11 -m venv .venv
source .venv/bin/activate
pip install -e '.[test]'
python -m stonkfly prepare
python -m stonkfly run

Defaults to paper trades on real public BTC-USDC data with a $100 simulated balance. No key needed. Ctrl-C stops it; the same command resumes.

Guardrails

Max order
$10 incl. reserved fees
Attempts
24 per day
Shorts / leverage
None
Drawdown stop
$20 halts new orders
API key scope
View + Trade, no Transfer
Default mode
Paper, $100 simulated

The repo does not come funded or connected to anyone's account. Live execution needs your own local credentials and explicit opt-in.