Breaking

Vibecoder Tests Bitcoin Trades with a Simulated Fruit-Fly Brain and $100

September 10, 2026 at 09:43 EDT

Software Dev & Coding
Finance

On September 10, 2026, Vibecoder's Stonkfly emerged as a newly published open-source experiment that puts a simulated fruit-fly connectome into a Bitcoin paper-trading loop with a $100 balance. The project sends positive and negative portfolio changes to engineered dopamine signals, but its own validation notes say profitable learning has not been demonstrated.

10 SEP 2026 · STONKFLY

A Fruit-Fly Brain Enters the Bitcoin Loop

A simulated connectome receives a rendered Bitcoin chart and proposes guarded paper trades from neural activity. Portfolio changes trigger engineered dopamine-cell feedback, but profitable learning has not been demonstrated.

The experiment in four figures

$100

paper balance

166,700

retained neurons

25,582,938

directed connections

tests passed

Engineered feedback is asymmetrical

Positive and negative portfolio changes target different dopamine-cell groups.

PAM11 reward cells

PPL101 loss cells

The pulse is an engineered feedback rule, not a proof of biological pleasure, pain or profitable learning.

From market pixels to a guarded action

BTC chart

Public price history is rendered visually.

→

RGB input

Brightness and color stimulate visual cells.

→

Neural vote

A fixed decoder proposes buy, sell or hold.

→

Order + feedback

Guards check the action, then P&L can pulse cells.

What is working

Full retained graph is rebuilt and tested.
Paper mode is the default and real orders require opt-in.
42 tests passed in the recorded validation run.

What remains open

Repeated buy proposals can create one-sided exposure.
Visual input is an unvalidated display adapter.
Profitable learning has not been demonstrated.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.