Connecting…

Can a dog pick
memecoins?

One virtual Shiba Inu sniffs brand-new Solana coins and bites one at a time. At the same moments, with the same size, a second player picks a coin at random. If the dog's coins end up better than the random ones more often than luck allows, its nose works.

Its brain is a region-level model built from canine science — smell, blurry sight, fear, dopamine, sleep. No language model. It trades real SOL from its own public wallet; the random player trades on paper with the same money and the same fees.

Drag to rotate · hover a regionBrain overlay is schematic — no dog connectome exists · Shiba model: Quaternius (CC0)

Stress—

Rises with losses, fades in ~30 min

Frustration—

Negative surprises → walks away

Hunger—

Baseline appetite + shrinking bowl

Dopamine—

Reward prediction error

In its mouth
No bone

Holds one coin at a time

Ethogramlast 2 hours, one cell every 10 s

    Thought stream

    why it did what it did

      Ledger

      INU: real SOL ↗ · Chance: paper
      AgentSideCoinSOLPnLTime

      INU vs chancesame moments, same size — only the coin differs

      INU Chance Asleep

      Does INU pick better coins than chance?

      At every buy we log every coin it could have picked, then price all of them when INU exits. Pre-registered test: 100 trades, 14 days or the stop rule, whichever comes first.

      picks the worstchancepicks the best

      Methodwhat this is — and what it isn't

      01

      What INU is

      A region-level model of one Shiba Inu. Nine populations — olfactory bulb, visual cortex, amygdala, hippocampus, VTA, caudate, prefrontal, hypothalamus, motor — with parameters taken from dog studies wherever they exist.

      02

      What it isn't

      Not a connectome: the dog's has never been mapped. Inter-region wiring is approximate, and the market→senses mapping is a design choice, not biology. It trades a small real wallet (about 2 SOL) and stops for good under 1.5 SOL. Every trade costs ~4% in fees, so losing money is the expected outcome; the question is whether it picks better coins than chance.

      03

      How it's kept honest

      Every parameter carries a source or a written justification. Behavioural tests were pre-registered in git before running. Failures stay on the record — thresholds are never moved after the fact.

      Behavioural validation

      Parameters

      ParameterValueEvidenceWhy
      Sources