The pre-mortem I run with Claude before I buy anything

Most investors fall in love with a story before they’ve tried to break it.

You read the pitch, the numbers line up, the narrative feels clean

and you start sizing the position in your head before you’ve asked what could actualy go wrong.

That’s not conviction.

That’s a thesis that hasn’t been tested yeet.

Here’s the process I run before I put any real capital behind an idea .

There’s a specific moment in research where the risk shows up.

It’s not when you’re reading the bear case.

It’s earlier than that, when the idea still feels good and you haven’t had a reason to poke holes in it yet.

Most people skip that moment.

They go from “this looks interesting” straight to “how many shares.”

The thesis never gets stress tested until the stock is already moving against

them and it’s too late to think clearly.

I changed the order. Before I buy anything, I run a pre-mortem with Claude.

Not a bull case. Not a summary of the filing.

I ask Claude to find the version of the future where I lose money, written before I’ve committed to the trade.

Three questions I run through Claude, in order.

About twenty minutes

It changes what the position actually looks like once I’ve done it.

Not “what are the risks.”

Every 10-K has a risk factors section and none of it is specific enough to act on.

I want Claude to find the one thing that, if it happens, means the whole idea was wrong from the start.

This forces precision.

A vague answer like “competition increases” isn’t useful.

A specific answer like “the company’s largest customer renegotiates pricing at the next contract renewal” is something I can actually watch for.

Claude is good at this kind of forced specificity if you ask for it directly instead of accepting a generic list…

These get confused constantly.

A great business at a bad price and a bad business at a great price can look identical on a screener.

Both show up as “cheap relative to something.”

The distinction matters because the two problems need opposite solutions.

A price problem fixes itself if you’re patient.

A quality problem gets worse the longer you hold it.

I ask Claude to separate the two explicitly, using the business fundamentals rather than the valuation multiple, since the multiple is usually what triggered my interest in the first place and I don’t want it doing the thinking for me.

Claude working from the raw numbers instead of my own framing is the whole point here.

This is the one that keeps me honest!

If a stock looks obviously cheap, someone smart has probably already looked at it and passed.

The question isn’t whether I found something interesting.

It’s whether I found something the market hasn’t already priced in.

I ask Claude to argue the other side, specifically the case for why the current price might already be correct.

Not a devil’s advocate exercise for its own sake.

I want Claude’s genuine attempt to find out what I might be missing before I find out the expensive way.

After running these three prompts through Claude, the position looks different than it did an hour earlier.

Sometimes the thesis survives and I buy with a clearer sense of what I’m actually watching for.

Sometimes it doesn’t survive, and Claude just saved me from finding out the hard way, six months and a drawdown later.

The point of running a pre-mortem with Claude isn’t to talk yourself out of good ideas.

It’s to make sure the ideas that do get through have already survived the version of scrutiny the market will eventually apply anyway.

Better to find the crack in your own office than in a drawdown.

These Claude prompts work best with the actual filing data in front of Claude, not a summary you’ve already written yourselff

The less pre-digested the input, the more honest Claude’s output.