ROBINHOOD CHAIN // AUTONOMOUS FUTURES DESK

166,000 neurons.
One futures desk.

Every creator fee $FLYPIT earns is swept to a wallet that you do not control and neither do we. A simulated fruit fly brain reads the market as optic flow, steers, and the steering becomes the order.

It does not know what a chart is. It knows the world is sliding past it and it wants to turn into the flow. That turns out to be a position.

CONTRACT not deployed yet

THE DESK--:--:--
EQUITY
--
TODAY
--
POSITION
--
LEVERAGE
--
LAST ORDER waiting for the desk to open
Shape preview only. The desk is not funded, no live orders exist, and this curve is not a track record. It switches to real state the moment the treasury address is set.
THE LOOP

A fly cannot read a chart. It can read motion.

The best documented circuit in the fruit fly brain is the one that turns vision into steering. Wide field motion across the eye drives the horizontal system cells, those drive descending neurons, those drive the wings. We do not teach the fly anything. We hand the market to it in the only language its optic lobe already speaks, and we read what comes out the other end.

01

PRICE

Mark price on the traded market, sampled every 250 ms and held in a rolling two minute window.

02

OPTIC FLOW

The window is rendered as a vertical grating drifting across a 720 by 360 visual field. Drift velocity is price velocity. Contrast is volatility.

03

OPTIC LOBE

The frame is injected into the medulla and lobula plate of the released connectome. Roughly 45,000 neurons per eye fire on it.

04

STEERING

We read the descending neurons that command turns. Net left minus right over a 30 second window is the bias. Magnitude is conviction.

05

ORDER

Bias crosses threshold, the risk layer sizes it, the relay sends it. One market, one position, no discretion.

LIVE READOUT

What the fly is actually looking at.

This is the real encoding running on simulated price, so you can watch the mechanism before the desk is funded. Top panel is the visual field as the fly receives it. Below it is the left and right horizontal system response. The steering bar is the raw output that becomes a position.

VISUAL FIELD 720x360 SIMULATED FEED
HS CELL RESPONSE L / R --
DECODE
PRICE--
FLOW--
LEFT HS--
RIGHT HS--
STEERING--
BIASFLAT
CONVICTION--

Conviction under 0.35 holds flat. The fly spends most of its life not turning hard enough to matter, and the desk respects that.

THE MONEY

Fees go in. Nothing comes back out to us.

$FLYPIT charges a 4 percent tax on every buy and sell on Robinhood Chain. The split is fixed in the contract and the desk wallet is public from day one, so the equity curve on this page is auditable against the chain.

2.5%

DESK TREASURY

Swept to the desk wallet. This is the only money the fly ever trades. It is never withdrawn to a team wallet.

1.0%

LIQUIDITY

Paired and locked. Keeps the book deep enough that the fee stream is not the thing moving the price.

0.5%

COMPUTE

The connectome sim is not free. This pays the GPU bill that keeps the fly awake and the readout live.

50/50

REALIZED PROFIT

Half of every realized gain buys $FLYPIT on the open market and burns it. Half compounds back into the desk.

NO WITHDRAWALS

The desk wallet has no withdrawal path to any address we control. It can trade, it can buy back, it can burn. That is the whole surface.

PUBLIC FROM BLOCK ZERO

Treasury address, order log and equity curve are published before the first trade. If the curve on this page and the chain disagree, believe the chain.

LOSSES ARE REAL

There is no backstop. If the fly draws the desk down, the desk is down. The only thing the caps do is stop it going to zero in one afternoon.

RISK RAILS

The fly steers. The rails decide how far.

A fruit fly has no concept of ruin. Everything between its steering signal and the exchange is there to make sure one bad afternoon does not end the experiment.

RAILSETTINGWHAT IT DOES
Max leverage3xHard cap in the relay. The fly cannot request more, and a malformed signal is clamped rather than rejected.
Position cap35% of equityNotional ceiling per position. Conviction scales size inside this band, never past it.
Daily loss cap8% of equityBreached, the desk flattens and sleeps until the next UTC day. The fly keeps steering, nobody listens.
Flip cooldown15 minutesMinimum time between direction changes. Stops chop from grinding the desk out through fees.
Conviction floor0.35Below this the desk stays flat. Most of the day sits below this.
Kill switch3 of 5 multisigFlatten and halt only. It cannot move funds anywhere. Published signer set.
$FLYPIT

One token. One fly. One position at a time.

Launching on Robinhood Chain. The contract address goes live on this page and on the X account at the same second. There is no presale, no allocation and no round.

ITEMDETAIL
ChainRobinhood Chain
Ticker$FLYPIT
Tax4% buy and sell, split 2.5 desk / 1.0 liquidity / 0.5 compute
Team allocationNone
PresaleNone
Burn sourceRealized trading profit only. No scheduled burns, no fake ones.
Contractnot deployed yet
QUESTIONS

The obvious ones.

Is this actually the Google connectome or is it a mascot?

It is the released wiring. The September 2026 drop from Google Research and HHMI Janelia covers roughly 166,000 neurons and 125 million synapses across brain, optic lobes and ventral nerve cord. We simulate the visual and descending pathways from that dataset. We are not simulating a whole living animal and we do not claim to. We are running a real circuit on a synthetic world and reading its output honestly.

Why would a fly be good at trading?

There is no reason to think it is, and we are not going to dress that up. What a fly is genuinely excellent at is committing to a direction in a noisy visual field and holding it, which is a trend follower with better reflexes and no ego. It will have bad months. The equity curve is public precisely so you can watch that happen.

Can the team touch the desk money?

No. The desk wallet can open positions, close positions, buy back and burn. There is no transfer path to an arbitrary address. The multisig can only flatten and halt. All of this is verifiable before you buy.

What stops it just being a random number generator?

Nothing stops it from performing like one, and if it does, the chart will say so. What we can promise is that the signal is not random by construction. The same price window fed twice produces the same steering output. It is deterministic, logged, and replayable from the published order log.

What happens if the desk goes to zero?

The rails make that slow rather than impossible. If equity hits the floor, the desk stops trading and the fee stream refills it from scratch. The token does not stop existing because the fly had a bad quarter.

Where does FLYHOUSE fit in?

Same brain, different experiment. FLYPIT puts one fly in front of a market. FLYHOUSE puts six of them in a house and streams it. They share the simulator and nothing else.