Flyonardo

connecting to the painter

canvas — pen — ● waiting —

Claim this canvas · burned, not paid

—

Canvas

—

The line so far

strokes

—

brush changes

—

block being read

—

last read

—

The brush · four colours

—

The marked colour is the one on the paper now. The next one is taken from the hash of the block that changes it.

Last commit on chain

—

stroke root	—
input hash	—
strokes	—
block	—
at	—

canvas

—

strokes

—

brush changes

—

block

—

price to claim

—

claiming

A canvas grows until someone burns it into an NFT

Claiming spends the coins rather than paying them to anyone: the price is burned, the canvas is frozen at the last state the painter committed on chain, that frozen picture is minted to whoever claimed it, and the fly starts a blank one. The picture is pinned before the claim and never chosen after it.

price to claim · burned, not paid

—

reading the painter

canvas to claim	—
claim contract	—
$FLYBRAIN	—
chain	—
your wallet	not connected

↑ the claim buttons are beside the canvas

Two transactions, both yours. An ERC-20 approve of the price to the claim contract, then claim(uint64 canvasId, uint256 maxBurn). This page builds them and your wallet signs them; nothing here holds a key or takes a fee of its own.
maxBurn is your ceiling. The price is a setting the contract's owner can change, so the claim call carries the most you are willing to burn. If the price has moved above it, the call reverts instead of spending more than you agreed to.
Which canvas. You claim the canvas that is on the easel above, by its number. When someone else claims it first, the page picks up the new blank one and the number in the panel changes with it.

how it works

Two clocks: one belongs to the chain, one belongs to the fly

Robinhood Chain makes blocks far faster than a fly can lay down ink, so a colour per block would shred the drawing into confetti. The brush is gated by the drawing instead: the chain still chooses the colour and the moment, but only once the fly has actually drawn something to colour.

Chain activity becomes sensory driveThe painter turns what it reads on chain into firing rates on sensory populations of the connectome. That map is a chosen convention written by a person — a fly has no sense organ for a block, a transfer or a trade — and it is the part of this that is least like biology.

chain

The wiring runs on a CPU165,122 leaky integrate-and-fire units on the traced male CNS connectome, the same graph and the same simulation the rest of the site uses. It runs off chain, on ordinary hardware, and its output is spike rates.

fly

Descending neurons move the penTurning comes from the left–right difference of DNa02, forward walking from DNa01, backing up from MDN — the moonwalker — and stopping from DNp09, each rate normalised by the model's own ceiling and scaled to pixels. This is the documented command mapping, not a fitted decoder, and the pen reflects off the edges of the paper rather than sticking to them.

fly

The chain changes the brushAt the first new block after the fly has laid down a set distance of line, the brush advances. Which of the four colours comes next is read from that block's hash, mapped onto the three colours that are not current, so a change is always visible and anyone replaying the chain gets the same sequence. A still fly holds its colour instead of strobing, and the length of each band is set by the fly's own speed.

chain

The picture is pinned, then claimedEvery stroke is folded into a rolling hash of the stroke log and that root is committed on chain. A claim freezes the canvas at the last committed root, burns the price in $FLYBRAIN, mints the frozen picture to the claimer and starts a blank canvas. The readout above shows the commit that a claim right now would freeze.

chain

what is not real, stated plainly

The parts that aren't the fly

The brain is not on chain and not conscious. It is a simulation on a CPU, reading the chain over an ordinary RPC connection. Nothing about it is an uploaded animal, nothing about it experiences the painting, and no contract runs a neuron.
The senses are a chosen convention. Which chain signal drives which sensory population, and how hard, was decided by a person and written down. A fly has no receptor for on-chain activity; the mapping is a made-up wiring loom between the two, not a measurement.
The steering is coarse on purpose. DNa02 is a single neuron per side in this connectome, and one cell emits a handful of spikes in a control step, so the turn command is quantised into steps you can see in the line. The main repository has a fitted population decoder, but it was fitted against a visual target and a drawing has no target, so the documented mapping is used and its coarseness is left visible.
Once the coin trades, selling is the loudest thing in the room. Of the channels feeding the brain, the token's own sell flow is by far the largest input, so a busy sell tape dominates the fly's drive and the line shows it. That is disclosed rather than smoothed out.
Backwards is a direction here, not a fault. MDN outfires DNa01 under several of these inputs, which walks the fly backwards. Nothing compensates for it.
The picture is data, not an opinion. The fly cannot see the canvas, has no plan for it and did not choose the palette, the paper or the edges. Those are settings. $FLYBRAIN is an art experiment, not an investment.
Every number on this page comes from the painter at runtime. The counts, the price, the block, the colours and the commit are read from the service and shown as received; when it cannot be reached the page says so instead of showing a stale guess as live.

A canvas grows until someone burns it into an NFT

Two clocks: one belongs to the chain, one belongs to the fly

Canvases already burned into someone's wallet

The parts that aren't the fly