- REGION
- —
- MODEL
- —
- SIGNAL
- —
- CONFIDENCE
- —
Choose a wire to begin a fictional defense-training run.
PONS LAUNCH SEQUENCE · RECORDING RIG
Watch it think.
Watch it launch.
Launch token
- Waiting for motor output…
A tiny army of 177-neuron defenders, training to stress-test AI systems and report weaknesses safely back to their deployers. Real connectomics. Extremely unserious commanders.
Recording prototype: no token is deployed and no displayed trade is real. Every choice is logged; the transaction boundary is visibly locked. Profiles and findings are fictional training data—not scraped people. “Defense army” is satire; the project does not infiltrate systems, decrypt models or extract private data.
WHAT IS REAL
A published wiring diagram is driving the display.
The graph comes from electron-microscopy reconstruction of one larva. Light stimulates named photoreceptors; activity propagates over the published weighted connections; named motor neurons produce the displayed left and right drive.
This is an exploratory point-neuron model—not a claim that the animal’s mind has been uploaded. Synapse polarity and cellular dynamics are incomplete, so assumptions must remain visible and auditable.
The researchers reconstructed one Ciona larva from ultrathin serial sections and identified its cells and synaptic contacts.
Rows are presynaptic cells, columns are postsynaptic partners, and each value measures cumulative contact depth.
This site converts that matrix into a directed graph. The light-field control changes photoreceptor input; activity decays and propagates across weighted edges.
Average left and right motor-neuron activity drives the displayed turn and tail motion. Small random noise keeps each run alive.