TL;DR: Today we’re releasing a new episode of our podcast How Do You Use ChatGPT? I go in depth with Jesse Beyroutey, managing partner at $600 million venture capital fund IA Ventures, on using Gemini Pro 1.5 and ChatGPT to find great investments. As we talk, we invest $1,000 in a stock we choose with AI. Watch on X or YouTube, or listen on Spotify or Apple Podcasts.

I made the greatest trade of my life with Jesse Beyroutey in 2019. We bought Nvidia shares when they were trading at $33, and they’re worth nearly $800 today.

In this episode, Jesse and I try to top that by making a real $1,000 investment live on the show—with Gemini Pro 1.5’s incredible 1 million token context window.

Jesse Beyroutey is a managing partner at IA Ventures, a $600 million seed and early-stage venture fund with investments in public companies like Wise and DigitalOcean. He’s also a very close friend and one of the smartest people I know. 

While making the Nvidia trade, Jesse and I believed that the company’s stock price was temporarily down because of the U.S.-China trade war, and from a long-term perspective, Nvidia was in a strong position because of its proprietary GPU architecture. We also thought that Nvidia’s market was likely to grow as machine learning and gaming increased the need for high-powered CPUs and computers.

In this interview, Jesse and I give ourselves 90 minutes to try to make an even better investment using the power of Gemini Pro 1.5 and ChatGPT. Our investment thesis is to find a company with strong potential that is trading down for temporary reasons.

To get us started, Jesse uses stock analytics platform TradingView and applies filters to surface underperforming public companies that have been relatively stable in the past few years, with a market cap of $2–100 billion and a positive gross profit. I opened an account on investment platform Robinhood and put in $1,000 as our principal. 

This is not investment advice, but it’s a must-watch for anyone who wants to leverage the power of AI to make smarter financial decisions. Here’s a taste:

• Setting the stage for Gemini. In our first prompt to Gemini, we briefly explain our investment thesis and copy-paste the data we found on TradingView, querying if anything about those stocks “pop[s] out” to Gemini.
• Be open to restarting from scratch. Gemini highlights three stocks from the data and recommends companies with “high growth potential,” but Jesse notes that we’re specifically looking for companies that are undervalued.” He recommends taking a “different angle” by asking Gemini “how to construct a [stock] screener to identify companies that might be artificially depressed in their prices.” 
• Get AI to make a decision for you. Gemini suggests a bunch of financial metrics and technical indicators as potential criteria for a stock screener. Jesse’s “immediate instinct” every time he “get[s] some kind of summary from ChatGPT or Gemini” is to “ask it to just make a decision.” 
• Leverage LLMs to get nuanced answers to your questions. When Gemini says it can only offer general investment advice, we prompt it to simulate Warren Buffett and help us choose a trading strategy. Jesse finds instructing AI to simulate experts valuable, explaining that it “elicits a different kind of perspective from the LLM than just asking it a question flat-out because it's trying to take advantage of whatever knowledge that it had about that person and things that co-occur with that person's name on the internet.”
• Find patterns in your data with AI. Based on Gemini’s response, we look for companies with low price-to-earnings ratios on TradingView, targeting those that are lately underperforming but have demonstrated relative stability in the past few years. We get 29 results from TradingView, and Jesse suggests prompting Gemini to identify “points of commonality” between them since “large language models are amazing at clustering things.”
• Using Gemini to decode consumer behavior. After Gemini organizes the results, we notice a few companies with poor reputations. Jesse thinks it might be an “interesting opportunity” to find unpopular companies that also have a “crazy distribution lock-in,” so we prompt Gemini to tell us which of these companies are likely to be trading poorly because their brand is disliked.
• Sift through large amounts of data. Out of the companies that Gemini surfaces, we choose to focus on in-flight internet provider Gogo. Leveraging Gemini’s large context window, we upload GoGo’s earnings transcripts from 2022 and 2023, using about 100,000 tokens (exceeding the capacity of the publicly available version of ChatGPT), and ask Gemini whether the management is talking about upgrading its technology.
• Understanding the stock market with Gemini. Gemini reports back that based on Gogo’s earnings transcripts, the company is actively discussing a technology upgrade. Jesse thinks this is a “pretty positive” outlook, and suggests getting Gemini to reason for us by asking, “What extent would that upcoming change already be priced into their stock and how would we figure that out?”

As Gemini generates a response, Jesse reflects on our original investment thesis of “finding things that were down for a clear reason.” He thinks we should search for stocks that are trading down because of “exogenous factors”—like inflation or industry trends—instead of filtering underperforming companies with a stock screener (which is what we did initially with TradingView). We think ChatGPT would be the best tool to answer this question, and it tells us that energy stocks are artificially depressed due to external reasons.

• Pitting Gemini and ChatGPT against each other. The next part of our investment thesis is finding companies that will benefit from “technology that they've built being adopted rapidly.” We ask ChatGPT and Gemini the same question, and find Gemini’s response to be better organized, even pre-empting our next query by including names of public stocks as examples.
• Consistently refine your search queries. The last piece of our trading strategy is a “highly scalable business model,” and we ask Gemini which publicly traded energy companies have this. Jesse chooses not to define “highly scalable” because he “find[s] that using LLMs is a task of continuously refining your question…and then you find some nuance and you kind of keep going and keep going.”

Even as we fine-tune our search queries, we realize that we aren’t getting very good results from Gemini or ChatGPT. So with just 15 minutes left on the clock, on Jesse’s recommendation, we switch tracks and choose a special purpose acquisition company (SPAC) to invest in. 

• New opportunities enabled by LLMs. While the size of Gemini’s context window is incredible, Jesse thinks that large amounts of content aren’t easily findable on the internet because a lot of people are averse to having their context indexed by LLMs. He thinks solving this problem and leveraging “anything that has a real time data source” to keep information updated are new avenues unlocked by large language models.

There’s a plot twist at the end of this episode, so stick around to see the epilogue Jesse and I recorded just days after we made our investment.

You can check out the episode on X, Spotify, Apple Podcasts, or YouTube. Links and timestamps are below:

• Watch on X
• Watch on YouTube
• Listen on Spotify (make sure to follow to help us rank!)
• Listen on Apple Podcasts

Timestamps:

1. Introduction 01:29
2. How Dan made the greatest trade of his life 03:50
3. Jesse’s strategy to use LLMs to get nuanced answers 05:27
4. Gearing up to orchestrate the best trade of our lives with Gemini Pro 1.5 09:20
5. How Jesse gets AI to make great decisions 17:52
6. Using Gemini Pro 1.5 to find patterns in data 22:38
7. How AI can provide deeper insights into the stock market 26:48
8. Leveraging Gemini Pro 1.5’s huge context window to analyze data 34:41 
9. Gemini Pro 1.5 and ChatGPT go head-to-head 46:33 
10. Choosing a stock with just 15 minutes left on the clock 1:10:11
11. What Jesse thinks are the biggest new opportunities enabled by LLMs 1:24:01
12. The epilogue Jesse and Dan recorded one week after making the trade 1:28:43

What do you use ChatGPT for? Have you found any interesting or surprising use cases? We want to hear from you—and we might even interview you. Reply here to talk to me!

Miss an episode? Catch up on my recent conversations with filmmaker Dave Clark, founder, author, and neuroscientist Anne-Laure Le Cunff, a16z podcast host Steph Smith, OpenAI developer advocate Logan Kilpatrick, clinical psychologist Dr. Gena Gorlin, economist Tyler Cowen, writer and entrepreneur David Perell, software researcher Geoffrey Lit, Waymark founder Nathan Labenz, Notion engineer Linus Lee, writer Nat Eliason, and Gumroad CEO Sahil Lavingia, and learn how they use ChatGPT.

If you’re enjoying my work, here are a few things I recommend:

• Subscribe to Every
• Follow me on X
• Subscribe to Every’s YouTube channel
• Check out our new course, Maximize Your Mind With ChatGPT

The episode transcript is for paying subscribers.

Dan Shipper is the cofounder and CEO of Every, where he writes the Chain of Thought column and hosts the podcast How Do You Use ChatGPT? You can follow him on X at @danshipper and on LinkedIn, and Every on X at @every and on LinkedIn.