I wanted to build an AI that trades stocks for me. I am building something better. My journey began as a biology student at Cornell University, where I took an introductory AI course that was widely considered “useless.” It wasn’t as mathematically rigorous as other classes, but the simplicity of its core ideas—Monte Carlo simulations, reinforcement learning, genetic optimization—felt like uncovering the secret mechanism behind a superpower. I was hooked. As a new trader who had lost more money than I’d made, I dreamed of a system that could make decisions for me. I devoured everything I could find—blogs, research papers, anything with the word “AI” and “trading” in it. Most of it was beyond my grasp, but the fire was lit. I became obsessed. I set out to build an AI trading bot. My first attempts were terrible. The models failed spectacularly. I even wrote a paper detailing how an AI trained on stock prices alone couldn’t learn to trade. If it weren’t for my professor’s personal review of the work and recognition of the effort, I would’ve gotten a B-. Instead, he gave me a B+—a rare gesture of encouragement in a notoriously tough course. But failure didn’t stop me. If anything, it fueled me. My vision was clear: build Stock-Jarvis—like Iron Man’s AI, but for the stock market. I started small. I built tools to backtest trading strategies. I tested how well ChatGPT could generate ideas. But I quickly realized that language models don’t think—they just predict the next word. They don’t form hypotheses, test them, or refine their approach based on data. So I built something different. Introducing Aurora—the AI Trading Agent. Aurora doesn’t just spit out ideas. She thinks. She reasons. She acts. When you ask her to “build the best rebalancing strategy between GLD and UPRO,” she doesn’t give a quick answer. She starts with your goal, then uses a structured reasoning loop—similar to the ReAct framework—to explore, hypothesize, test, and optimize. She searches financial data, applies economic principles, considers risk tolerance, generates multiple strategies, backtests them across bull and bear markets, and refines them using genetic algorithms—just like the ones I learned in my Cornell class. She doesn’t just suggest a strategy. She builds it from first principles, evaluates its robustness, and delivers a well-validated plan. And she does it all autonomously. The most powerful part? You can try it for free. Create a free account. Go to the agent page. Give her a goal—simple or complex—and watch her work. Why give it away? Because I believe in a feedback loop. The more people use Aurora, the smarter she becomes. Just like OpenAI improves ChatGPT through user interactions, I’m training Aurora with real-world usage data. I’m using advanced techniques like offline reinforcement learning and Decision Transformers—algorithms that can learn from historical data without constant real-time interaction. These are the same tools used by top quant firms. My personal Robinhood account is up $30,872 year-to-date—over 123%. That’s not typical. Past performance doesn’t guarantee future results. But other users are seeing similar gains. One member of our free Discord community said they were up 91% in six months—enough to cover their subscription for over five years. I’m not promising success. The market can change overnight. A crash could happen. People will point fingers. But I don’t care. Because I already won. I started not as a finance expert, not as a computer scientist, not with a degree or a Wall Street pedigree. I started with passion. With refusal to quit. With the lucky timing of ChatGPT’s release. I built an AI that trades better than I ever could. Now, I’m asking: can it trade better than the entire world? Give me five more years. I’ll find out.