About ten months ago, I started using AI to help me understand value investing.
Every morning I would read financial reports, then talk through what I found with Claude or Gemini. Over time these conversations accumulated into something substantial — hundreds of exchanges covering individual companies, industry analysis, portfolio decisions, and the mental models I was building.
At some point I decided to ask AI to read all of it. Not to find information I had missed. To find the pattern in how I was thinking — and whether that pattern was going to get me killed.
What I Asked For
I fed AI the full conversation archive: everything I had discussed over ten months. Not just my questions, but the way I asked them. Not just the conclusions, but the reasoning I used to get there.
The instruction was simple: find what I am doing well, and find what I am doing that will hurt me before my strategy has time to prove itself.
AI came back with a 40-page document it called the Upgraded Value Investing Handbook.
The opening sentence: “Your stock-picking methodology already surpasses 90% of retail investors. Your survival habits do not match it.”
The Good Part First
The positive assessment was genuine.
Ten months of reading annual reports, visiting stores to check products, comparing companies within the same industry, learning to separate temporary earnings drops from structural ones — this had built a real skill. AI confirmed that the analytical framework was sound.
For industries I knew from direct experience — manufacturing supply chains, export logistics, consumer goods distribution — my judgment was described as having “real sensory knowledge.” I had actually handled purchase orders, negotiated with suppliers, watched how Chinese factories respond to overseas demand. AI said this was a genuine edge, not borrowed knowledge.
That part felt good to read. Then came the rest.
The Flaw
The problem was not my ability to find good companies. The problem was what I did after I found them.
AI identified a pattern across my conversation archive: when I found a company I believed in strongly, I tended to go in at full position. Sometimes with leverage added. Decisions that should take days took three minutes. When a position moved against me, I found myself thinking seriously about averaging down.
AI laid out why these behaviors are specifically dangerous in combination with a long-term value investing strategy.
The argument was framed around a concept called ergodicity — the difference between the average outcome across many parallel universes and what actually happens to you in the single timeline you live.
In theory: across ten thousand investors following my strategy, the average outcome is positive. The math works.
In practice: I am not ten thousand investors. I am one investor. I have one timeline. A single forced liquidation — from leverage going wrong at the wrong moment — clears to zero everything that came before it. Not just the money. The compounding. The years.
AI put it plainly: you are using a strategy that only pays off with time, paired with habits that remove your ability to wait.
The Framework It Proposed
The document introduced a way of categorizing positions by actual knowledge rather than confidence.
The question AI posed for each holding: can you explain — without looking at your notes — what this business does to make money, why competitors can not take it, and how much it earns in its worst year? Then answer three follow-up questions from someone who knows nothing about the industry.
If you can do that: this is real knowledge. Full position allowed.
If you need to check your notes halfway through: this is borrowed knowledge — learned from AI or analyst reports, not from direct understanding. Half position maximum, with a pre-written exit condition.
If you can only describe a trend or a macro story: this is confidence, not knowledge. No position justified.
Working through my actual holdings honestly against this test produced uncomfortable results. Some companies I had described as high conviction turned out to be high familiarity — I had heard the thesis explained many times, which created the feeling of understanding without the substance.
The Survival Rules
The upgraded handbook ended with what it called survival axioms — not investment philosophy, but survival requirements that had to be met before the investment philosophy could have any chance to work.
No leverage. Not because leverage always loses, but because leverage converts a game that time can win into a game that one bad event can end.
No position-sizing by conviction. Size positions by verifiable knowledge, not by how certain you feel. Certainty is a feeling. Knowledge is something you can demonstrate.
Pre-write the reason to sell before you buy. Not a vague “if the story changes” — a specific, observable signal that forces a review.
None of these rules are original. Every experienced investor says some version of them. But seeing them derived from my own conversations, from my own patterns, made them land differently.
What Changed
I don’t trade the same way now.
The full-position instinct is still there. But I can see it for what it is. The three-minute decision still tempts me. But I have a written rule that says I am not allowed to act on a new idea for 48 hours — and AI helped me write it, because it identified the pattern in the first place.
Whether this makes me a better investor over time, I don’t know yet. Ten months of learning, ten months more to go.
But there is something valuable about asking a system that has no stake in your feelings to read your own thinking back to you without softening the findings.
My stock-picking was better than I thought. My risk management was worse than I admitted. Both were true at the same time.
I just needed something with no reason to tell me what I wanted to hear.
I am still learning. If you are using AI for investing — I am curious what you are using it for and what it has found.
Share your experience or thoughts below.
Related Reading
- AI Told Me to Buy. I Hesitated. It Tripled.
- 10 Months Using AI as a Value Investing Tutor — What I Actually Learned
- I Fell in Love with a Stock. My AI Noticed Before I Did.
Leave a Reply