AI Trading Works (You are just doing it wrong)

The AI Trading Fallacy: From Hand-Trading to Systematic Incubation

This video features a trader who initially believed that learning to code—and later, using AI—would automatically lead to algorithmic trading wealth. He shares his personal journey from building a failed first bot to developing a structured framework, arguing that knowing how to code (or using AI to code) does not equate to knowing how to trade. The core message is a warning against the common misconception that AI can simply be told to "go make money."

The Core Misconception: AI as a Money-Making Oracle

The creator emphasizes that most people approach AI trading incorrectly by asking it for direct profits. He points to his own experiment at moonab.com/ai, where multiple AI models were given the same code and data to trade autonomously. The results showed that only one AI (OpenAI) was marginally profitable, proving that AI cannot reliably generate returns on its own. The key distinction is that AI should be used for coding and backtesting systems, not for making discretionary trading decisions.

The Framework: Research, Backtest, Incubate

Instead of seeking a "magic bullet" bot, the video advocates for a scientific, process-driven approach. The framework is designed to kill most ideas quickly because most trading strategies fail. The process involves:

• Research: Gathering ideas from books, podcasts (e.g., Chat with Traders), academic papers (Google Scholar), and YouTube. The goal is to build a large list of potential strategies.
• Backtest: Using AI to code backtests rapidly, testing multiple timeframes and symbols. A key tip is to look for "bubbles" where a strategy works across several close parameters, indicating robustness.
• Incubate: Running the most promising strategies live with small position sizes (e.g., $10) for a few weeks to see if they hold up in real market conditions.

The Business of Trading: Fees, Leverage, and Mindset

The video highlights often-overlooked costs that erode profits, particularly taker fees, which are significantly higher than maker fees. It also warns against the psychological trap of hand-trading, which is described as an "addiction" fueled by leverage and emotional decision-making. The creator argues that moving to automated systems removes stress and forces a business-like perspective, where the goal is to compound small edges over time rather than chase one massive win.

"Stop asking AI what to trade and ask it to build the system."

Key Takeaway

The path to algorithmic trading is not about finding a single profitable bot, but about building a systematic pipeline that continuously generates, tests, and incubates multiple strategies. AI is the ultimate accelerator for coding and backtesting, but the edge comes from a disciplined process of research and small-scale validation. Success is not guaranteed by technology alone; it requires treating trading as a business of probabilities, where one winning strategy out of ten is a great result, and the market—not opinions—provides the final verdict.