Can artificial intelligence really help you beat the stock market? One student from Oklahoma decided to find out the hard way—by handing over €100 (well, $100) to ChatGPT and letting it call all the shots. What followed was a sharp learning curve, a tidy profit, and a surprising amount of buzz for someone still juggling schoolwork and summer holidays.
Letting AI take the wheel
Nathan Smith didn’t exactly have Wall Street ambitions. Like many teens, he stumbled across a flashy ad online claiming AI could outperform the market. Curious, and just savvy enough to know how to build a basic system, he decided to hand full control of a micro-cap stock portfolio to ChatGPT.
His instructions to the chatbot were crystal clear: pick only U.S. companies with market caps under $300 million, manage the entire portfolio—position sizing, stop-loss rules, and all—and aim to maximise gains from late June to late December. Nathan’s only role? Executing the trades and keeping records.
Four weeks in, the results were surprising. His GPT-powered portfolio showed a 23.8% return, outpacing both the Russell 2000 (3.9%) and the biotech ETF HBI (3.5%). Not bad for a system with no emotional bias and zero human intuition.
Building a bot from scratch
What set Nathan’s experiment apart wasn’t just the use of ChatGPT—it was how he structured the whole operation. He designed five core tools to support the project: manual trade confirmations, transaction tracking, real-time performance charts, daily result updates from Yahoo Finance, and a dashboard to keep everything in check.
The coding itself wasn’t rocket science, he claims. “The setup is fairly straightforward,” he said, referencing the use of Pandas dataframes to analyse price data and select new stocks each week. His only real challenge? Preventing ChatGPT from getting confused—a quirk known as “hallucination” in AI speak.
Interestingly, one of the most profitable moves came when ChatGPT abruptly sold a position in CADL, banking a tidy 50% gain. According to Nathan, the AI showed a level of discipline most human traders would struggle to match. “It just knew that gains in micro-caps can vanish in a blink. It didn't hesitate.”
The risks no one talks about
Before you start dreaming of turning your next pizza budget into a fortune, let’s get one thing straight: this isn’t a foolproof get-rich scheme. Even Nathan admits the timeframe is too short to mean much. Four weeks is barely a market blink, and with such volatile stocks, one or two lucky trades can skew the results entirely.
And then there’s the danger of mistaking short-term gains for long-term skill. Markets are notoriously cyclical. Without a deep understanding of trading psychology, risk management and data interpretation, using AI for investing could do more harm than good.
From summer project to serious passion
Nathan’s not banking on early success. He’s already looking ahead to extending the project to a full year and refining the model further. What began as a curiosity—sparked by online tutorials and a brief flirtation with Harvard’s CS50 course—has turned into a genuine passion for quantitative finance and Python programming.
But he’s also clear-eyed about the limits. “AI won’t make you rich overnight,” he says. “If you don’t understand what you’re doing, it can just as easily wipe you out.”
For now, though, Nathan’s story serves as a fascinating snapshot of where tech, curiosity and a bit of clever coding can take you. Just don’t forget—when it comes to the stock market, there’s no such thing as guaranteed success. Not even with a chatbot on your side.
