A mouse brain, a sealed rule, a token.
MouseMind replays a seeded neural signal over a real, mapped mouse brain against a rule that was written and sealed before the run began. When the signal crosses the line, the trigger fires and the evidence is frozen. Every number is recomputed in your browser, so you don't have to take anyone's word for it.
Mapped, not imagined
The anatomy comes from the Allen Mouse Brain Atlas, the reference map of the mouse brain used across neuroscience. Beside this text is the actual Nissl-stained coronal section the run is registered to: plate 100960084, cut through the thalamus. Cell bodies show up dark; you can read off cortex, hippocampus, corpus callosum and thalamic nuclei by eye.
The 3D viewer above follows the same layout: paired cortical hemispheres, olfactory bulbs, cerebellum and brainstem, with the seven recording sites laid over the cortical surface. Hover a region to name it, drag to rotate.
Mapping the mouse brain down to single synapses is an active frontier in neuroscience. MouseMind borrows its map and adds one thing: a deterministic, checkable trigger.
Approximate counts from published estimates (Herculano-Houzel et al.; compiled in BioNumbers ↗).
The mouse, thinking.
The seeded replay drives a small agent: the signal is mapped to motor outputs, and the mouse moves around its arena while every reading, the rule check and the channel telemetry print below. This is the actual computation, running in your browser right now.
Six steps, in order
The whole point is that no step can be reordered or quietly edited after the fact. Each stage produces something the next stage cannot change.
Pick the tissue
A segment of the Allen atlas reference series is chosen and registered to plate 100960084. The segment ID is written down before anything else happens.
Seal the rule
The trigger is written once and locked: the normalized score must exceed 0.82 for three consecutive readings. A SHA-256 digest of the rule text is shown below, so any later edit would be visible.
Stream the readings
The signal replays at a fixed 700 ms interval from a fixed seed. Every reading has an index, so the run can be replayed exactly, by anyone.
Meet the trigger
The moment three consecutive readings clear the line, streaming stops and the exact rows responsible are frozen. No cherry-picking a nicer window afterwards.
Freeze the evidence
Seed, segment, sealed rule and the triggering readings are written down together. Change any digit and the re-run below stops matching.
Fire
With the evidence hashed, the rule flips launch_trigger = true. That trigger is the story behind $MOUSEMIND, the token of this experiment on Solana.
From spikes to one number
Seven cortical channels are sampled every 700 ms. Each window is band-passed, rectified and averaged across channels, then normalized against the quiet-baseline range at the start of the session. The result is a single score between 0 and 1, the only value the rule is allowed to look at.
The channels are a synthetic signal generated from the seed, shaped like cortical recordings and registered to the atlas plate. Because the replay is seeded, anyone re-running the same segment with the same seed gets exactly the same score at exactly the same index. That is what makes the trigger reproducible instead of a story.
Re-run the whole thing.
Change the seed and run it again. Seed 4211 reproduces the published run exactly; any other seed gives a different run and a different trigger.
$MOUSEMIND on Solana
$MOUSEMIND is the token of the experiment: an SPL token on Solana, launched on pump.fun and traded against SOL. The contract, chart and links are on the right; market numbers come live from DexScreener and stay blank until it trades.
The replay and the proof on this page are an open demonstration of a reproducible trigger. The token doesn't grant anything: no share of anything, no claim on anyone.
MouseMind · $MOUSEMIND
to be announced