Senpi 2.0 in the Wild: One Week, Real Wallets, Real Capital
We just opened the Senpi 2.0 waitlist with $100 in free AI credits
Before we ask you to reserve your @senpi_ai Hyperliquid Agent, here’s the honest version of what one looks like in motion: a real strategy we’ve been running on Senpi 2.0 in dev for the past week, on real wallets, with $2k in real capital on the line.
No backtest.
No cherry-picked screenshot.
Every trade below is onchain and verifiable.
Here’s the week.
How it started: a sentence, not a spec
The whole thing began the way Senpi 2.0 is meant to:
In plain English.
The thesis was simple to say and historically painful to build:
The world is splitting into winners and losers, and the gap keeps widening. Go long the structural winners — AI and its infrastructure — and short the laggards. And separately, catch the explosive price discovery when new tokenized equities list on Hyperliquid.
That sentence turned into two distinct autonomous strategies running side by side on Hyperliquid.
No user code.
No config files.
No infrastructure setup.
One strategy trades a K-shaped long/short thesis.
The other trades new-listing events.
Senpi built both, sized them, and has been running them 24/7 since.
🦁 Lion — the K-shaped long/short
Lion’s bet is not simply that the market goes up.
It is that the market is becoming increasingly two-speed — and that the spread between the fast lane and the slow lane is the edge.
The long book holds the structural winners on Hyperliquid XYZ:
These are names Lion expects to outperform regardless of overall market direction. The short book sells the laggards:
Broad U.S. market exposure
Weaker crypto alts
Names showing deteriorating relative strength
The profit does not require the market to fall.
It comes from dispersion:
Long winners rise faster
+
Short laggards fall or underperform
=
Spread capturedThis is long-short, market-neutral-ish, harvesting the gap.
Both books are trend-confirmed and fully autonomous. A producer scores the universe, the runtime handles entries and exits, and every position is protected by a dynamic trailing stop.
No static stop-loss to get hunted.
No human panic button.
Just a floor that ratchets up as the trade works.
Per-name conviction sizing.
No human in the loop.
🐦 Magpie — the new-listing event trader
Magpie plays a different game entirely: the pre-IPO → listing → graduation arc of tokenized equities on Hyperliquid.
When a new name lists on the trade.xyz pre-IPO perpetual market, it has a specific signature:
Extreme funding
Capped leverage
Violent, short-lived price discovery
A narrow window where being early matters
Magpie auto-discovers those listings and trades the event.
This week, Magpie did something important:
Nothing.
No qualifying listing met its bar, so it sat in standby, watching, with capital intact.
That is a feature, not a bug.
An event strategy that forces trades when there is no event is just a fee generator. More on that discipline below.
What’s live right now
Three wallets.
All onchain.
All verifiable.
The open positions: Lion long book
All three are protected by trailing stops with floor prices already set.
No raw stop-loss orders.
No raw take-profit orders.
The exit engine manages it dynamically as price moves.
The week’s results — the good and the ugly
Here is the part most “AI trading” posts skip.
The full ledger: winners and losers.
All-time, since June 17:
Lion long book — 7 closed trades
🟢 Samsung +$51.52 Cleanest trade of the week
🟢 SpaceX, round 1 +$41.65 Closed in profit, then re-entered
🟢 Intel +$40.87 Rode $128 → $131
🔴 ARM -$34.14 Wrong, cut fast
🔴 NBIS ×2 + SpaceX ~-$4.00 Small, quick exits
Lion short book — 5 closed trades
🟢 AVAX short +$87.72 Standout trade, 4× short, $6.61 → $6.03
🔴 ETH short -$8.06 Wrong side of a bounce, closed quickly
🔴 LTC / SPX / NEAR -$0.71 Chop, cut before it mattered
Today, the long book is up +$31.83 with zero drawdown.
All three open positions were entered in the last few hours. Two are already green.
Why the losses are the point
Look at the win rates and you might wince.
The long book won 3 of 7 trades: 43%.
The short book won 1 of 5 trades: 20%.
On paper, ugly.
But that is the wrong number to fixate on.
Look at the sizing.
LONG BOOK
Average winner: ~$45
Average loser: ~$13
The winners were allowed to breathe.
The losers were cut before they mattered.SHORT BOOK
Largest winner: +$87.72
All losses combined: -$8.77
One asymmetric win paid for every mistake almost 10× over.One asymmetric win paid for every mistake almost 10× over.
That is the entire game.
The discipline is not being right more often.
It is cutting losers fast and letting winners run.
And that discipline is enforced by trailing stops that do not have feelings.
ARM went wrong and got cut at −$34 before it became −$100.
AVAX worked and was allowed to compound to +$88.
A human staring at a 20% short-book win rate would have switched it off on Tuesday.
The agent did not, because the math was working.
This is what “the agent gets smarter and keeps emotion out of it” looks like in a ledger instead of a tagline.
The honest caveats
Because we do not do promise outcomes, and you should not trust anyone who does:
This is one week. A 10% return on a small base over seven days is not something to annualize, and we are not going to pretend otherwise.
The size is small — under $2,000 across all three wallets. Deliberately. This is a dev run, not a fund.
Magpie made $0, and that is in here on purpose. A real strategy spends most of its time waiting. We are showing you the boring wallet too.
It is all onchain. The wallet addresses above are real. You can verify every trade yourself — which is exactly the point.
We would rather show you a verifiable +$175 with red trades included than an unverifiable hockey stick.
This is the thing you’ll be running
Two specialist strategies:
Lion
K-shaped dispersion capture
+
Magpie
New-listing event optionalityBuilt from a sentence.
Running autonomously.
Cutting losers.
Riding winners.
Watching while we sleep.
That is Senpi 2.0.
When we open, you describe your view in plain English and your agent does the rest.
Reserve one before launch and get $100 in free AI credits to put to work in your first 30 days.
Reserve your agent → waitlist.senpi.ai 🥷