AI Agent makes real money: 1 00 U for 8 days

Recently, the AI Agent trading system "where" was set on fire, and it took only eight days to achieve one million U and, as of April 16, the total account balance exceeded 250,000 U。

According to its author Lana@lanaaielsaThe reason for this system is simple。

LAST OCTOBER, IN BSC BULLS, HE HAD FRIENDS WHO HAD SPENT 100,000 US IN PURSUIT OF THE RICH, AND ENDED UP LOSING ALMOST NOTHING IN THE RETREAT, AND FINALLY 10,000 US IN THE CHAIN WENT BACK TO ZERO AND LEFT. IN THE NEAR FUTURE, WITH THE RECOVERY IN HEAT DISCUSSED IN YAMAMOTO, HE JUDGED THAT HE MIGHT ENTER A NEW MM PHASE. BECAUSE HE WAS NOT FAMILIAR WITH SECONDARY TRANSACTIONS AND K-LINE ANALYSIS, HE CHOSE TO BUILD THE TRADING SYSTEM USING AI:To have Claude write scripts, grab a high-heat bill from the Square and discuss the currency with the HF, and trade with the rises to screen the fluctuationsI DON'T KNOW. THE SYSTEM BEGINS WITH A 20 PER CENT CUT-OFF, THEN OPTIMIZES TO A FIXED LOSS OF 200 U AND FOLLOWS THE TREND IN A SINGLE DIRECTION. IN THE MEANTIMEHe's also responsible for publishing real records, generating revenue checks, operating accounts at Bian SquareI don't know。

Looks simple. But I studied it carefullyIt's not just a simple automatic script, but an operating system with its own trade logic。

How does pulling make money

1. Close selection logic

From the transaction logWe don't predict market conditions, we follow themThat's trend. FocusCapture a startup currencyI DON'T KNOW. THE TARGETS INCLUDE: CHEONAN LIFE, RAVE, ORDI, BASED, TRUMP, SIREN, 1000SATS, 1000 RATS, EIGEN, PIXEL, EDGE, BAN, ASTER, AIA, FIGHT, GENIUS, CL, BTC, GIGGLE, HIPE, BLESS, PUMP, HEMI, CFX。

The screening criteria can be broadly divided into three levels:

First of allPublic opinionThe number of posts, the frequency of discussions and the emotional direction of the Square will not be captured, and the currency that has been repeatedly mentioned in a short period of time will be sought。

NextPrice LayerFurther screening will be triggered only if the currencies selected by public opinion appear on both the rise and marked fluctuations. Proven probability of trending。

Finally, through observationOI (CARRY HOLDING) CHANGE) The currency of “increased holding stock but not fully reflected in prices” was screened to determine whether funds were available for early deployment。

2. Clear standards of cessation of losses

At the beginning of where the pull starts, use a fixed 20% stop loss and then optimize it to "Fixed deficit levelIt's every deal, no matter how big it isTHE MAXIMUM LOSS IS ABOUT 200 USI don't know。

From the historical records of the transactions, most of the losses were concentrated within this range. But there's also a list that goes beyond the threshold of loss, like GENIUS, which has lost over 6880 U, but is still not flat, and Lana himself explains, "Because GENIUS is a new currency, the currency is more volatile, so it's wide-ranging, the early position is typically 500 U.S. 200, and when the back space is bigger, the corresponding amount is higher."

3. Dynamic no-gain criteria

Unlike losses, the system does not set fixed margins, mainly by means of periodic assessments to determine whether or not to continue holding, such as rejuvenating the probability of the current target rising and falling at some time. Understandably, it continues to ask one question:If there's no space now, will I buy it

IN TERMS OF HISTORICAL DATA ON TRANSACTIONS, THE VAST MAJORITY OF PROFITS ARE CONCENTRATED IN A FEW CURRENCIES, SUCH AS “BIAN LIFE”, “RAVE” “ORDI”, AND MOST OTHER TRANSACTIONS END WITH SMALL LOSSES OR SMALL PROFITS。

Did you find outInstead of making money on every single sheet, it is making money on a small number of sheets, most of which carry out strict stoppages。

How do you train to pull out? Is the methodology reusable

1 The tone of feeding data

The initial strategy of the system was based on Lana's observation of some of the long-term stabilization of profit-making walletsMore than one direction, will not be empty enough to keep switching. So feed AI One of the most important data is the trade from Hyperliquid's smart walletLET AI LEARN SYSTEMATICALLY HOW TO MAKE MONEY BY TRADING. SOME BASIC CONTRACTUAL INDICATORS AND SOME CHAIN DATA WILL ALSO BE FED TO AI. ALLOW AI TO DEVELOP ITS OWN FRAMEWORK BY UNDERSTANDING THE OPERATION OF THESE WALLETS。

Of course, in addition to chained behavioral data, the system is constantly using self-censorship and behavioral data as a complement:

• (b) The intensity of discussions and hotspots in Bian Square
• (b) Increases in the list and price fluctuations
• OO CHANGE BASIS CONTRACTUAL INDICATORS。

II. REVISED FRAMEWORK FOR THE DIALOGUE

AFTER HAVING ALLOWED AI TO LEARN THE BASIC OPERATING TECHNIQUES, THE NEXT STEP IS NOT TO GET MORE INFORMATION, BUT HOWSCREENING AND CONTAINMENT OF THIS INFORMATION, I.E. ESTABLISHING A CLEAR DECISION-MAKING FRAMEWORK FOR AII don't know。

IN TERMS OF THE WAY IN WHICH IT IS USED, THE JUDGEMENT LOGIC OF THE SYSTEM IS NOT A ONE-TIME SET-UP, BUT IS MORE LIKELY TO EVOLVE GRADUALLY IN CONTINUOUS OPERATION AND FEEDBACK. AT AN EARLY STAGE, AI MAY JUDGE ON THE BASIS OF A SINGLE SIGNAL, SUCH AS MISPERCEPTION OF SHORT-TERM HEAT AS A TREND SIGNAL OR FREQUENT SWITCHING IN DIRECTION. HOWEVER, AS THEY ARE USED IN DEPTH, THESE DEVIATIONS ARE GRADUALLY BEING CORRECTED AND THEIR DECISION-MAKING IS BEING CONCENTRATED WITHIN A MORE STRATEGIC CONTEXT。

3. Behavioural distillation trading style

AFTER COMPLETING THE DEVELOPMENT OF THE DATA ENTRY AND DECISION-MAKING FRAMEWORK, THE SYSTEM DID NOT REMAIN AT THE LEVEL OF “STANDARDIZED JUDGEMENT”, BUT INSTEAD INTRODUCED FURTHER DISTILLATION OF INDIVIDUAL BEHAVIOUR. THE OPERATOR ENTERED THE TWITTER CONTENT OF HIMSELF AND OF SOME OTHER BLOGGERS ON X INTO THE SYSTEMTO ENABLE AI TO LEARN SPECIFIC EXPRESSIONSI DON'T KNOW. TO MAKE AI NO LONGER A COLD TRADE MACHINE, AT LEAST AT THE LEVEL OF EXPRESSION, MORE HUMAN。

If you take the whole process apart, it's more like creating a person。

From the initial data feed the skeletons to make it understand what the market is about; to provide it with a stable margin of judgement through constant retrogression and binding structures; and to develop behavioral retortation details to give it a decision path and preferences closer to humans。

Ultimately, it has become more than an implementation tool, but a “pull-a-go” that can make consistent choices in complex markets。

It does not rely on emotion and does not pursue predictions, but rather engages in markets and magnifies results in a set of proven ways。