Can You Trust AI Despite Hallucinations?

Brian urges skepticism but shares rules and guardrails that make AI outputs reliable for him.

Play episode from 07:21

chevron_right

Transcript

chevron_right

Transcript

Episode notes

#663: We’re living through the first era in which an investor can ask a machine to read a decade of SEC filings in seconds. That sounds powerful, but also a little terrifying. Can we trust it? And how do we use it without falling for hallucinations or built-in optimism?

In this episode, we dig into the practical, real-world ways AI can strengthen our investing process while avoiding its biggest pitfalls. If you’ve ever wondered how to blend old-school fundamentals with new-school tools, this conversation will open up an entirely new mental model.

Our guest is Brian Feroldi, an investor who has spent more than twenty years doing classic, deep-dive fundamental research. He reads SEC filings for fun, and he’s embraced AI not as a stock picker, but as a force multiplier that can turn days of research into minutes.

We talk about the specific guardrails that make AI useful for fundamental investors, including restricting sources to trusted filings, designing step-by-step instructions, and assigning the AI a role so it knows how to “think.” We also explore how to stress-test optimism bias, how to analyze companies like a forensic accountant or a short seller, and how to build prompts that match your own investing personality.

Whether you’re an index-fund loyalist with a little “fun money” or a hands-on analyst, this conversation will expand the way you evaluate businesses and make decisions.

Key Takeaways

How a single prompt can transform AI from a loose generalist into a sharp, reliable research assistant. 
The surprising way optimism bias shows up in AI tools, and how to flip it to your advantage. 
Why limiting your data sources can make your analysis dramatically stronger. 
The role-play trick that helps you see a company the way a short seller, value investor, or even Warren Buffett might. 
A simple reframing that turns AI from a stock picker into something far more powerful for decision-making. 
The moment in the demo that revealed a blind spot even seasoned investors often miss.

Resources and Links