Archive / Use AI as an investment research team, not a trader

Use AI as an investment research team, not a trader

A prompt that has AI write a research memo on one company, with the case for, the case against and what would prove it wrong. It never tells you to buy.

Nothing here is financial advice or a recommendation to buy, sell or trade anything. Do your own research.

For: Solo founders, New to AI

Topics: Investment

Why I'm sharing it

The safest way I've found to use AI for investing is the boring one: research, not trading. It reads, it argues both sides, it writes a memo. It places nothing.

I found an open-source project called AI Berkshire that turns Claude Code into a value-investing research team. It's research only and places no trades, which is exactly why I like it as a starting point. I haven't run it myself yet, so this post is my own simple version of the idea, one prompt you can use in any chat.

What makes it work

It argues both sides. A bull case with no bear case is a sales pitch.

It writes down what would make you wrong. That's the part you'll want to reread later.

It separates facts from guesses. Every number has to say where it came from, or say that it's an estimate.

It ends with questions, not a verdict. The decision stays yours.

The prompt

Prompt

Act as a small investment research team writing a memo on one company: [company name and ticker]. Use only information you can point to, and label anything you estimate. Don't recommend buying or selling. 1. The business: what it sells, who pays, and how it makes money, in plain words. 2. What protects it from competitors, if anything, and how durable that is. 3. The bull case: the three strongest reasons it could do well. 4. The bear case: the three strongest reasons it could do badly, argued just as hard. 5. The numbers that matter most for this business, where each comes from, and what's missing. 6. What would prove the bull case wrong. Name signs I could watch for. 7. Five questions I should answer myself before forming a view. Keep it to one page. Mark every estimate as an estimate.

How to use it

Check the numbers against the company's own filings before you lean on them. AI can get figures wrong, and confidently.

Save each memo with the date. Reading your old "what would prove me wrong" list a few months later is where the learning happens.

If you want to go further, look at the AI Berkshire project on GitHub, read its own documentation, and try it on a company you already know well first.

This is research practice, not advice. It isn't financial advice, and nothing here is a recommendation to buy, sell or trade anything.

Inspired by AI Berkshire, an open-source project on GitHub by xbtlin that sets up value-investing research skills in Claude Code. The prompt here is my own, not taken from the project.

Use it for: Doing structured research on a company you're curious about, with the risks written down before any decision.

Back to the archive