How I learn an industry in an afternoon (part 2 of 3)

Last week I made the case for starting with the sector instead of the stock, and mapping it until you can see the whole chain. That raises an obvious problem. What do you do when the sector is one you know nothing about, and every sentence is packed with terms that make your eyes glaze over?

For most of my investing life, the answer was slow. I'd buy a textbook, read forum threads, and piece it together over weeks. That changed in 2024. AI tools got good enough to act as a patient tutor who never sighs at a basic question, and the time it takes me to learn a new field dropped from a month to a single focused afternoon.

Here's the exact routine, using optics as the running example. Optics is one of the harder sectors to crack, which makes it a good test. If the method works there, it works almost anywhere.

The first wall is always the vocabulary. Optics people toss around terms like CPO, InP, and EML as if we were all born knowing them, and a single paragraph can hold five of them. So I don't power through. I paste the confusing sentence into an AI tool and ask for a plain version, the kind you'd hand a sharp tenth grader.

Take CPO, short for co-packaged optics. In plain English it becomes this: instead of putting the part that turns data into light off at the edge of the board, you move that light engine right next to the main chip, so the signal travels a much shorter distance. Same reason moving your router into the same room speeds up your wifi.

I do this for every term until the fog lifts. InP turns out to be indium phosphide, a special material lasers are built on, like the particular soil a certain crop needs. EML is a type of tiny laser that sends data down a fiber. None of these are as scary as they look. Once the words stop blocking you, the ideas underneath are usually simple.

Now that I can read the sentences, I want the shape of the whole sector. So I give the AI a direct task. Lay out the optics supply chain from raw material to the finished cable that plugs into a server, and name the public companies at each step.

What comes back is a rough draft. For optics it usually runs like this: the special material, then the lasers built on it, then the little modules called transceivers that package those lasers up, then the chips that clean and manage the signal, then the cables and connectors that carry it between machines. Each layer has its own set of companies.

Then I treat that first map as a sparring partner, not an answer key. I push back. Why does this company sit in that layer? Who competes with it? What did you leave out? The back-and-forth is where the learning actually sticks, far more than reading ever did for me. After a few rounds, a messy field feels organized, and I can see which layers are crowded and which have only one or two names holding them up.

AI has one dangerous habit. It sounds exactly as confident when it's wrong as when it's right, and it will hand you a tidy answer that happens to be made up. So I never stop at what the model tells me.

I match its claims against the real record: company filings, earnings call transcripts, product pages, and what I hear from engineers here in the Valley. When the two line up, my confidence grows. When they disagree, that gap is a gift. Either I just caught the model inventing something, which saves me from a costly mistake, or the model is behind and I've found a piece of the story the crowd hasn't caught up to yet.

More than once, an AI tool has told me with total confidence that a company makes a certain part, and a five-minute look at that company's own filings showed it doesn't, or does so only through a partner. Small distinction, big consequence. Buying a company for a business it doesn't actually own is exactly the mistake this step catches.

A quick search hands you the number one player and stops there, the name everyone already owns and talks about. The more interesting work sits one and two steps back from the front.

So I ask the AI to walk me down the chain. Who supplies the supplier? Who makes the material the lasers are grown on? Who builds the tool that makes the part? In optics, that pulls up the companies behind the lasers, the makers of that indium phosphide material, and the quieter names that sell into several big players at once. Some turn out to be tiny and fragile. Others are quiet toll booths that get paid no matter which brand wins at the front. Either way, you only find them by asking a question basic search never asks.

I'll name two you've likely seen in the optics conversation, Lumentum ($LITE) and Coherent ($COHR), as examples of the laser layer rather than the finished-cable layer most people picture first. Placing a company on the right layer of the map changes the whole story you tell about it.

Here's the step that makes it all compound. Every term I decode and every fact I confirm goes into one running note, in my own words. My optics file holds the jargon, the map, the layers, and the open questions I still want answered. I also jot down the analogy that finally made each term click, because months later that's what I remember, not the textbook definition.

The next time optics shows up in the news, I'm not starting from zero. I open the note, refresh in five minutes, and add whatever's new. Do this across a few sectors for a year and you end up with something valuable: your own private encyclopedia of the market you care about, in language you actually understand. That growing stack is the real asset, more than any single stock in it.

You don't need anything fancy. Plain requests work best. Four I lean on, roughly in order.

"Explain this paragraph like I'm a smart tenth grader, and define every technical term." Clears the vocabulary wall.

"Map the optics supply chain from raw material to finished product, and list the public companies at each step." Gives me the first draft to argue with.

"What are you least sure about in that answer, and what would I need to check myself?" Turns the tool into a partner that flags its own weak spots.

"Who supplies the companies you just named, and who competes with them?" Walks me down the chain toward the quieter names.

Copy those, swap in whatever sector you're studying, and you've got most of an afternoon's research mapped out.

None of this turns me into an engineer, and it doesn't need to. What the routine gives me is speed and a working map. A sector that used to cost a month of evenings now takes a focused afternoon, and I come out able to ask sharper questions and notice when a story doesn't hold together.

One warning, because it matters. AI is a tutor that likes to think they’re brilliant. It's thorough at explaining and organizing, and unreliable on specific facts, especially recent ones. Use it to understand, then verify before you act. The people who get burned blur those two jobs, treating a confident explanation as a confirmed fact and skipping to the buy button. Keep them separate and the tool becomes a real advantage instead of a hidden trap.

Next Thursday I'll close the series with the part everyone actually asks about. Once you understand a sector and can map it, how do you judge a single company, and how do you decide when to buy or sell? That's where the process turns into decisions.

One note, since this issue leaned on AI. The AI is just how I move faster. The method underneath, building a high-conviction portfolio and doing real research, is what my book Make Your Own Alpha covers start to finish, and it works with or without any of these tools.

Get the book by clicking here.

Also, I’m hosting a community meetup in London, UK next Tuesday, July 28th at 6-8 pm. I’m taking headcount, so I can figure out which location will serve us best. It will likely be somewhere in Central London. If you’re based in London, I would love to meet you in person!

Save your spot for London meetup with this link.

Stay disciplined, Koh

Disclaimer: Nothing in this newsletter constitutes investment advice or a recommendation to buy or sell any security. Numbers and observations are as of publication. I may hold positions in companies discussed above. Always do your own research and consult a licensed financial advisor before making investment decisions.