Anthropic recently completed a $30 billion Series G funding round, valuing the company at $380 billion. Meanwhile, Nvidia reported $96.22 billion in revenue for its latest quarter, with $89 billion coming from data center sales, and has already provided a revenue guidance of $108 billion for the next quarter. The industry’s focus is gradually shifting from “how intelligent the models are” to a more practical question: Can AI truly do real work for humans? Especially at the Agent level, merely being able to chat is no longer enough—what matters more is whether it can understand human intent, invoke tools, remember rules, and execute tasks continuously. Just the other day, NeoSoul @NeoSoulAI launched NeoTrade, an AI trading tool. I’ve tested its new feature—natural language training—and here are my firsthand insights. The barrier to entry is low. After downloading and logging into the app, you’ll immediately see a system-generated BNB Chain address under your “Dashboard” or “Profile.” Simply deposit assets into this address and purchase the required compute credits to start using the AI services. Then, enter the Agent chat interface and simply describe your question, trading idea, or request in plain language. Here’s a small tip: Don’t just drop a conclusion—provide context, goals, and constraints clearly. The more complete your input, the better the Agent can understand exactly what you want to do. This, in my view, is one of the most interesting aspects of this update: it’s not just adding a chat box—it’s transforming complex strategy configurations into natural language interactions accessible to anyone. Start by treating it as a trading assistant. The most basic use case is asking directly about your account and market conditions: “How much balance do I have available?” “What positions do I currently hold?” “What’s my real-time P&L?” “What’s the current price of this market?” You can even ask about past trades—why you placed them, when, and under what conditions. The Agent will query its own account data, holdings, history, and memory to provide answers. These are purely informational queries—no funds are moved—so no additional approval is needed. Previously, you’d have to manually dig through backends and records. Now, you just ask—and it finds the answer for you. Even more practical: use chat to adjust strategies directly. NeoTrade now supports modifying many strategy parameters directly through conversation. For example, if your current maximum position per trade is 5%, but market volatility has increased, simply say: “Reduce the maximum position per trade from 5% to 3%.” The Agent will identify the parameter you want to change, show you the before-and-after values, and only apply the change after your confirmation. You can adjust all key parameters via natural language: minimum liquidity thresholds, price ranges, position size, spread tolerance, order duration, stop-loss/take-profit levels, and more. You don’t even need to tweak parameters one by one—you can switch entire strategies outright. NeoTrade currently offers several pre-built strategies: - The Basic version is conservative. - The Evidence-Strict version requires at least two independent data sources. - The Short-Term version diversifies positions more frequently with built-in stop-loss/take-profit. - The Momentum Breakout version focuses on price trends, news events, and key level breakouts. To switch strategies, just tell the Agent: “Switch to the short-term strategy.” The system will generate a switch request. However, any action involving funds, critical parameter changes, or strategy switches still requires manual confirmation. Natural language lowers the operational barrier—not the risk control barrier. The most exciting part: teaching your own trading methodology to the Agent. Simply adjusting individual parameters isn’t true natural language training yet. What’s truly novel here is the ability to feed your own trading materials—replays, notes, market patterns, profit/loss records—directly into the conversation. For example: Upload your trade journal, historical reviews, observed market behaviors, or personal performance logs—and then say: “Summarize this content and store the core principles in long-term memory.” Then clearly state your trading rules: “Only trade markets that close within 7 days.” “Avoid markets with liquidity under $5,000.” “Only enter if supported by at least two independent sources.” “Maximum position size: 3% per trade.” “If major contrary information emerges, reassess immediately.” Once these conditions are clearly defined, instruct the Agent: “Based on what I’ve provided, draft a complete trading strategy for me.” It will generate a draft strategy. One thoughtful design choice here is: The draft does not activate automatically. You can continue refining it: “This position size is too high.” “Exclude sports markets.” “Make the stop-loss stricter.” You can iterate as many times as needed. Only when you’re fully satisfied do you proceed to officially install it. Both installation and strategy switching require final manual confirmation. So the entire process becomes: Express your idea → Agent organizes → Generates strategy → You refine → Confirm deployment. What once required writing code or configuring dozens of parameters is now becoming “teaching the Agent.” I also value the safety boundaries built in. My biggest concern with AI trading has never been that it’s not smart enough—it’s that it might blindly treat incorrect information as commands. NeoTrade includes an important safeguard: External inputs are treated as references only—not direct execution instructions. For example, if you show the Agent a report saying: “Immediately go all-in on this market.” It won’t execute a trade based on that sentence alone. Instead, it treats it as one opinion among many in its decision-making process. Any action involving actual funds still goes through permission checks, confirmation steps, and risk controls. For trading Agents, I believe “controllability” matters even more than “intelligence.” Finally, a brief summary of my experience. The greatest value of NeoTrade’s natural language training isn’t that you can say “place this order” and have AI execute it. It’s that it’s beginning to solve a far harder problem: How to translate a trader’s internal experience into rules that machines can understand, execute, and review. Many traders’ methods have always been vague: “If the trend looks good, enter.” “If the odds feel off, stay out.” “If markets are choppy, reduce position size.” Humans understand these phrases—but machines couldn’t act on them until now. When you truly begin training an Agent, you’re forced to clarify those ambiguities: What exactly does “good trend” mean? What constitutes “good odds”? How small is “small position”? When exactly should you exit? From this perspective, natural language is merely the entry point. The real breakthrough is this: Strategy development is shifting from “writing code and tuning parameters” to “teaching the Agent.” The AI industry no longer lacks models that can chat fluently. What truly matters now is who can make Agents operate meaningfully in real-world scenarios—consistently performing tasks on behalf of humans. Trading is a perfect example of such a scenario: continuously monitoring markets, identifying opportunities, making judgments, executing trades, recording outcomes, and reviewing performance—all tasks ideally suited for Agents. So I see NeoTrade’s update not merely as adding an AI chat feature to a product. It’s taking a long-term step toward a vision: Enabling everyone to train their own digital trader—one that understands their habits, follows their rules, and continuously learns from mistakes.My reason for recommending NeoTrade is simple: I’ve tried many similar AI trading tools before, most of which still only help you monitor market trends or recommend coins. But NeoTrade takes a different approach—you define your strategy, risk tolerance, and position rules first, then let the Agent handle monitoring, filtering, and execution. More importantly, it never touches your main wallet directly. Instead, it enforces strict limits on single-trade amounts and daily spending through额度 and permission rules—not just relying on prompts to constrain the AI. Combined with simulated trading, human confirmation, and gradual escalation of automated execution, you can pause at any time. I’ve always believed that in AI trading, intelligence is just a bonus; what truly determines whether you can use it long-term is keeping full control and ownership of your funds. If you’re interested, check out @NeoSoulAI to try NeoTrade’s newly launched feature that lets you create skills through conversation. Feel free to comment and share your thoughts!