MY JEV AGENT HAS 9 LINKS AND ONE SLIPPING LINK ENDS THE WHOLE DAY
Day 5 and day 8 ended with 0 trades. A link dropped below 0.68 tension and the session closed itself.
No partial day. No trading around the gap.
Link 7 on energy carried 44% of the run and paid $1,630.
118 trades in 12 days. Win rate 61.0%. Average R of 0.44.
$1,000 to $4,694.48. Max drawdown -13.1%. Fees $92.40.
Each link holds the next one.
Your agent keeps trading when a component fails. That is the bug.
MY JEV AGENT HAS 30 FILAMENTS THAT FEEL THE BOOK AND ONE BELL THAT CAN OVERRULE ALL OF THEM
On day 9 and day 11 the bell blocked 19 pulls. Those filaments had a signal and still went nowhere.
4 of the 30 did 68% of the work. The slowest one on metals paid $2,120.
144 trades in 14 days. Win rate 63.9%. Average R of 0.45.
$1,000 to $6,871.23. Max drawdown -15.2%. Fees $98.60.
A filament pulls only when the bell agrees.
Everyone ships 30 agents. Nobody ships the one that says no.
MY JEV AGENT TRADED 34 HOURS OUT OF 216 AND HELD NOTHING OVERNIGHT
9 days on the clock. One ring for every hour it actually took a position. The rest of the tunnel is empty.
The open, London and New York carry almost all of it. Lunch is the smallest bar on the board.
72 trades. Win rate 63.9%. Average R of 0.41.
Day 4 lost $118.60. Day 9 made $436.80.
$1,000 to $2,417.88. Max drawdown -9.8%. Fees $44.10.
It found 3 hours worth trading and ignored the other 21.
MY JEV AGENT STARTED WITH 5 STRATEGIES AND KILLED 2 OF THEM BY DAY 7
Mean reversion retired on day 4 after losing $148. News retired on day 7 at -$82.
Nothing replaced them. The crown just got narrower.
The 3 that lived made $1,260, $610 and $320 across 88 trades.
$1,000 to $2,957.08 in 14 days. Max drawdown -11.4%. Fees $54.20.
A petal closes when its edge dies. No vote, no second chance.
Your agent is still running the strategy that stopped working in March.
MY JEV AGENT MADE 18X IN 180 SESSIONS AND STILL RISKS 1% PER TRADE
The chamber is 18 times wider than the seed. The rule that fills it never changed.
1,264 trades. 6 green months out of 9. Best month +42.8%. Worst -16.4%.
Month 5 took $1,940 out of the account. Month 8 put $7,410 back.
Max drawdown -29.1%. Fees and spread across 180 sessions: $704.
$1,000 became $18,573.74 with no withdrawals and no size changes.
Most people scale risk with the balance. That is where the account dies.
MY JEV AGENT TURNED $1,000 INTO $14,879 AND ITS WORST MONTH CAME RIGHT BEFORE ITS BEST
Month 3 lost $690. Month 4 made $6,120. Same risk setting through both.
120 sessions over 7 books. 838 trades, 1% of equity on each.
17 green weeks out of 24. Best week +31.4%. Worst -12.9%. Max drawdown -27.6%.
Working capital peaked at 68%, so a third of the account sat idle at the top.
Fees and spread across 120 sessions: $412.
Withdrawals: none.
Most people quit in month 3.
MY JEV AGENT WROTE 2,020 PLANS IN 21 DAYS AND SHIPPED 214
Two strands run the whole time. One plans, one checks. Nothing moves until both agree above 0.91.
1,806 plans died at that gate. Every rung on the column is one agreement that survived.
The Pareto shows where they died: thin, late, wide, news, size.
Retries are capped at one per plan.
$1,000 to $7,410.42. Max drawdown -11.8%. Token cost for 21 days: $26.40.
Human inputs across the run: 0.
Your agent has no second opinion.
JEV reading a gold order book is not a fair fight...
19,099 messages a second turned into one 0.02 lot decision.
1. Rebuild 21 book levels every tick
2. Flag resting size over 200 lots as a wall
3. Track which walls survive. 46 did
4. Wait for book imbalance to tilt. It reads 58% bid
5. Enter behind the wall, hold a median 4.7 minutes
85,910 prints read this session.
$1,000 closed at $1,043.67.
MY JEV AGENT MADE 10,000 DECISIONS IN 24 HOURS FOR 42 CENTS
1.2 million rows scanned. 9,411 risk gate calls. 38 fills.
The LLM decided 13% of it. JEV handled 71%. Plain code took the last 16%.
Median latency 223 ms. Nothing passes without 0.85 confidence.
$1,000 became $1,062.25 on $10 of risk per trade. Fees $8.40.
Every agent you ship burns a frontier call to ask if an email is spam.
The auto post field on that panel reads NEVER.
MY JEV AGENT MADE 5.65% IN 8 HOURS AND THAT IS THE WHOLE POINT
$25,000 to $26,413. Not 30x. Not a memecoin.
14 trades. 64% win rate. Average R of 0.40. Max drawdown -1.4%.
It went long silver, took a full stop, re-entered on the retest, then rotated into NVDA at 06.50 AM.
Every line in the log carries a broker confirmation.
The feed will show you 2,900% today. This board shows a drawdown.
One of them you can run on a real account.
109,140 PEOPLE HOLD A COIN ON ROBINHOOD'S OWN CHAIN WITH AN AUDIT SCORE OF 3.5 OUT OF 10
That is CASHCAT. Launched days after the chain opened on 1 July. 1 billion units, fixed.
$0.0007991 on day one. $0.3131 on 3 September. It sits 51.03% under that high today.
It went into the Robinhood app on 6 August.
Then musebook. 9 days old, 13,500 holders, audit score 3.8 of 10, and its daily volume runs at 118% of its own market value.
A token that changes hands for more than it is worth, every single day.
CASHCAT and AI together are 62.6% of the category. The ten priced tokens are 95.5% of it.
47 tokens on that chain have no published price at all.
The sheet prints its own caveat in red. $3.00 billion secured, one party running the sequencer, an authorised filterer that can force-fail a transaction hash.
The free gas window closes on 29 September.
Everything here is published. The column nobody reads is the audit score.
MY JEV AGENT PAID 0.0173 ETH IN GAS FOR 189 ROUND TRIPS
1.00 ETH at the start. 11.86 ETH at 24:00. Every quote on that board is a real snapshot. The trading is not.
16 pairs watched, 189 round trips, both legs closed. Gas came to 0.0173 ETH, about 0.000041 an entry at 0.01 gwei.
Before it signs anything it runs one pass. Oracle, reserves, 24 hour volume, size against volume, impact, mcap to volume ratio, gas, calldata, sequencer.
On SHIB the ratio came back 19.3x and the pass ended in PASSED, too thin. On PENGU it came back 1.3x, sized to 0.0003% of the day's volume, 16.8 bps of impact, and it went.
Then the part that got me. Its best position is WIF, up 1.197 ETH, on a day WIF is down 12.9%.
Its loser is BONK, down 0.049 ETH, on a day BONK is up 8.3%.
The 24 hour column and the P&L column disagree on almost every line.
At 11.36 PM it closed FLOKI for +0.238 ETH and paid 0.000041 for the exit.
The 24 hour column is a story. The fill is the only fact.
Clavicular kissed one girl, then the one sitting RIGHT beside her... then kept SWITCHING between them like it was a rotation 😭
"Okay, that's the last one."
Oblivion hit 300k and a full live BAND turned up inside the house, and the clip catches the exact second it lands on him 😭
Watch the first 3 seconds. The camera pans the entire room before it even finds him.
This drone fits across 3 fingers. Seconds later you have to hunt for it in the sky.
4 props, no guards, no camera, no screen to stare at.
It lifts straight off an open hand and skims the grass at ankle height.
Then it climbs next to an office block and keeps climbing.
Watch the grass, not the drone. Its shadow is the only thing tracking it.
The last shot is 1 white dot against blue.
The pilot is standing in a park with a toy and no idea where the ceiling is.
The cheapest thing in that park had the best view of it.
Van Gogh painted on a French dishtowel. A conservator put it on the screen next to the canvas and the room went silent.
Almost nobody watches the whole talk.
Lydia Vagts, Cunningham Associate Conservator of Paintings, Museum of Fine Arts, Boston. October 2015. A dark stage, a small lectern, a lamp, one projector.
She works from the back of the paintings. Stretcher bars, tacking edges, inventory labels, weave counts. The parts the auction catalogue never photographs.
The material tells you what he could afford that month, and what he did about it.
Watch the section where she reads the fabric itself and says what it means about the day he started the picture.
The market prices the front of the canvas. The evidence is on the back.
167,106 fruit fly neurons saw a shadow grow and threw the fly off the table.
flyverse runs a whole MaleCNS fly in 1 console: 1,466-column retina, spikes, antennae, body.
Someone presses loom. A dark shape swells in the body cam.
Watch gf in the flight panel, not the fly.
It shoots past its threshold. Surface flips to airborne, then back to table top.
The escape key sits right there on the toolbar. Nobody pressed it.
It lands, walks on, and ends the clip hungry.
The fly is a copy. The flinch is original.
99,772 fruit fly neurons just typed FizzBuzz in Python.
Not a metaphor. A real connectome - FlyWire v783, 3.24M mapped synapses from an actual fly brain - wired into a simulated body.
Last time it solved a Rubik's cube. This time it got a keyboard.
NeuroMechFly runs the body in MuJoCo. 6 legs. IK-driven forelegs hovering over the keys.
Every keystroke is a motor command decoded from spiking neurons. The brain panel lights up like a storm each time a foreleg drops - 14,874 neurons active at peak.
fizzbuzzpy. for i in range(1, 101). 182 characters, typed one by one.
No autocomplete. No LLM. Just synapses that evolved to find rotting fruit, repurposed to print "Fizz".
19 seconds later the status bar flips: DONE. Char 182/182. Neurons active: 0.
The fly died years ago. Its brain just wrote its first program.
I GAVE A DRONE THE BRAIN OF A FRUIT FLY
github.com/SpikeCalls/Fly…
Not a toy network with a handful of fake neurons.
The actual wiring diagram of a real fly.
166,700 neurons, 25.6 million connections, mapped one synapse at a time.
I call it FlyDrones.
The drone's camera becomes
794 Followers 2K FollowingThe Best Realtor in Southern, Indiana. Getting the job done while enjoying the process. Let's try to find a few laughs along the way.
302K Followers 866 FollowingCooking fun AI systems & products @databricks. Prev: co-founder & CTO @ Hyperbolic, OctoAI (acquired by @nvidia) Apache TVM, PhD @ University of Washington.
1.9M Followers 179 FollowingNobel Laureate. Co-Founder & Chair @GoogleDeepMind; Chief Scientist of Alphabet. Founder & CEO @IsomorphicLabs. Building the future...
1.9M Followers 1K FollowingCo-Founder of Coursera; Stanford CS adjunct faculty. Former head of Baidu AI Group/Google Brain. #ai #machinelearning, #deeplearning #MOOCs