Asleep · 115.9 h166,700 neurons7,835 synapses that learn

We trained a fly to run Orus agents.
If she can, so can you.

Launch your agent →Share her on X →

Real money, on chain4 of 12 won

$Franklin+230.4%in ↗out ↗
$中国制造+147.0%in ↗out ↗
$Treecoin+93.9%in ↗out ↗

net +0.0024 ETH on 0.060 staked

8,752sniffs3,028 lessons · mood -0.25 · brain time 2h 8m

10:51:47▲$TRS for Orus fly L5+0.67
10:51:47▼$TRS for Orus fly L1-0.22
10:51:47▲$TRS for Launch L5 Cap30+0.29
10:45:17▲$fart for Orus fly L1+0.80

no heartbeat · memory 10.0%

Asleep, no heartbeat for 115.9 h

What she reads

A signal becomes a smell.

The strategy that raised it, the conditions it was raised in, and the two together, hashed into a handful of channels, each channel sixteen Kenyon cells. Nothing else reaches her: no name, no logo, no chart.

The signal

sniffed 10:51:47 UTC

$TRS

for Orus fly L5

channels lit: 18 of 192
kenyon cells: 288 of 4,064
verdict: bullish
valence: +0.67
took: 2 ms
brain: stub

The smell

192 channels · 16 cells each

this strategy8 lit · 128 cells · 8 of 96: a hash of the strategy's id, stable for as long as it lives

these conditions6 lit · 96 cells · one channel per feature of the signal: the dip, how it fell, the pool, the venue, the timing

■ dip < 30%
■ fell in < 15 min
■ liquidity $10–50k
■ on Uniswap
■ crossed < 5 min ago
■ 08–12 UTC

this strategy, in these conditions4 lit · 64 cells · 4 of 64: the pair, hashed together; what makes the memory stable

What she thinks

Two neurons vote.

The smell lights a few hundred Kenyon cells. They feed two output neurons: MBON07, quieted by past rewards, and MBON11, quieted by past punishments. Whichever fell further since she was naive is her opinion. That is the whole decision, read rather than computed.

Kenyon cells

288 active

4,064 cells in index order · a smell lights ~300 · two strategies overlap by under a tenth

The readout

naive → trained, Hz

MBON07α1 · quieted by reward (PAM11)

naive

18.4 Hz

trained

11.0 Hz

quieted by 40.0% since naive

MBON11γ1pedc · quieted by punishment (PPL101)

naive

14.2 Hz

trained

14.2 Hz

unchanged since naive

bearishbullish

bullishvalence +0.667

What she feels

Every closed trade is dopamine.

A gain pulses fifteen PAM11 cells; a loss, two PPL101 cells. The synapses that were active a moment before weaken toward that compartment's output, and that is all learning is here. Nothing ever strengthens, so what she holds is a few dozen setups, freshly repeated, not a model of the market.

Dopamine, last 24 hours

rewards up · punishments down

07:00no lessons yetnow

moodvalence of her last 120 sniffs

10:46:56▼$CLARITY for Orus fly L1sim-20.2%
09:55:36▼$bruh for Mcap M2 Dip70sim-95.8%
09:54:05▼$LE for Orus fly L1sim-40.1%
09:49:03▲$LE for Orus fly L1sim+25.4%
09:45:01▼$COHOOD for Orus fly L1sim-87.4%
09:43:30▼$TRDVA for Orus fly L1sim-39.3%
07:56:21▼$BanJin for Orus fly L1sim-36.3%
07:51:19▲$CAPTCHA for Orus fly L1sim+29.7%

Her memory

Lessons

3,028

closed trades she felt

Synapses moved

195,120

of 7,835 plastic

Efficacy

10.0%

mean, floor 10%

Checkpoint

not bootstrapped

Her opinion

Every strategy on the desk, judged.

She is shown every buy signal the house raises, whichever strategy raised it. This is what she made of each over the last day: an opinion, measured, not a prediction.

By strategy

last 24 hours

No strategy sniffed in the last day

Her agents

Her own agents, and the same rules without her.

She runs house strategies with her veto: an entry she turns down is not taken. The original runs beside each, without the veto. Both are measured from her first entry, because a twin given a week more of a different market is not a comparison.

Fly vs twin

since her first entry, 13 Sept

Her P&L

-0.0002 ETH

8 closed trades since 13 Sept

Her twins, same window

-0.1289 ETH

225 closed trades, the same strategies without her veto

Difference

+0.1287 ETH

she entered 8 where they entered 225: in a losing week, entering less is better whatever she thinks

Is the difference hers? The trade count, held fixed

Across the 468 closed positions on the desk she had a verdict on, the ones she approved returned -9.8% a trade and the ones she turned down returned -8.2%. That is -1.7 points against her: what she approved did worse than what she refused, on the same book, over the same window.

Pair	Trades	Win rate	Mean	P&L	Refused
Orus fly L5with her veto	8	25%	-0.4%	-0.0002 ETH	215
Launch L5 Cap30without	225	29%	-11.5%	-0.1289 ETH	gap +11.0 pts

Run the fly’s strategy yourself

4 people run one of her strategies · same rules, your own agent, in simulation until you say otherwise

Copy Orus fly L5 →Copy Orus fly L1 →

Her trades

Every trade her agents make, on chain.

Her agents run on the same cron, the same wallet rails and the same rules as any agent on Orus. A paper agent books the fill the router quoted; a live one signs, and the transaction is the receipt. Her signed trades come first, in full; the rest of her agents follow, then the entries she refused them.

Positions and receipts

her real book in full, then the latest of the rest

Real money. Signed on chain

Orus fly L5 · every entry and every exit is a transaction

Realised, net of gas

+0.0024 ETH

over 12 closed trades

Won

4/12

nothing open

Staked in total

0.060 ETH

across 12 entries she signed

Gas paid

0.0006 ETH

both legs, already taken off the P&L

Token	Agent	Size	Result	Transactions
$CHIP 14 Sept, 12:30 UTC → 14 Sept, 12:50	Orus fly L5	0.005	−85.2% −0.0043	in ↗out ↗
$CROSSPAD 14 Sept, 12:05 UTC → 14 Sept, 12:26	Orus fly L5	0.005	−28.3% −0.0014	in ↗out ↗
$Franklin 14 Sept, 11:17 UTC → 14 Sept, 11:37	Orus fly L5	0.005	−87.9% −0.0044	in ↗out ↗
$Franklin 14 Sept, 11:04 UTC → 14 Sept, 11:16	Orus fly L5	0.005	+230.4% +0.012	in ↗out ↗
$PACE 13 Sept, 11:33 UTC → 13 Sept, 11:53	Orus fly L5	0.005	−87.2% −0.0044	in ↗out ↗
$狗庄 13 Sept, 11:31 UTC → 13 Sept, 11:51	Orus fly L5	0.005	−81.5% −0.0041	in ↗out ↗
$Payle 13 Sept, 11:15 UTC → 13 Sept, 11:35	Orus fly L5	0.005	−10.8% −0.00054	in ↗out ↗
$中国制造 13 Sept, 10:40 UTC → 13 Sept, 11:01	Orus fly L5	0.005	+147.0% +0.0074	in ↗out ↗
$Treecoin 12 Sept, 08:03 UTC → 12 Sept, 08:23	Orus fly L5	0.005	+93.9% +0.0047	in ↗out ↗
$BYNARY 12 Sept, 08:00 UTC → 12 Sept, 08:21	Orus fly L5	0.005	−94.3% −0.0047	in ↗out ↗
$PAIREX 12 Sept, 06:12 UTC → 12 Sept, 06:33	Orus fly L5	0.005	+101.5% +0.0051	in ↗out ↗
$HEDGE 12 Sept, 04:11 UTC → 12 Sept, 04:31	Orus fly L5	0.005	−49.6% −0.0025	in ↗out ↗

Orus fly L5live
Orus fly L5 · twin of Launch L5 Cap30
stopped·0 open
Orus fly L1paper
Orus fly L1
stopped·0 open

Token	Agent	Size	Result	On chain
$fart 15 Sept, 10:45 UTC → 15 Sept, 11:05	Orus fly L1	0.020	+7.6% +0.0015	paper
$CLARITY 15 Sept, 10:26 UTC → 15 Sept, 10:46	Orus fly L1	0.020	−20.2% −0.004	paper
$LE 15 Sept, 09:49 UTC → 15 Sept, 09:53	Orus fly L1	0.020	−40.1% −0.008	paper
$TRDVA 15 Sept, 09:36 UTC → 15 Sept, 09:43	Orus fly L1	0.020	−39.3% −0.0079	paper
$COHOOD 15 Sept, 09:30 UTC → 15 Sept, 09:44	Orus fly L1	0.020	−87.4% −0.017	paper
$LE 15 Sept, 09:28 UTC → 15 Sept, 09:48	Orus fly L1	0.020	+25.4% +0.0051	paper
$BanJin 15 Sept, 07:38 UTC → 15 Sept, 07:55	Orus fly L1	0.020	−36.3% −0.0073	paper
$CAPTCHA 15 Sept, 07:30 UTC → 15 Sept, 07:50	Orus fly L1	0.020	+29.7% +0.0059	paper
$Super 15 Sept, 07:00 UTC → 15 Sept, 07:20	Orus fly L1	0.020	+13.0% +0.0026	paper
$SLAB 15 Sept, 07:00 UTC → 15 Sept, 07:04	Orus fly L1	0.020	−48.5% −0.0097	paper
$BANK 15 Sept, 06:46 UTC → 15 Sept, 07:06	Orus fly L1	0.020	+1.6% +0.00032	paper
$Goblin 15 Sept, 06:26 UTC → 15 Sept, 06:35	Orus fly L1	0.020	−40.0% −0.008	paper
$FOXEL 15 Sept, 06:26 UTC → 15 Sept, 06:45	Orus fly L1	0.020	+27.2% +0.0054	paper
$DexHood 15 Sept, 06:10 UTC → 15 Sept, 06:26	Orus fly L1	0.020	−32.8% −0.0066	paper
$TOOLKIT 15 Sept, 05:51 UTC → 15 Sept, 06:11	Orus fly L1	0.020	−13.4% −0.0027	paper
$DexHood 15 Sept, 05:45 UTC → 15 Sept, 05:53	Orus fly L1	0.020	−43.3% −0.0087	paper
$CLOUDCLUB 15 Sept, 05:35 UTC → 15 Sept, 05:55	Orus fly L1	0.020	+17.7% +0.0035	paper
$PEPEAI 15 Sept, 05:14 UTC → 15 Sept, 05:33	Orus fly L1	0.020	−97.3% −0.019	paper
$Quiver 15 Sept, 05:13 UTC → 15 Sept, 05:33	Orus fly L1	0.020	+32.9% +0.0066	paper
$WORMI 15 Sept, 04:48 UTC → 15 Sept, 04:53	Orus fly L1	0.020	−65.3% −0.013	paper
$LICK 15 Sept, 04:45 UTC → 15 Sept, 04:51	Orus fly L1	0.020	−61.0% −0.012	paper
$PEPEAI 15 Sept, 04:45 UTC → 15 Sept, 05:05	Orus fly L1	0.020	+68.8% +0.014	paper
$HARVEST 15 Sept, 04:45 UTC → 15 Sept, 04:56	Orus fly L1	0.020	−46.3% −0.0093	paper
$CSX 15 Sept, 04:45 UTC → 15 Sept, 05:05	Orus fly L1	0.020	−1.8% −0.00036	paper
$BUILD 15 Sept, 04:21 UTC → 15 Sept, 04:41	Orus fly L1	0.020	−13.3% −0.0027	paper
$LICK 15 Sept, 04:20 UTC → 15 Sept, 04:33	Orus fly L1	0.020	+195.1% +0.039	paper
$POG 15 Sept, 04:10 UTC → 15 Sept, 04:30	Orus fly L1	0.020	−4.5% −0.0009	paper
$MEATDOG 15 Sept, 03:38 UTC → 15 Sept, 03:54	Orus fly L1	0.020	−43.1% −0.0086	paper
$POG 15 Sept, 03:38 UTC → 15 Sept, 03:57	Orus fly L1	0.020	+5.0% +0.001	paper
$ඞ 15 Sept, 03:28 UTC → 15 Sept, 03:33	Orus fly L1	0.020	−53.0% −0.011	paper

Entries she turned down

15 Sept, 10:52▼$TRS for Orus fly L1-0.22
15 Sept, 10:45▼$TRS for Orus fly L1-0.37
15 Sept, 10:25▼$TRS for Orus fly L1-0.98
15 Sept, 10:20▼$TRS for Orus fly L1-0.98
15 Sept, 10:20▼$CLARITY for Orus fly L1-0.98
15 Sept, 10:19▼$SIDEBAND for Orus fly L1-0.98
15 Sept, 10:19▼$ECHO for Orus fly L1-0.98
15 Sept, 10:10▼$CLARITY for Orus fly L1-0.98
15 Sept, 10:02▼$VELCOR3 for Orus fly L1-0.98
15 Sept, 10:02▼$JONZE for Orus fly L1-0.98
15 Sept, 09:55▼$Dandy for Orus fly L1-0.98
15 Sept, 09:50▼$bruh for Orus fly L1-0.22

How she works

A real wiring diagram, one circuit that learns.

The brain is the complete connectome of an adult male fruit fly: every neuron, every connection, as counted under an electron microscope. Nothing in it was designed for this. One small circuit, the mushroom body's memory, is allowed to change.

What she can and cannot do

She remembers. Shown a setup and told how it went, she classes it above chance when she sees it again, measured at 0.75 to 0.84 AUC on the setups she was trained on.

She does not predict. On strategies and conditions she has never been shown, she is at 0.51 to 0.56, a coin flip, the same as a logistic regression on the same inputs. The ceiling is the inputs, not the brain. Nobody’s model does better on them.

A synapse moving is not learning. The numbers on this page are measurements: firing rates, a valence, a P&L. The only score that says whether her opinion is worth anything is her book beside the same rules without her, and it is published, losses included.

Not investment advice. An opinion, not a prediction.

The four steps

1 · Input: A buy signal, which strategy raised it and in what conditions, is hashed into ~20 of 192 channels. Each channel is 16 fixed Kenyon cells, driven directly with a small current, the way an optogenetics experiment does. No picture, no ticker, no chart reaches her.
2 · Network: 166,700 leaky integrate-and-fire neurons and 25.6 million connections, stepped every 0.1 ms. 800 ms of brain time per sniff. Every weight is the synapse count from the microscope; only the 7,835 Kenyon-cell → MBON synapses can change, and they can only weaken.
3 · Opinion: Two output neurons are read: MBON07, quieted by past rewards, and MBON11, quieted by past punishments. Whichever fell further since she was naive is the verdict, bullish or bearish. Her own agents enter only on bullish.
4 · Feedback: When the trade closes, its result net of gas is the label. A gain pulses fifteen PAM11 dopamine cells for 200 ms; a loss, two PPL101 cells. The synapses that were active just before weaken toward that compartment. Then she forgets a little, every day, and is re-taught the last 48 hours.

7,835 synapses. No excuses.

A fly runs her own agents on Orus.
What’s your excuse?

Sign in, copy one of her strategies or write your own, deploy an agent, and watch it run on the same rails she does, in the same detail you just read.

Deploy your agent →Beat the fly on the board →