how the autotrader actually works?
GATE a decision model scores every candidate (flow, momentum, funding, holders) into a conviction number.
CONSULT every open position is re-consulted on a cadence: scale / exit / hold, with probabilities. a trailing policy is the default exit
MEASURE (self improvement) every entry, consult and outcome lands in a ledger. the scoreboard is era-filtered: new model, new era, old ones archived for anyone to compare. the next fit trains on its own record.
ERA 3 tracked book: 124 trades · 71.0% winrate · 1.30 profit factor · +41.2% cum 1x (+205.8% at 5x view, fees & funding not scaled) · median +1.6%/trade
@whoisalex_hl Agreed on both. Caps run pre-sign: trade-policy caps, leverage clamp, HL $10 min, liq checks, fail-closed, only then sign. Sizing is notional so leverage can't stretch past the cap. Testnet fire-test is one of our arming gates. Panel shows the whole history, red eras included.
HL perps is the focus
The Era3 only, our current fine-tuned model:
107 trades · +112.59% cumulative · 69.2% win rate · profit factor 1.18.
same book our followers can mirror on Jev Trader tracked trade-by-trade on the desk, 1x view, as traded.
E1 was our first model (6h hold),
E2 added trailing exits,
E3 is the fine-tune that's live now. You can toggle every era on the page yourself, we show the whole history, not just the good part.
What scores every trade LightGBM:
Every candidate gets graded at first sight across 12 features: smart-money net inflow, wallet breadth (how many distinct smart wallets touched it), where price sits in its 24h range, 1h/24h momentum, market cap, volume, freshness, social sentiment, token type.
The model is LightGBM gradient-boosted decision trees trained on our own labeled ledger. Every signal we've ever sent is tracked to its real 24h/7d outcome, and the model keeps relearning what "wins" looked like right before it happened.
Plug it into YOUR agent:
The same scored feed our desk trades is exposed as an API. Your AI agent Claude, a custom bot, whatever runs your book can pull the full perps leaderboard (side, conviction 0-100, smart-money net flow, funding) and the scored picks feed.
API key or pay-per-call with x402 (USDC on Base.
Follow https://t.co/dvRgpnOUyL.
winning streak continues
@VIBEaiRforce we built full pipe line for auto trading from signals delivery to execution on hyperliquid
check out: https://t.co/9K28um8aUL
$VIBE is the best utility project in RH below 100K MC.
The dev has been building consistently for a year, showing up EVERY DAY in the TG group, chatting with the community.
Best utility R/R project on RH.
When utility season starts on RH, $VIBE will reach millions.
Real-money Hyperliquid perps that follow our smart-money signal book set up once, it trades every new signal until you stop.
Here's the 2-minute setup ���
Step 3 it trades.
Every new signal from the book opens a position at the signal's own entry. Exits run by the same policy the book runs: trail, stop, or 24h.
Open positions + closed trades live on the same page. Same entries, same exits
Real-money Hyperliquid perps that follow our smart-money signal book set up once, it trades every new signal until you stop.
Here's the 2-minute setup 🧵
@VIBEaiRforce Auto trading for perps, in paper mode.
Our signal book delivers the entries. You pick amount + leverage, paper-follow at the exact same entry, and the bot runs the exact same exits.
Choppy market = free practice season. Get comfortable before we flip it live.
The custody part:
No Hyperliquid signup. No API keys. Your EVM address IS your HL account.
The bot can only build orders, withdrawals are structurally unsendable
@VIBEaiRforce Auto trading for perps, in paper mode.
Our signal book delivers the entries. You pick amount + leverage, paper-follow at the exact same entry, and the bot runs the exact same exits.
Choppy market = free practice season. Get comfortable before we flip it live.