His BTC decision-making system is still in development, offering a concrete look at the work that comes before fully automated trading.

Xiaolong wants the AI system he is building to trade Bitcoin around the clock one day. During XT’s Chinese-language AMA, “The Next Step in AI Trading: Exploring Agent Applications Through NeoSoul (DNA),” the Bitcoin market analyst discussed AI’s advantages in execution, then returned to the trading logic behind its decisions:

“But real judgment still comes from the trading rules and strategies designed by people.”

Quote translated from Chinese.

His own system has not reached full automation because he does not consider its decision-making mature or stable enough yet. He is continuing to refine it and is willing to use limited live trading tests to find problems. His experience raises practical questions: what AI bases its decisions on, how it interprets a trading strategy, and what needs to change before it can take on more responsibility.

Automating orders doesn’t solve every problem

Xiaolong sees automated trading and risk monitoring as parts of execution. In his approach, whether a tool can place an order promptly is a separate question from whether the strategy behind that order is ready.

He sees potential in AI’s speed and ability to keep running. When discussing how well a system might ultimately perform, though, he kept returning to the trading strategy. Tools with the same ability to place orders automatically can produce different results if they use different strategies. Faster execution cannot make up for flawed judgment.

That is where he is focusing his own development work. He described incorporating trading methods, on-chain data, and market structure analysis into the AI’s decision-making process, hoping it would gradually learn his approach to analysis. How the system uses that information and turns a judgment into a trade still needs work.

Breaking round-the-clock trading into specific jobs

To work toward round-the-clock trading, Xiaolong proposed dividing responsibilities among different agents. Some would gather market news and on-chain data. Others would use that information to form judgments, with additional tools supporting backtesting and trade reviews.

That division gives the development process specific tasks. Collecting more information only helps if it contributes to a trading decision. Each step between analysis and execution has its own problems to solve. Xiaolong described these as areas he wanted to explore; he did not say his system already had all of those capabilities.

One thing he is already doing is using AI to help analyze the market and anticipate moves, then checking those judgments against actual trades. He supplies the information available at the time, his trading logic, and details of how the trade unfolded so the system can analyze what it got right and where it went wrong. He described this repeated adjustment as a way to bring the tool closer to his own trading approach.

He is hopeful about the process. Still, whether a system’s analysis becomes more like its user’s and whether its trading results become more consistent are questions that need separate testing. An error can suggest what to change. Whether that change helps has to be assessed through subsequent performance.

Learning through limited tests

Xiaolong is open to letting AI buy and sell on its own. Within the small amounts he has set aside for testing, he is willing to give the system considerable freedom. He worries that intervening too often would constrain the tests and could cause the system to miss trading opportunities.

That willingness is consistent with his acknowledgment that the system is unfinished. He wants limited experiments to reveal the problems that arise in actual use. For larger amounts, he still prefers to retain control while the system remains unstable. Any decision to expand its use would depend on his further assessment of it.

He was discussing his own development and testing experience. Xiaolong explicitly said that he had built a decision-making system himself and had not used the commercial agent products being discussed. His comments should therefore not be read as a review of NeoSoul.

His eventual goal is clear: AI that can trade on his behalf around the clock. Based on the progress he described in the AMA, his immediate task is still to improve the decision-making system and make its analysis work more closely with his trading methods. Further automation depends on how that work progresses.

---

This article is based on Xiaolong’s remarks during XT’s Chinese-language AMA, “The Next Step in AI Trading: Exploring Agent Applications Through NeoSoul (DNA).” It presents his personal views and development experience and does not constitute investment advice.

Listen to the full replay.

About XT Exchange

Founded in 2018, XT Exchange is a global digital asset trading platform serving more than 12 million registered users across over 200 countries and regions. Its services span spot, margin and futures trading, TradFi market exposure, a real-world asset marketplace, and everyday payments through XT Pay. Guided by Xplore Crypto Trade with Trust, XT Exchange enters its ninth year with Build the NeXT, reflecting its focus on trust, broader market access, and practical uses for digital assets.

Join the XT Exchange Community: X (Twitter) | Telegram | Facebook | Instagram | LinkedIn | Medium | YouTube

Disclaimer: XT Exchange reserves the right, at its sole discretion, to modify, amend, or cancel this announcement at any time for any reason without prior notice.