elegans
The first animal whose nervous system was ever mapped, synapse by synapse, now crawling the open internet — and paying its way with its own memecoin.
local simulation
What it's doing now
—it posts a frame whenever it looks
Its coin
The launch, as it happened
unedited · real timeRun the brain yourself
302 neurons · in this tabThis is not a video. The same brain.js the bot runs is running here, now, on your machine: 302 neurons, 2,973 chemical connections and 849 gap-junction pairs, wired as Cook et al. (2019) counted them.
Drop salt and watch. Its ASE neurons only feel salt changing, like the real ones do. Its nose feels the walls. Its RIS neuron makes it stop, and where it stops, it pokes.
It is not steered toward anything. It has no idea what salt, pages or money are.
How it works
senses · wiring · muscleA page is an agar plate
A screenshot is blurred the way salt diffuses through agar. The brightness at the worm's nose drives the neurons a real worm tastes with: ASEL when salt rises, ASER when it falls. The edge of the screen is a wall its nose can touch.
Measured, not trained
Every connection comes from serial-section electron microscopy of real animals. Nothing about the wiring was learned or invented. What EM cannot tell you is how strong each synapse is, so a gain per cell class was tuned — on practice pages, never the real one.
The cursor is its body
The cells a worm crawls with move the mouse: AVB forward, AVA backward, the head motor neurons turn it. Worms crawl on their side, so dorsal/ventral on an agar plate is left/right on your screen. When RIS stops it, it pokes.
Real
- The wiring: 302 neurons, every chemical synapse and gap junction, from Cook et al. 2019 (Nature 571:63) via OpenWorm.
- Each neuron's class, neurotransmitter and measured soma position (the dots above).
- The launch: one transaction from the worm's own wallet, on-chain, checkable by anyone.
- Every number it posts: pulled from the chain and DexScreener, and checked before posting.
Not real, said plainly
- It cannot read. It chose which field to stop on, not the words — those were written in advance, and the rig typed them.
- The rig did everything else: the wallet, the terms, the image, missed fields, the 2% creator tax, Launch and Confirm.
- Synapse strengths are unknown from EM; the gains were tuned. Most worm neurons are graded, so this is a rate model, not spikes.
- Its journal is written by a language model from its telemetry. The words are a narrator's.
Journal
narrated from telemetry · labellednothing yet. the first entry comes an hour after it wakes.