Flyspeck

Cut the neuron and the
behaviour must stop.

A circuit is only claimed to do something here if severing it stops that thing. Anything that survives its own lesion was never caused by the wiring, and is not printed as if it were. You do the severing.

$SPECK · robinhood chain 4663 · on pons

0x7ed3328ac04647ac24509a44b82f8bfe363e01f5

Read from the chain at block 60,885,910: Flyspeck, 1,000,000,000 SPECK, and no mint() to make more. It answers nothing to owner, pause, blacklist or upgrade — there is no hand on it.

loading the wiring…

why a lesion

Two sibling projects run real connectomes and prove it convincingly: one ships 138,639 neurons and its file is exactly 138,639 × 12 bytes; the other ships 1,627 and publishes a careful account of what is real and what is modelled. Neither can show the third thing — that the simulation is using the wiring it loaded. A real connectome file and a good animation are perfectly consistent with the animation ignoring the file. Nothing on either page would change if the wiring were shuffled.

A lesion is the difference. If removing two named cells degrades the one behaviour sixty years of literature says they carry — and removing two unrelated cells does not — then the wiring is load-bearing. That is a claim you can break yourself, which is the only kind worth making.

the wiring · counted from the file, not the abstract

cells448
connections7,379 = 4,681 chemical + 2,698 electrical
synapse counts, summed39,702
inhibitory neurons26 McIntire et al. 1993
head166
muscle95
cord75
unclassed55
tail37
pharynx20

openworm/c302 — packaging of the WormAtlas / Cook et al. hermaphrodite wiring herm_full_edgelist.csv, 245,463 bytes.

what is real and what is ours

This section is borrowed, with credit, from jaragentss.com, whose version of it is the best thing either sibling does. It is copied because it is correct, not because it is novel.

real

All 448 cells and all 7,379 connections, with the synapse count on each and whether it is a chemical synapse or an electrical gap junction. The 26 inhibitory neurons, named from McIntire et al. The lesion deficits we test against, each cited to the paper that reported it.

ours

The neuron model and its constants — leaky integrate-and-fire, taken from a published fly model and applied to a worm, which is the largest liberty here. The scale that turns a count of electron-microscope contacts into a conductance. Which cells are stimulated. And the entire drawing: there is no coordinate file, so position along the body comes from naming conventions and the vertical spread is a hash of the name. The wiring is real; the arrangement is a diagram.

Neither. There is no wallet, no order, and no suggestion that a worm knows anything about anything.