The Research Diary
Daily unfiltered notes from a buy-side analyst's journal - covering tech, semis, and time management
What I Worked On/Thought About This Week
Each day, I spend 5–30 minutes writing a stream-of-consciousness “brain dump” about whatever I worked on or thought about. I do a quick edit (or sometimes none at all) and post those daily reflections to X (@stockthoughts81). At the end of each week, I compile them here in one place.
The goal is to clarify my own thinking, keep a record of my day-to-day process, and hopefully spark conversations with others tackling similar problems. These entries are closer to raw journaling than polished essays—maybe I’ll flesh them out more at some point, but for now, I keep it simple.
If you enjoy this kind of journal-like content and want to keep following along, feel free to sign up below. And please don't hesitate to reach out with thoughts on any of the companies or topics discussed.
1/27/27
Today was not a productive day. Prospect meeting tomorrow so did a little bit of prep time for that, mostly going over PRKS and CPAY.
The July 2024 VIC writeup on CPAY is really good. It very clearly identifies what is important (the increasing importance of the Corp Pmts segment) and spends basically all the time on that. They did a really good job of breaking down the different subsegments, competitors in each, and CPAY’s positioning.
PRKS is just cheap.
I made a couple mistakes on the productivity front. Once you enter the hive mind of the Internet it is hard to pull away. First I went into Teams to get some data I needed to update the PRKS packet. I should have put a pin in it to do later. Because once I was on Teams there were a couple of other things to address and obviously I think “let me get these off my plate”. Then later on went on Twitter to RT my own tweet about the blog post that collects last week’s daily updates. That maybe could’ve waited until EOD. IDK. Or scheduled it to go out in morning, and then retweet later in day when I do admin stuff. And then anyone on here knows how Twitter was today. It was hard to stop reading/thinking about DeepSeek and that was what really threw things off. I don’t really have anything valuable to add about DeepSeek. I am still thinking it through. There is a ton to read. I could waste the next 30 days reading DeepSeek stuff. IDK if that’s a good use of time. It could impact several of our holdings but I think I need to sit on this for a few days. See which analyses rise to the top. It may be that the key questions are unknowable and so that makes it not worth spending time on.
How about ORCL today, huh? Down 14%. Safra Catz sold $500m after exercising options. Maybe buying TikTok. And who knows what DeepSeek implications are for Stargate.
And then some random email/admin stuff.
I don’t quite understand why the WFE players like ASML MKSI AMAT and LRCX are down so much. If you had asked me before how exposed they are to AI, I probably would’ve said something like, “less than they want you to think they are… it’s a small but increasing portion of revenue that will drive some growth in years to come.” Apparently in 2025 ~25% of DRAM output expected to be HBM. But even still - there are just so many different ways all this could play out. E.g. Gavin proposed maybe this makes for a huge smartphone/PC supercycle because inference moves to the edge. I don’t know. It’s all a lot to process. I don’t feel compelled to rush to a conclusion. Though I will admit there is something about this, as far as headline risk goes, that feels more real and relevant than most things that cause “markets in turmoil”.
Tomorrow we have a prospect onsite in the afternoon but I should be able to get some work done in the first half of the day. Not sure what second half looks like. Probably mostly admin and lighter research with meeting stuff mixed in.
I think part of the issue, too, is that I just pivoted suddenly to this small semi company and I don’t really know how to attack this or what I am expected to do here. Working on it with someone else. Need to just build out a plan. Again - need to decide - what is the concrete next thing to do on [small semi co]? “Come up with plan” is not a deep work thing, that is more an admin thing. So that is part of what I will do rest of day.
I like this daily writing thing. Engagement has been slow but I think with consistency it may increase over time. I may make it my “thing”. E.g. put in bio “I write a daily unfiltered journal entry about my work as a research analyst each day” or something like that.
1/28/25
-OTA AI risk
-Small semi co
-Zichen Wang (must follow) on DeepSeek
Watched a webinar on OTAs/AI Agent risk. Long story short, it seems like this is still very, very early. I don’t see many people using this in the near term. Medium-long term, I see a couple possible risks it poses to OTAs. Increased price transparency is one but I don’t think the models are really at the spot they need to be to drive that in a big way, yet (to actually price compare all the available inventory). There could be take-rate ramifications over time. In a typical OTA B2B contract you might have a 15% take rate, and the traffic partner (e.g. OAI) might get 3-4pts of that. So you could see take rate compression if this is a big portion of traffic. But then again, that is sort of replacing a Google CAC, which can be quite expensive. 50%+ of take rate. From a p*q perspective, you have some risks around p but people are unlikely to travel more just because of an AI agent so not enough offset from q. So I don’t know. We will see, but it all still seems very early. Will be interesting to follow.
Worked on small semiconductor company. Have much more to say about this but sitting on it for a bit. I hate even mentioning something like that but I feel like it’s necessary for this exercise? I feel like that guy who is always like “met with a 60 trillion dollar fund manager the other day and he told me…” Oooohhhh tell us! Tell us! Please!
Prospect in. Thought it went well. Good job team. Again tomorrow.
Listened to Zichen Wang from Pekingnology on a podcast talking about DeepSeek. He is really sharp. Worth following along. I need to read more of his stuff. See the thing I just tweeted.
1/29/25
-Semis $LRCX $AMAT
-What Semi-Analysis is doing is really cool
-Waymo market share
-DeepSeek, AI capex, etc.
More semiconductor stuff. Looked over the LRCX call briefly. It seems like things are starting to pick up. Maybe the WFE market has finally bottomed. NAND starting to pick up. It’s interesting to juxtapose LRCX/AMAT diversification with the small company I am looking at. It plays specifically within the deposition space, and niche within deposition is what they are particularly strong at. They’ve been around for decades but it is still tough to tell where that goes over 5-10y given how fast everything is moving. I don’t exactly know how to diligence it, or if it’s more just a position sizing thing. Don’t get overly confident.
As part of diligence on that, had a call with the Semi-Analysis sales team. Unfortunately the more institutional level stuff, which I didn’t even know they had, is outside of the budget. I had no idea how big that Semi-Analysis team was. I thought it was just Dylan and Doug writing a newsletter and maybe a couple supporting ops people. Nope. They have sales. Multiple analysts. Going to several dozen conferences a year. It is awesome to see what they have done. The rise of independent research is really inspiring. Took long enough. It is meritocratic. I like to see people like that succeed. Like Mostly Borrowed Ideas. Scuttleblurb. Etc. I think someone recently tweeted about that - how that is more of an option these days for people who don’t like their seat. Especially if you’re willing to make it more of a lifestyle thing. That had always been my framework for that kind of work, it was interesting to see sort of an upside case via Semi-Analysis. Very cool to see what they have done in an industry that typically only pays top dollar for name brand. Rooting for them.
I realized that Waymo SF market share from Yipit is only for the Waymo operating zone. I guess that explains the discrepancy between the numbers in that chart that went viral on X and then I think it was the LYFT CEO saying their share is much higher than what was shown?
I still don’t know what to think about DeepSeek. I don’t know how game changing it will or won’t be or how thing shake out. Their are so many opinions out there. Some highly technical but well explained, some very jargony, some overly simplistic, and many very confident. I don’t know how you get to those conclusions that quick. I think I need to see how it plays out. I am unsure how much time I should be spending on it in the interim.
Relatedly, this AI capex is insane. I wonder how each $1 breaks down between - you know, you have land/server cabinets/DCs, then network infra, then servers, etc. And then within that - NAND, DRAM, logic, whatever else. Or like if you could make a pie chart of all the companies that the spend is actually going to. And then click into them to see who benefits from it second hand. But… these capex numbers are huge. What is the risk of it? I guess they could shut them down if something really crazy happened e.g. DeepSeek makes it all totally useless. Sunk cost at that point. I don’t know.
Prospect in today. Think we did good. It is in God’s hands now.
1/30/25
-Semi stuff
-A couple questions about pods/Surveyor for anyone who is familiar
More semi stuff today. Just going through materials still. Need to start building out the skeleton of the report soon and figuring out what gaps need to be filled via expert calls, etc. I don’t know, though. Sometimes with e.g. PEP I basically go right to the skeleton. With something like this I need to spend more time reading about it. Partially just as a catalyst to think about things, but also because I don’t know what I don’t know. So idk. Probably a few more days of company materials and some sell side stuff. We will see. Kind of want to act quickly too, though. Trying to balance the two. Don't want to get ahead of myself.
I am still trying to figure out what the UBS analyst was referring to on the Lam call when he referenced a peer saying NAND will double this year.
Read the Micah Nance article. Seems like a cool guy and a cool place to work. Some comments, as well as some questions for people who know anything about pods and Surveyor. Stuff I thought was interesting in the article: has a wife and two kids. Wakes up at 530am to play tennis, and then office at 730am. Home in time for dinner but usually sending emails etc. at night too. Doesn’t sound like terrible work/life balance. Talks a lot about the emotional side of things mattering, too. Doesn’t want to add extra pressure to get results, screws up analysis. “I want to inject zero excess stress into our decision-making.” Pushback is good, but don’t be argumentative or overly emotional. Some questions:
1) it says he and his team of 9 analysts/associates cover 150 tech and media companies. ~15 each. Do we think that is within the book or total coverage? I hear about pod people covering ~40 names each and I am like - how do you keep up in any significant depth with 40 names each quarter. But IDK it is a different process and structure to some degree. 150 names in the book? Or coverage universe/watchlist? 15 is a heck of a lot more manageable. It’s like - a lot of people run single person shops and might have a 15 name book they follow really closely and others tangentially but you aren’t like reading 50 calls and updating 50 models per Q. Idk. Any color into this is appreciated.
2) And then he says - I “build my conviction and think with duration.” There is a stereotype that pods trade around Qs or maybe a year. MAYBE 18 months. Is this not the case at Surveyor? What is “duration” to him?
1/31/25
Mostly semiconductor company. Lots of questions. Will put skeleton of report together next week and kick off Tegus project soon. HEI does not look cheap. Interviewed an intern candidate. Need to plan out week better to make sure I have enough hours to get key task done. Idk. More to say but no time. Going to play soccer.
2/1/125
The boys are in town. Not much was done. Two things I will be thinking about/monitoring for info on more: 1) the nature of AI/DC capex changing in a way that negatively impacts $CMI $ETN etc., 2) L/O use of factor modeling.
2/2/25
I am trying something a bit different this week where I plan out my whole week at once, at least to a large extent. For Monday - Friday I already filled in time blocks for anything that is a fixed time commitment as well as my deep work hours. This compares to usually figuring out the whole day just the night before. This can sometimes lead to issues where I don’t realize I have certain things throughout the week so I plan to get more done that week than I actually can. Or sometimes I don’t realize until the night before that the next day has barely any time for good deep work slots. Having it all planned out ahead of time gives me a better high level view of how the week will unfold, which days will be shallower or deeper, how much I can get done, etc. Currently the plan should allow me to finish the skeleton of the small semi co report by EOD Tuesday and leave a few hours for miscellaneous further reading on Wednesday with additions to the report as needed. We will see what interruptions and obstacles come up and inevitably throw this off track!
I am going to launch a Tegus project on the small semi co. We have a process where we, before considering valuation, must reach a team consensus that the company is high enough quality to justify investment. Only afterward do we do the valuation. And when we first introduce the company, it’s just a quick elevator pitch from the analyst that brought it and then the rest of the team takes a week to review, ask questions, etc. before we debate the quality question the following week. So - I haven’t even submitted the report for Intro yet. Usually Tegus comes later in our process. But I think it makes sense to move it up. If we cannot agree it qualifies, perhaps that discussion will guide us in what to ask on the Tegus calls. Or if we do agree it qualifies, we can use the Tegus calls to help us with valuation. If everyone shoots the idea down right away, perhaps we will end up wasting and/or cancelling the Tegus calls. We shall see.
Tried a new church this morning. Kind of looking for a new one but I don’t know. Figuring things out. Then spent most of the day handling errands and getting ready for the week.