AI in Investing: Guest outlines a practical AI-driven equity research framework with 13 tested prompts, emphasizing it as a tool to enhance speed and breadth without replacing human judgment.
Idea Generation: Describes using AI for Phil Fisher-style screens and Special Situations (spin-offs, restructurings) to surface candidates, accepting noise to find a few high-quality prospects.
Research Tools: Highlights Google’s NotebookLM to ingest 10-Ks and expert transcripts for context-only analysis, uncovering themes, omissions, and between-the-lines insights.
Devil’s Advocate: Uses AI to critique thesis logic and counter confirmation bias, pairing machine checks with a human-led final decision process.
Data Centers: Warns of heavy AI-related capex and low server utilization, drawing parallels to the late-1990s dark fiber overbuild and potential poor returns on massive infrastructure spend.
Market Outlook: Applies the Gartner hype cycle lens to AI, noting frothy valuations, earnings quality concerns, and challenges finding bargains; long-term return expectations should be tempered.
Overall Perspective: Human remains in the driver’s seat; AI boosts productivity and risk control but is not a magic alpha machine, and caution is warranted amid elevated asset prices.
Transcript
Hello and welcome to the Stansbury Investor Hour. I'm Dan Ferrris. I'm the editor of Extreme Value and the Ferris Report, both published by Stanberry Research. And I'm Cory McGlaclin, editor of the Stanberry Daily Digest. Today we talk with Gary Murus, chief investment officer of Silver Ring Value Partners. Gary is an old friend and he is a great investor and a very smart guy and he's figuring out how to use AI to do investing to analyze stocks and make recommendations and do research and we will give you a link to download his PDF so that you can see what he's up to and learn how to do it too. So let's talk with Gary. Let's do it right now. Let's talk with Gary Mashurus. Gary, welcome back to the show. Always good to see you. Likewise, thank you very much for having me. >> I was really happy that you reached out to us um with a topic that of course everybody's fairly well obsessed with these days and I thought, "Wow, this is exactly like you picked the exact perfect moment uh for this." So, so good on you. I mean, I know you're not in a marketing business, but if you were, this would be the this would be the time for that message. And the message um just for our listeners benefit, Gary reached out and said, "Hey, I have a process that uses AI, you know, for my investment analysis and I want to share it with you." So I thought yes, because I'm trying to do that myself and I haven't necessarily come up with anything really super organized. So I'm I'm a listener today as much as a host. So, um, >> well, I think we're going to have a good conversation. It's a good topic that I think everyone is struggling with right now. >> Yeah. So, I I feel like I should put it in your hands. Um, wherever you want to start with this, um, you know, your your AI investment analysis process, just go for it. >> Yeah. No, absolutely. I mean, I think look for some background, right? you know, when I started as Fidelity as a young analyst 25 years ago, um, you know, Excel was a pretty common tool, but you had, you know, some grizzled veteran portfolio managers who were still using the legal pad, right? And I literally remember as a young analyst seeing them do modeling on a yellow pad. And I think pretty soon, if not already, if you're not using AI, that's what it's going to feel like. meaning that it's maybe hardcore that you're doing these things manually, but I think that it's not optimal, right? And I think there are two extremes like people are people are going to either the extreme of hey AI is going to solve everything. It's going to be a genie in a bottle and it's going to give you like these amazing stock pickicss and you know all you have to do is figure out the right magic prompt and rob it the right way and the genie will come out, right? And then there's the other extreme of like it's all hype, it doesn't work. Real investors don't use AI. They just sit there like Warren Buffett and you know read 10Ks ideally physically printed, not even in, you know, digital form, right? Um so in somewhere in the middle, I think there's an actual opportunity to add value without sacrificing quality. And that's I think what I've been working towards. >> It's ultimately a tool. So, you know, if you don't have a hammer available to you and someone gives you one, um, it feels great and it and it makes putting nails into wood a lot easier, right? >> But, uh, you know, it doesn't actually it doesn't really solve all your problems. It doesn't tell you how to design a house that won't fall down, etc., etc. So, >> um, yeah, it's a great tool. Yeah. >> And and I think, look, there's two [clears throat] types of use cases. two as I call them category one and category two and I think category one is basically something where you're trying to save resources and that could be primarily time could be money but you're saving you're doing the same thing you could have done as a human you're just doing it faster cheaper some using less of something right and that's probably the most common and then there is this elusive what I call category two and category two is you're actually doing stuff with AI that you couldn't do as a human or couldn't do nearly as well even if you had nearly unlimited resources. And I think that's where maybe the holy grail is down the road. I think most of the use cases I've been able to figure out right now are category one. But I think I have one one or two use cases which kind of enters category two. And I think you you mentioned the framework. I mean, so I put together this AI equity analyst framework that um you know, I think, you know, I've made freely available partly for selfish reasons. Like I want people to read it and tell me either what I'm missing, where I'm wrong, or what I can do even better because this is the wild wild west. But that being said, I think I have a pretty good starting point, which is what I think it is. It's not the final answer to anything. It's a working kind of a framework that I actually use dayto-day because it's just amazing once you roll up your sleeves how much AI can improve the process of a long-term serious investor. >> Amen to that. Um even just uh for me just like as a search tool through multiple 10Ks and stuff, it's that it's unbeatable, >> right? If you just feed it 10ks, if all you ever do, you know, at this point for me is feed the thing 10ks and use it for search for various, you know, prompts. >> That alone and that's not wrong. I think that's a great use case for sure. I just think there's so much more. And I think, you know, >> if you I think that actually even though it's just a step back, the way you use AI right now is actually probably of secondary importance to like start using it. So if you're listening to this right now, like just roll up your sleeves and start building it into your process because if you don't you wait for some iteration of Chad GPT17 or Gemini 12 or whatever you you're not going to have the experience. It takes reps. It takes like you have to kind of do trial and error because by the way there are some pretty atrocious results obtainable with AI as well. you know I you know so in this framework in the AI equity analyst I've tested each prompt and some of the versions I had to go through to get to a what I think is a working version they were atrocious you know and some of the use cases AI is just not ready for but you're only going to find that out if you start doing it now but and if you do you're going to start layering AI into your process and I think that could be different for you it might be know a lot more than me or a lot less. But I think start doing it is the main message. I would say even more important than some clever prompt that prompts AI just the right way or something like that. >> I'm really glad you said that. Yeah, I'm I'm I'm glad you said that too because the conversations we've been having lately or even I've been just been thinking about about AI now now compared to say even a year or two years ago when it was what is this how is this going to change the world and take everybody's job now some people still you know think that last part but um the conversation now is just more practical like what you're saying you could have more practical conversation about how does this actually work uh what are the benefits etc if you start trying it is the key thing and so I'm glad you brought that point up to the table >> yeah no absolutely and I would say again there's a lot of focus on prompt engineering and as I just mentioned it can be important like it can definitely make a you know a bad use case good but I think and this is my you know I don't have any data to prove it but in my subjective experience so far like having the right use case so asking the right question is like 80%. And then figuring out how exactly to prompt AI for that question is data 20 because I think that you know like some of my best aha moments with AI in the last six months and I've literally spent like over a 100 hours building this out um because it's the return on that time is just so huge. Like I invested, you know, 100 200 hours but I'm going to save hundreds of hours per year. So the ROI is incredible. But like some of my biggest aha moments were like, "Wait, what if I use it for this?" Uh, and and then I would have like this blocking. Oh no, no, it's going to be terrible at this. No, I'm not going to do it. Well, let me try. And maybe at first it's terrible or maybe it's 80% terrible, but there's a 20% useful component. And and this is another very important point. I think people like talk about AI hallucinating, right? or you know AI giving you the wrong numbers and then they use it as a reason to not use it. I think that that is actually like completely backwards. You have to figure out where is it okay to have hallucinations like where is it okay to have wrong answers. Let me give you a very straightforward example. So at the top of my investing funnel you know I have idea generation right? So, I'm looking for potentially attractive investment candidates. And it actually does not matter if the list AI generates uh and I have two prompts in the framework, you know, special situations and a kind of a Philisher compounder prompt that I put together and I kind of test it out. It doesn't matter if gives if it gives me 20 ideas and 16 of them are terrible. Like what matters is are there four good ones in there, right? Because I'm not saying you want to delegate everything to AI and go to the beach and come back and see how the portfolio is doing. That's not it, you know. I'm saying you use it as a tool. You stay engaged with it and you basically go and say, "Okay, now I have these 20 candidates that AI generated that I would not have had." And again, using both my own, you know, knowledge and also other AI tools saying, "Okay, go through one by one. Garbage, garbage, garbage. Oh, interesting. garbage interesting interesting maybe you know okay and now I have generated something I wouldn't have had even though m most of the output might have been not great so that's an important point I would say >> yeah it's a it's a learning to use it is a process of trial and error that I agree and it's worth it and there's no other way to go about it you're riding a bicycle you're going to get on the thing and you might fall and that's the way it is um I'm glad you said that it's I'm glad you emphasized early in early in what you just said um that asking questions and and you know querying prompting is is really important because finally um it there's finally a place in the world for somebody who might not be able to write code and might not be a highly technical individual but who has a lot of questions about what they're doing and and let's face it if you're looking as we are, as you are um in the in our somewhat different roles, investment ideas, you have a lot of questions all the time and you want to know, you know, the best company in this or that industry or the company that has the least debt in this, you know, industry or anything like that and much more that's much more complicated than that. Um, you know, it it's a it's just a wonderful it's a wonderful thing I think that's happened to us that we have so you know this resource that we can just ask questions to without needing a degree in in you know computer science to do it. Um, >> yeah. No, absolutely. And I think it's funny you mention it because I have my my teing here. My I do have computer science degree from back a quarter century ago, but I don't think it's useful because it teaches me how to code something. it I think you know and I think this is an important insight right people think that okay AI maybe will flatten kind of completely level the playing field and the return to skill will go away because everything will everyone will have access to these like you know genius AI systems that they can just replicate that >> no that's not what we're saying at all >> yeah in the driver's seat >> I think the return on the real skill and insight is going to get magnified by AI AI, right? Exactly. Because >> if you can add, you know, what what is going to go away is the grunt work required or at least a chunk of the grunt work required to execute you know uh on those insights. But if you imagine if you could you know think about you know yourself as okay you have the true value adding things you do whatever they are might again different for you than for me and different for you to listen to them for me as well but you know you have them and then there's all the blocking and tackling to actually make those those insights useful and make them get into your investment portfolio right and so now you can just leverage those insights so much more And you you can skip some of those steps, but again, you're not and this I call it when I wrote the framework, you know, there's like a big orange warning on like page two or something of the PDF, which is this is not a shortcut. This is not meant if you're lazy and you just want again that made magic genie in a bottle, this is not it. What this is is it forces you to be explicit about your process and allows you to do more quality work faster, not just cut corners and hope for the best, which I think is an important distinction. >> Right. Right. So, I'm glad we agree that, you know, the the the human being is in the driver's seat. This idea that AI makes, you know, our humanity like redundant is ridiculous. All right. So, shall we? We should get into this a little bit. And um I'm sure our listeners like, "Okay, what how does how does Gary really use this? What is he doing? What's the secret?" And you have a you have a um a document that you've created that that tells this information, but but we'll, you know, we'll we'll introduce the listener to that. Okay. So, where where would you where should we begin? Where where do you begin? You said idea generation. Um >> yeah I mean no absolutely and I think idea generation is a great use case because when you think about what is AI good at one of the things it's good at is synthesizing a lot of information right rapidly and so I think that so I have two prompts you know so in that PDF you mentioned I have 13 well tested prompts that that work you know and um you know two of those prompts are idea generation prompts so you start at the top of your beginning of your research funnel Right? And you say, "Okay, what candidates should I consider?" And one of them is a Phil Fischer prompt. So Phil Fischer, famous, you know, growth investor, you know, I would think of him as a just a long-term intrinsic value investor with a focus on growing companies, but he's known as a growth investor. So I'm not going to quibble on that. And you basically tell AI, okay, act like Phil Fischer. You know this is not the exact wrong but go and find ideas within certain parameters. Let's say I want certain countries certain market cap ranges what have you and have that and and give me a list of things that fit. And Bill Fischer by the way in his book common stocks uncommon profits had these 15 points kind of a checklist. So you're basically having AI run through the entire market or some subset of the market and have it judge each of these companies on on that those 15 points. Um you're going to get some list. It's going to take a long time. By the way, this is not like, you know, you're going to like you're going to go, you know, put it in. >> Yeah. >> Yeah. you it's going to be it's not going to be instantaneous but you're going to get a list and you're going to go through that list and you may find some hidden gems and I don't know New Zealand that you never would have heard of otherwise so that's qu a very cool use case um again a lot of human oversight you don't want to assume that it's right some of them you'll be like huh really and there's the I the intelligence in AI wait that doesn't like this can't possibly be a fit for what I'm looking for but ignore that don't worry about it because you only care about the ones that are interesting, not the mistakes. And a completely different approach, you know, like let's say you are like into just deep value. You want, you know, special situations, things like that, which I know I love, a lot of my investor friends do. And people like there are shops that have a full-time special sits analyst. All they do is find and kind of prep these ideas for the PM to look at, right? Well, you can go and have AI do that for you. like find all the special situations, spin-offs, you know, restructurings, whatever. You know, we don't need to get into the the nitty-gritty, but the point is, you know, come up and run that monthly. You can run or you can even create, you know, a task to have it update you for uh when new ones come. So, I think at the top of the process, the sky is the limit, right? And I think I think that is just a terrific use case where again, not saying that that's the only way you should generate ideas. I think there are other ways as well, but I I found these two ways to be pretty pretty cool to be honest. >> I find it pretty cool, too, man. [laughter] >> So, I agree. That's good. >> I'm glad you mentioned Phil Fischer, too. >> Yeah. And and that's my point, too, is that this approach is really agnostic to how you invest. Like, it really doesn't matter if you're looking for, you know, completely different set of things that make an investment attractive to you than would make it attractive to me. Like that's what I mean make AI your own like it's there to serve you like execute your process not my process you know and not someone else's process but another question is okay so you find this list and what do you do with this you know okay and you can go hardcore go to the library put on your favorite music and your headphones and start reading 10ks but that is and there's a place for that I think still but I think it's too early you're getting this giant list you're pairing it down to some subset And now you want to go through it and figure out which companies are actually worth the effort. And so I have a whole bunch of prompts from, you know, basic company history and description to like put, you know, your Michael Hoer structural analysis head on and analyze the quality of the business. All kinds of stuff that the main purpose is for you to get to a quick note, right? because you get this giant list and you don't really want to spend your manual labor hours equally on every company on that list. That would be foolish. So you what you do is you essentially use AI to come give you enough analysis to say no no no maybe ooh this is interesting. And now when you say, "Oo, this is interesting." This is not, "Oo, I'm going to buy some." This is, "Ooh, that's when I'm willing to spend the human research uh time and go deeper." >> That I want I want our listeners to know that that moment right there that Gary just described is it's the one I don't know how he feels about it, but I feel strongly about it. That is the moment you pray for every day. Oh, this is interesting because so much of the time you say no, no, no, too much debt, commodity business, no moat, whatever, you know, whatever it is you're looking for. Like you said, it can be any any investment style. But [clears throat] that moment when you go, oh, this is worth looking for. There's something happens where you're confident that even if you wind up saying no, the next hour or two of your time is going to be really well spent. >> Yeah. And and by the way, again, I've done this for a quarter of a century now. And even though I'm a process guy and I'm very like rational, methodical, I've just learned to trust my gut a little bit. It's okay. Sure. And you get this feeling like you have to put every at least I do. I have to put everything else aside and start researching this company. That's a good feeling. That's how you know. >> Yeah. Yeah. >> That's what I'm talking about, man. It's like finally I because otherwise >> um >> I don't know how you feel about this but but I hate the feeling of being unfocused. Sometimes I just push back from my desk and say, "Whoa, whoa, I need to I need to step away." Then I sit down and I write with a pen on a piece of paper what I'm doing next to kind of refocus because that focus, >> man, focusing on something and really putting your aiming your your power at aiming your highowered rifle of your mind at it. Um hopefully with a good accurate scope and [laughter] a good marksmanship um is is what we really live for. Um we analysts, we equity analysts. Um so beyond so so here we are, we are we have you know we've gotten into idea generation. We're now at that moment where we say hey this is interesting. Um and and then you're talking about trusting your gut and you know reading the 10Ks and all that. So what you know what's left? What's left for an AI? >> So there's one more step. There's one more step, you know. Yeah. And I think there's a tool. I'm sure you've heard of it, but you know, I don't know how many people have actually used it. Seriously. And I honestly think it's the most powerful free AI tool like bar none. Like, and it's not Chad GPT. It's not Claude. Um, it's not even Gemini. It's Google's notebook LM. >> And if you >> You're the second person to tell me that. >> Yes. I'll introduce you to the first. I want to get to know them. you know, >> it's our our research director, Matt Wine Shank, told me. >> There we go. >> But I think it's so important because I've had all these companies pitch me on their AI products and, you know, I have nothing bad to say about any of them. I might use some of them down the road. Um, but notebook LM totally free. Number one. Number two, it it basically So, if you don't know how AI works, right, just like a step back, there is something called the context, right? An analogy. So AI the reason like a lot of AI so large language models hallucinate is they have this this basically they sucked in hoovered up the internet for some portion thereof into their context and then when you type in you know you adding context and they're basically kind of combining those two and you know finding the answer for you I'm simplifying there other steps there's thinking models and all that so nom is like a blank slate the only context it has are the source sources that you upload and you can upload up to 300 sources and so I'll tell you exactly how I use it. So for example, you uh I start with downloading you know let's say 20 annual reports if they're available PDFs are fine. You throw them into your notebook. So nobook alm you create a new notebook again it it has not sucked in the internet that its knowledge is limited starts at zero literally starts at zero and then you pop you tell it what you wanted to learn. So now rather than prompt engineering, you're context engineering. you're engineering what context you want the LLM to assess and you let's say you put these 10 annual reports or 20 annual reports and then you ask it okay first you can just have it say tell me the story you know tell me a quick story you know of how this company you know evolved over 20 years right by the way if you really want to save time you can create a podcast you know speaking of podcasts right you can create a a podcast with two hosts uh using just information that you've uploaded and go to the gym, take a walk, whatever, and then come back and you'll have no know learn a bunch of stuff about this company. There's another level to the game here, which is I mentioned earlier on that there are some of these category 2 uses like things that AI might be able to do that you and I can't, right? No matter how much time and money we throw at it. And so what I've done is, you know, so I use expert interviews as part of my research process. So as I go deeper into a company, I don't want to take management's word for it. I want to go ahead and say, okay, what are other people saying, customers, former employees, suppliers, and so forth, right? So let's say you have access to an expert network, and you can download the transcripts, and you add those to notebook. And I have a really cool prompt in that AI equity analyst framework that I mentioned which actually I think is a kind of a category two prompt in the sense that what's happening there is you're asking it to find common themes and trends among the you know among among the interviews and let's say if you are had a 500 you know 50 transcripts or something like that it would take you a long time but even then you you might not connect the dots like oh this happened here and that happened here. This former employee said that and the customer said that. So that amazing use case. And so when I was a young analyst at Fidelity, they brought in CIA former interrogators and they kind of try to teach us how do you interview uh people? How do you detect if someone is lying to you? And one of the things that always kind stuck in my mind is that you know they told us when people lie they don't actually tell you a complete falsehood usually in in relation to uh the question you're asking. What they frequently do is they will answer a slightly different question. So let me give you an example. Let's say you're interviewing a former you know employee of a company you're researching and you say hey how's the culture? Right? and and the person says something like, I don't know, you know, I really like the guys. We really enjoy going out for beers after work, right? So, and you might just move on and feel like no cognitive dissonance whatsoever because you feel like you gotten an answer and you got a good answer. But if you pause for a little bit, you might realize, wait, that that wasn't a answer to my question. I didn't ask you if you like guys the guys you work with or if you went out for beers. I asked you how the culture is because maybe you're going out for beers with the guys you like to and moan about like how bad the culture is, right? You know, you know, and so AI has this amazing way of, you know, and the prompt literally tells AI find things that were unsaid that are literally between the lines. And I think it's it I've seen it tease out some pretty neat things in my experiments. Um, and I think it's an awesome use case that's perfect for the tools that are available right now. Oh yeah, I love that idea. I love the idea of, you know, feeding it conference call transcripts and what aren't they telling us? >> Yep. Exactly. Omission. I mean, so I mean, I think om omission uh m uh stuff is also a big way people lie, right? You know, they they tell you part of the truth but not the whole truth, you know. And uh I mean man, you know, I was just talking to a former I IR officer of a a large company that we've known each other for a while. And I'm not going to mention the company, but the things that the person told me, you know, the things you don't know as on the outside of a public company that are going on on the inside sometime sometimes that's a big it's a big category of things, you know, let let me tell you it's and so I think that especially look as a concentrated investor which I am, you know, if you have like 20 basis points in each idea, maybe you don't care, maybe you investing based on no I don't know quality factors, that's a different game than what I do. But if you're a concentrated long-term investor like myself, then having like a big blow up can be very costly. And I'm not saying this will prevent all of them, but if you can reduce them, that is very valuable, you know, at least to me. >> Absolutely. Um, [clears throat] so like how is there a point in your process, Gary, where um beyond which AI just simply, you know, you've you've still got to, you know, tough it out the old school way. >> Yeah, absolutely. I mean, I don't want AI to do no the thinking for me. So once I kind of sunk my teeth into an idea, I do still do all the deep research. I I do read now some I do faster, you know, uh but I still do it. um I still come up with my own range of values. I still make the decision and there's actually a uh so all of that is me and in that you know in that framework I literally have a flowchart which has steps in the process and uh you know blue steps are AI blue and orange steps are combination of human AI and orange is purely human and what you'll see as you follow go further and further through the process is that at the beginning of the process there's a lot more blue or blue and orange towards the end it's mostly orange it's mostly me right but that way I can do that mostly orange the human part on a lot more ideas >> so this is a very interesting topic because I've seen studies where um medical diagnoses were done with AI only human only and then the combination and the AI outperformed the other two so you know um a medical diagnosis is not an investment decision you I get the But uh you know you understand why it's an interesting question at least right? >> No for sure. I mean and by the way listen maybe that's where we're going right maybe in some number of years you know AI will be hiring me you know and AI will be say Gary this is what I need you to do today. I will make I will I got this right. So you know and by the way I've seen some LinkedIn posts where like they show the org chart of a company and all the seuite is AI you know CMO is an agent you know you know right so you've probably seen those too but I think today we're not there I also frankly I'm a little bit old school like I know what the studies might say but I don't know when there'll be some glitch when there will be some issue when there will be something right um and so may I know again I you know see this these gray hairs you know I've gotten you know through the not hard knocks you know in the markets over the last quarter century and I just feel like when the people are trusting me with their money they're not trusting me to go to the beach and turn on some AI algorithm they want me to actually make the decision so I think it's super important to I mean different people will come out come out at different places on this but for me I bring actually AI in the end and let me tell you how I do it because I think I think it's a very important. So you probably know that I'm a big fan of behavioral financing. My substack is be called behavioral value investor for that reason. I think like essentially in investing we are frequently our own worst enemies. So there's two more steps at the end that you can do. No. So you you've done all that you know early stuff then you've done the d uh deep digging and then you end up um you're going to pull the trigger right? Well no wrong. You not you don't pull the trigger. You do two things. number one. And so you write up your thesis uh and you put it into AI and you have it check it like you know check the logic check I'm not talking about spellch checkcking or grammar here you know talking about you know are there mistakes of omission you know is it internally consistent for instance maybe you say at the beginning I think investment XYZ is great because of A B and C but then you only go and show A and B or maybe there is evidence in your own report that C is untrue right So things like that I think is a very easy and useful way um to find to have AI point out mistakes before you actually put money to work behind them. Right? The other thing is something I call the devil's advocate. And this is an idea I had back in the day. We got a group of, you know, grizzled investors together, friends of mine, and I got them together and I said, "Okay guys, we, you know, we all know behavioral biases are real. We all know we make mistakes. So, how about we kind of create a pack, a little club where once a year you'll be asked to spend a couple of weeks, you know, coming up with a serious opposite case, the devil's advocate case based on the information someone sends you on the company and and your and your work buys you that once a year you can ask, you know, uh, someone else to do that for you. And it worked probably like six months, right? And then people got busy. I mean, listen, it's a heavy lift to ask someone to spend a couple of weeks on an idea they don't truly care about, right? It's like we're human, right? You know, it's like I talk to college students, what I teach at BAPS and I talk to other groups about investing. I always have the these slides in my deck where, you know, about comparing investing to dieting, right? So, you have, you know, I have one slide with this really fit young woman and with healthy foods and exercise and which is what we know we should be doing. And then there is the next slide which is this big fat guy chomping on a donut which is what happens in reality right and so like investing is a little bit like that like there's what we wish we were doing ideally and then there's like the reality hopefully it's not as bad as that but you know you I think you get the gist and so now I have this AI devil's advocate prompt where its goal now is not to stay limited to just my report its goal is to go and be this hunter killer, you know, of my thesis and help me find things that are missing like, you know, obvious common bias called confirmation bias. What is it? You kind of seek out things that agree with your already predetermined conclusion and you kind of ignore things that go against it, right? We all do it except for me. No, I'm just kidding. We all do it. That's the whole point, >> you know? So, so I think AI is amazing at that. So you start with AI, you use it a little bit less as you go deeper and deeper into your idea and then you do the pure human part at the tail end. You kind of use AI to help you avoid mistakes. So that's kind of my process. I mean I'm, you know, if I'm sure there are people who have, you know, found things I haven't. Again, it's the wild west, but that's part of the reason I'm sharing this because it's so exciting and I think there's so much we have to gain from each other in terms of collaborating and looking at how different people are doing this because we can only get better together. That's the only outcome. >> Maybe before we go any further, we should we should tell our listener how they can get to this document of yours because, you know, you want everybody to read it, right? >> Yeah. No, I mean, I think I think it's worth I think it's a it's a it's a good one. You know, I think I've had a lot of positive feedback and I've had people kind of, you know, give me some good push back on some of these things and at some point there'll be a version two that will be even better because of it. But I'll I'll share a link with you that you'll uh you'll have. It's just a, you know, it's a 33page PDF where I kind of go through everything step by step. And also, you know, for some reason you can't find the link, if you go to the behavioral value investor Substack, it's uh there was an article a couple of weeks ago that I wrote which describes the overall framework where you and then you have a link to download the um the PDF. >> Sounds good. Sounds good. Thank you for that. Um where do we go from here, Gary? I mean, um, it sounds like, you know, this is sounds like early, you put a lot of work into it, but this is early days like are you you you're going to get feedback from, you know, perhaps dozens, hundreds, I don't know, of people and then you're going to come up with a 2.0 and you are just to be I just want to be very crystal clear in case this isn't clear to anybody. You are using this to allocate real money for your investors. >> Yeah. No, absolutely. Like I was I just had my annual meeting for the partnership uh on Friday and I I was talking about this with my partners and I literally you know pulled up the screen of a notebook I have with a gazillion sources of a British company I'm researching and you know showed some of the kind of live stuff like it's great. It's not it's not like a oh let's design some theoretical thing and see you know throw it out there on the internet and get some feedback. It's like actually what I'm doing understanding that you still have to iterate and what I'm doing right now will probably be better when I do it in 6 months and 12 months and 18 months and so forth. But it's absolutely um a useful tool today and it will only get better. I mean I think one you know so you know I think you and I met in Vale right the friend friends friends bali's um conference for the first you know bunch of years back right and I was in veil um this summer and I gave a talk on AI there and you know and so one of the things I had you know you know you know how these things go it's a short talk and then you do Q&A from other experienced investors and and so right where I was supposed to take questions I said, "Wait, actually this time we're not going to do that." And the reason is I already asked AI what questions you guys are going to ask ask and here they are you know and you know so we had we all had a good laugh and of course I took questions but I kind of you know it was interesting to see what AI thought the objections were or the challenges were with using AI and one of the a few things that stood out. One was, hey, is this somehow going to dull your ability to do research? Like you're going to atrophy your like primary research muscles. And I think the answer is could if you use AI incorrectly, but if you use it correctly, you're actually going to get more reps on the things that are the most value adding. And you're going to spend less time going, let me put it this way, you're going to spend less time going to the gym and setting up the weights and doing all this other stuff and changing into your clothes and more stuff on the bench, you know, you know, uh, pushing at the weights. So, I think that's how I view it. Um, >> I was going to use the analogy of a carpenter who gets better with with his tools over time and becomes more creative. >> No, absolutely. The other thing AI point, and this is the million-dollar question, or maybe the billion dollar question is [laughter] AI, one of the questions was like, hey, it's awesome that there's all these tools and you can do all this stuff, but is there any evidence that AI actually improve return improves returns, right? And the honest answer is we just don't know. Um I think the head of uh Citadel was on Bloomberg a couple of months ago saying ah nope you know it's nice making junior analysts more efficient but it's not going to add alpha and maybe he's right. I don't know. Um and I think he doesn't know to be honest it's a it's still being kind of worked out and decided. I can tell you with near certainty though is if you don't use AI you'll be at a competitive disadvantage. So maybe you, you know, if you use AI, well, it's not going to magically give you 300 uh basis points of alpha extra per year, but if you insist on doing things kind of the old way uh completely, I think chances are you'll be at a disadvantage. And I don't think I don't think you need to do that because it's it's not that hard to do it. Well, >> right. And if we're if we're um just sort of uh how does one say analogizing about about how this works um or might work out in light of the Citadel guys comments. The first thought that came to my mind when you were talking was Renaissance technologies. I mean if you would have told me before that happened that a bunch of mathematicians and physicists were going to get together and go through a period I think it was 20 years of 80% annualized returns. 80 80 um something truly insane. It might even be 60 or 80, but any of those numbers is just off the charts insane. Um if you had told me that, I would have said, "Oh, that's silly. They're not Warren Buffett." You know what I'm saying? It just would have been crazy. But that's exactly what they did. So, who who knows? This is a tool just like the mathematical tools they were using. maybe we could say um or similar enough to and and maybe there's a you know an AI renaissance out there that's going to just you know shoot the lights out. >> Although you know I'm going to you know like just push you back just a little bit because I think that there's a here >> you know in that like I'm a process guy. I'm a long-term guy, but there's so many hypeers out there, and I know it's not you or any any of your listeners, but like, you know, you ever watch YouTube and get these like commercials like, "Hey, bro, get my trading system and make a billion bucks from your couch, right?" And I feel like AI and the hype surrounding it just gives more ammunition to those like you know shysters bas you know like people who are uh you know selling hope and like it's a confidence scheme right but and I think this is why like no I'm not out here saying hey I have this magic AI tool go buy it for $300 and you will get amazing returns. I'm saying I have basically essentially a very useful public white paper that's completely free and that you can go and do whatever you want with even if you get one idea from it. It's it's going to be helpful. But I think there's going to be a lot of other people who are going to try to make money not investing using AI but taking advantage of the gullible by selling them some you know pipe dream um that AI is going to be a magic solution. And I think you have to be careful. You know, you don't want to fall for that. >> Yeah. Human beings are vulnerable to that type of an appeal. They want to hear that there's a button they can press and and and they and and if you're um if you're a novice, let's just say, or just really don't know anything about how investing works, you may think that there is some way to just kind of um figure it all out so that you never ever lose money and you're never wrong and you're always making lots of money and that just doesn't exist. >> Yeah. [laughter] >> And that's how you lose money. Yeah. >> Yeah. No, >> thinking that is how you That's right. Yeah. [snorts] >> And I, you know, it's funny. I actually saw a cartoon post on LinkedIn a few weeks ago, you know, and and in the first like it's like kind of two scenes, right? In the first kind of panel, uh this guy, you know, P gives the interviewee the pen says, "Sell me this pen." You know, you probably all seen those like, "Sell me, uh uh this pen and you know, kind of tests, right?" And the se and the second p uh p p p p p p p p p p p p p p p p p p p p panel the guy just uh the interviewer says it's aic you know and it's such a you know you know for those you know like you know it's a right the reference to like the hype around AI agents and all right so uh you know so it used to be social proof it used to be I don't know Matt Damon you used it to sign his latest contract you that that used to be like the right answer now it's agentic right you know is like the the punch line and and it's so true right there's so much hype and so much I mean I think it's good that we are experimenting as a society and as a community and investors and even beyond investors um but I think that only a small fraction of the things we're working on are going to actually work out and you know it's important to be realistic about what AI can and cannot do at given point in time. >> It is an exciting time isn't it? Um, on the one hand, uh, lots of people like you and and and us are excited about how to use this new tool. And on the other hand, um given the business that we're that we're in, you know, you and me and and overwhelmingly every guest we ever interview, um we can't help noticing that um lots of people are investing a lot of capital in um data centers that are um I previously represented them as being highly utilized. the load factors, the power usage is is great, but the server utilization is not mostly, as far as I can tell, under 20%. So, the moment strikes me as more similar to the buildout of fiber in the late 1990s, early 2000s than I ever thought, you know, with when there was a lot of dark so-called dark fiber. So there's a lot of dark dark server capacity and and no return in sight on an investments approaching trillions. You know, there there's going to be this trillion dollar multi-t trillion dollar asset out there if they keep going the way they're going and building the way they're building. Um and [clears throat] it's uh it's just I don't know how to think. Like we just spoke with Ben Hunt and we were saying, "Boy, Ben, it sounds like you don't think 2026 is going to be a very good year." and and um it's I don't know do do you just now that we've talked about AI there you know let's talk about it as investors like do you do you have a macro view do you have a view on how this is going to play out in the next 6 12 18 24 months do you indulge that sort of thinking >> yeah no I hear you I I'm a much more of a bottomup investor but listen I think about these things as well but and I think like the Gartner hype cycle kind of framework is perfect for this, right? Because I think like we've seen this so many times where there is a promising technology. It is capable of doing a lot of new things, but then the expectations for what it actually can do and how quickly it can get those things done, you know, is know it's go it they're way over overhyped, right? And I think that the some of the studies I've seen like the amount of extra revenues you need um you know to actually generate a return on the trillions of investments that are going into AI. Those are huge numbers. And like and I know like Elon Musk is out there saying there's going to be an army of robots and uh 10 years from now, you know, no one's going to need to work or something like that. But you know, it's Elon. He likes he like he's an amazing creator of wealth but he's also likes to fantasize about the future and you know a lot of his forecasts don't quite come true. So I don't know. I mean I also know the reason I'm sure you've been following the Michael Bur um saga with you know his you know related to your capex question is that these companies are like manipulating their earnings like they're of balance sheet debt their depreciation is you know insufficient. So there's a lot of froth in the market. I think that there's a lot of gullible people who frankly you know the last 15 years we haven't had a real bare market like we had many corrections where the government quickly stepped in and flooded the market with liquidity and so there's a whole generation of investors who like they know the stocks can go down but they in the back of their minds they expect that dip buyers will come and they will buy all the stocks and it'll be all okay and that might be true but as a student of financial history and markets like we you and I both know there are like decades where things were not that great right and we just happened to be exiting a decade and a half where the markets have been great and very benign and forgiving so I think you combine that with the AI hype and despite the fact that AI is quite real as I I think we spend most of our conversation talking about I think at the end of the day I think there's a lot of danger in the markets and I am very nervous that asset prices almost across the board are very frothy uh or at least very full maybe not frothy in every you know corner of the market but certainly not a lot of distress or cheapness going around and expectations are built for perfection but you know what do I know at least that's my view they are built for perfection yeah um that's but it's funny though because given the um generally you know higher market multiples and things that have prevailed really this century this century really post you know post 2000 um you you could have said that and I did say it I have said it many times you know so it's I guess I'm kind of I want my listener to know [clears throat] and and if you disagree by all means weigh in like noting that something is priced for perfection is not a timing call Gary is not calling the top. >> That's why I hed in the beginning like valuation is probably the worst, you know, indicator of what's going to happen in the next 6 to 12 months. >> Yeah. I just want to >> land on long-term expectations. That's for sure. And I think that if you if you're listening to this and you think, "Oh, I heard markets return 10% per year." And by the way, recently that's been more than that. And you think that that's like, you know, guaranteed from this starting point. I would I would go and recheck your assumptions because you know that I don't think that that's likely to happen for a while. >> Yeah. Long long term um from this moment lots of people have studied it. You know if you go back in history from this kind of a price for perfection moment um the S&P 500 has generally done really poorly and been you know flat or even down over many years like a decade 12 years numbers like that. And again, I'm not forecasting that, but I am cautious. I can tell you as a bottom up investor, I'm having trouble finding new ideas. You know, it's not easy. And I know I'm not supposed to say that because >> old investor, you know, most investors are like, "Oh, I can find amazing ideas in any market. You know, just give me some money." I rather be honest and say it's tough. If you want quality and if you want a good price and you want a business you understand and you want there to be something misunderstood, mispriced, then that that's not big opportunity set right now. At least not one that I'm finding anyway. >> Right. We um in in our extreme value newsletter, Mike Barrett and I have done really well over the past five years. Um but you know, just like everybody else, right? >> Um wow. >> And and I mean much better than the market, you know. >> Uh many many you know tens of basis points more just blow it out but we feel like we're struggling every month now it's it's really we're going ohh god you know can we really recommend this with a straight face yeah we can because of this and this and I wonder when we're going to just say can't do it we we just can't do it one more month you know [laughter] but it is difficult >> I I think your honesty is refreshing right because I think Again, I think it's important to be honest both with yourself and with others. And I think look, I always tell my partners, you know, it's not about how the portfolio, how attractive the portfolio is today because what I'm doing is I'm sitting there day in and out and I'm hunting for opportunities and maybe to bring it back full circle, you know, using both AI and other methods as well. I haven't given up on, you know, you know, the the good old just, you know, hard work and elbow grease, right? Um, and I think the opportunity set will change. And I think if you're being properly cautious right now in how you're investing, that's going to pay dividends when you are able to take advantage of the more attractive opportunity when a lot of people maybe they're not. >> All right, Gary, it's time for our final question, which is the same for every guest, no matter what the topic. Um, you've answered it before. I kind of hope you forgot because I think it works better that way. [laughter] >> No, I'm good. No, that's good. No, don't be nervous. I I'll I'll tell you. The question is this. It's it's for our listener's sake um to give them a takeaway. So, if there's one thought or one takeaway that you would like our listener to have, um what would that be? >> Yeah. Well, since it's uh since the main topic of the conversation is AI, I would say whether you download my uh framework or not or do whatever you do, start using it uh for investing now, at least part of your process and start learning how you can build that for you in a way that's right for you. So, if you uh walk away with nothing else is that just getting the reps in on building AI into your investing is going to be something you're going to be grateful for for years to come. >> Thank you for that and thanks for being here, Gary. Uh it was really great to talk to you. >> Thank you for having me back. I appreciate it. Folks, the White House is on a stock buying spree that is comparable to a hedge fund that has raised billions. And the stocks in question are surging as a direct result. 200% within 24 hours in one case, 90% in another, and that's just the tip of the iceberg. Which is why we're hosting an urgent financial summit, the stocks that save America, on Tuesday, November 18th. That's tomorrow. This event will feature a man you may know well, 50-year investing titan Rick Rule, along with my guest today, Nick Hajj. Together, they'll reveal why the success of some stocks could now be a matter of national security, how a select group of US companies could soon play a historic role in rebuilding the nation's industrial backbone and the wealth of regular people, and the name and ticker of a free stock recommendation they believe could soar as this story accelerates. A stock I personally love, by the way. If you enjoyed today's episode, this briefing will be a mustsee followup. Now, head over to www.whousetocks.com to learn more and reserve your spot. Again, that's www.whoustocks.com. Don't delay. Well, it was great to talk with Gary and it was great to hear somebody sort of get into it um with using AI in an investing process and I'm so glad that he has this PDF. I have not read it yet. He just gave us the link like shortly before we did the uh did the recording here and I have not gotten into it. I I am going to tear that thing apart because I'm struggling right now to sort of figure out AI in my investment process. I use I use a lot of queries, but they tend to be kind of general and I wind up having to do a lot of work to get the garbage out of it. Um, but that was really, you know, as always, Gary's a very thoughtful processoriented guy. So, that was perfect, I thought, and perfect for this moment. >> Yeah, very uh a great resource uh that he's put together. It seems like I I haven't uh checked it out myself yet either, but I will. And it's um just reminds me of kind of how I'm trying to think about all these AI tools that are out there now. Um about really looking at it as like an assistant or um uh like a kind of a entrylevel uh assistant job uh sort of role. uh which is which is good news if you have a job already and are kind of mid or or upper you know career you could you can afford to do this and put these uh let these tools help you. We're seeing right now also that's a problem for people who are trying to find kind of entry level jobs in kind of the knowledge uh economy uh you know maybe junior level analysts and and those sorts of things. Um, but you know, there will be I there's got to be ways to to use it uh from an like entry level perspective as well too. Um, and that'll work itself out over time. But, uh, what what what Gary was talking about is super interesting practical ways of putting this technology to work. I loved his his thing about if you if you didn't know anything that he was saying his points about prompts and asking the the models questions and then notebook LM um just kind of a tool where you could dump a ton of information like the information that you want and uh it's like a second brain you know it's like all the stuff you you wanted to ever remember uh you could dump into like a single folder essentially and have the AI you know pull pieces of information or analysis from it. So, um those two things alone are super valuable to do if you're trying to figure this stuff out, >> right? Um I just realized as you're speaking, you and I, um you of course you write the Stanbury Digest, you know, virtually every day and I do it once a week usually. Um, and at Stanbury generally, even in our newsletters when we make recommendations, we're mostly targeting uh, and we're overwhelmingly our subscribers are self-directed individuals um, who are managing their own money. There are plenty of investment professionals running other people's money among our subscribers, but overwhelmingly um, you know, it's the self-directed individual. So we do a lot of storytelling and we try to get our analysis um you know we we don't just throw a lot of numbers at people. We try to do the analysis and then [clears throat] make it far easier to understand the thesis than a Wall Street analyst might make it. And that storytelling we do I think is really important. I have found that I use AI and I do like general, you know, throw it out to the internet kind of queries to hunt for this stuff. I'm using it to find the story almost more than to do analysis more definitely more at this point. And I'm wondering you you you do a lot of that same kind of storytelling every day of the week just about. Uh, and I'm wondering what your, you know, specific use case looks like. >> That's interesting you say that. I I I haven't really used it at all to to to find stories or narratives. I did early on. This is probably at least over a year ago. Early on, I was like, "All right, let me see what this if how I could incorporate this to daily, you know, the daily work." And >> actually, let me let me um qualify. It's not like I use it to find the story or the narrative, but I [clears throat] definitely use it to hunt down as many sort of facts and details, right, >> to amplify the narrative, right? So, just to just to clarify. >> Yeah. Yeah. Yeah. I get it. And and early on, you know, I was I was like curious, okay, how how great can this can uh you know, this a AI search be essentially? So, you know, I'd say, "What are the top what are the top five stories people are talking about in the stock market today?" And, you know, it'd give you some vague representation of it. Uh, which wasn't very helpful for me at all. Um now so so I use it more as from like a writing editing perspective uh on as like I was saying as like an assistant like um you know you would have we have editors too um and we will but to uh have something written and and then have an AI look at it uh and may be able to fill in some gaps or tell you if you're missing something or um did you think about this kind of what Gary was saying about you know towards the end uh you just kind of yeah use it as a filter. Um so um that's really mostly um you know from from a daily perspective it's different because or for me I feel like because there's so much happening every day that a lot of times for me it's just easier I'm still at the point where it's easier for me to think with my own brain and just put it on the page and instead of that the interstep of going to the going to a technology to tell me what's going on rather than me just doing it myself. You know, I'm still at that point with a lot of a lot of this. Um, if that makes any sense. But, um, I definitely see the the the value in adding it to a workflow um to, uh, just kind of do do the do things that you're not necessarily mundane tasks, but stuff that's like been routine um, over time. And just once you start using this thing, I think you even start realizing, oh, I don't need to do that anymore. And that's okay. Like I'll I don't need I don't need it anymore. And and then you just do other stuff. And then you'll just do other things. You won't even be realizing it. >> Interesting. Okay. Well, I just wanted to see where you were. Um and I'm sure it'll be different a year from now, right? >> Oh, yeah. Yeah. Probably be different in a couple days, but [laughter] >> Yeah, that's right. Yeah. Me, too. Um, but one thing I definitely am going to do is is learn how to use Notebook LM. Um, which I really haven't haven't truly done yet. Um, to focus just, you know, so much of what we've done certainly in Extreme Value and in other publications focus on focuses on really high quality sources, 10Ks, company reports, you know, on the ground due diligence, etc., etc. And um doing that faster is like a dream come true and that's what you could do with with Notebook LM. Anyway, that was awesome. I hope everyone else got lots out of it and took some good notes and we'll download Gary's PDF and take a look at it. Um it was a really great fun interview and a great episode. I hope you enjoyed it as much as we really really truly did. Opinions expressed on this program are solely those of the contributor and do not necessarily reflect the opinions of Stanbury Research, its parent company or affiliates.
AI Is the Edge You Can't Afford to Ignore
Summary
Transcript
Hello and welcome to the Stansbury Investor Hour. I'm Dan Ferrris. I'm the editor of Extreme Value and the Ferris Report, both published by Stanberry Research. And I'm Cory McGlaclin, editor of the Stanberry Daily Digest. Today we talk with Gary Murus, chief investment officer of Silver Ring Value Partners. Gary is an old friend and he is a great investor and a very smart guy and he's figuring out how to use AI to do investing to analyze stocks and make recommendations and do research and we will give you a link to download his PDF so that you can see what he's up to and learn how to do it too. So let's talk with Gary. Let's do it right now. Let's talk with Gary Mashurus. Gary, welcome back to the show. Always good to see you. Likewise, thank you very much for having me. >> I was really happy that you reached out to us um with a topic that of course everybody's fairly well obsessed with these days and I thought, "Wow, this is exactly like you picked the exact perfect moment uh for this." So, so good on you. I mean, I know you're not in a marketing business, but if you were, this would be the this would be the time for that message. And the message um just for our listeners benefit, Gary reached out and said, "Hey, I have a process that uses AI, you know, for my investment analysis and I want to share it with you." So I thought yes, because I'm trying to do that myself and I haven't necessarily come up with anything really super organized. So I'm I'm a listener today as much as a host. So, um, >> well, I think we're going to have a good conversation. It's a good topic that I think everyone is struggling with right now. >> Yeah. So, I I feel like I should put it in your hands. Um, wherever you want to start with this, um, you know, your your AI investment analysis process, just go for it. >> Yeah. No, absolutely. I mean, I think look for some background, right? you know, when I started as Fidelity as a young analyst 25 years ago, um, you know, Excel was a pretty common tool, but you had, you know, some grizzled veteran portfolio managers who were still using the legal pad, right? And I literally remember as a young analyst seeing them do modeling on a yellow pad. And I think pretty soon, if not already, if you're not using AI, that's what it's going to feel like. meaning that it's maybe hardcore that you're doing these things manually, but I think that it's not optimal, right? And I think there are two extremes like people are people are going to either the extreme of hey AI is going to solve everything. It's going to be a genie in a bottle and it's going to give you like these amazing stock pickicss and you know all you have to do is figure out the right magic prompt and rob it the right way and the genie will come out, right? And then there's the other extreme of like it's all hype, it doesn't work. Real investors don't use AI. They just sit there like Warren Buffett and you know read 10Ks ideally physically printed, not even in, you know, digital form, right? Um so in somewhere in the middle, I think there's an actual opportunity to add value without sacrificing quality. And that's I think what I've been working towards. >> It's ultimately a tool. So, you know, if you don't have a hammer available to you and someone gives you one, um, it feels great and it and it makes putting nails into wood a lot easier, right? >> But, uh, you know, it doesn't actually it doesn't really solve all your problems. It doesn't tell you how to design a house that won't fall down, etc., etc. So, >> um, yeah, it's a great tool. Yeah. >> And and I think, look, there's two [clears throat] types of use cases. two as I call them category one and category two and I think category one is basically something where you're trying to save resources and that could be primarily time could be money but you're saving you're doing the same thing you could have done as a human you're just doing it faster cheaper some using less of something right and that's probably the most common and then there is this elusive what I call category two and category two is you're actually doing stuff with AI that you couldn't do as a human or couldn't do nearly as well even if you had nearly unlimited resources. And I think that's where maybe the holy grail is down the road. I think most of the use cases I've been able to figure out right now are category one. But I think I have one one or two use cases which kind of enters category two. And I think you you mentioned the framework. I mean, so I put together this AI equity analyst framework that um you know, I think, you know, I've made freely available partly for selfish reasons. Like I want people to read it and tell me either what I'm missing, where I'm wrong, or what I can do even better because this is the wild wild west. But that being said, I think I have a pretty good starting point, which is what I think it is. It's not the final answer to anything. It's a working kind of a framework that I actually use dayto-day because it's just amazing once you roll up your sleeves how much AI can improve the process of a long-term serious investor. >> Amen to that. Um even just uh for me just like as a search tool through multiple 10Ks and stuff, it's that it's unbeatable, >> right? If you just feed it 10ks, if all you ever do, you know, at this point for me is feed the thing 10ks and use it for search for various, you know, prompts. >> That alone and that's not wrong. I think that's a great use case for sure. I just think there's so much more. And I think, you know, >> if you I think that actually even though it's just a step back, the way you use AI right now is actually probably of secondary importance to like start using it. So if you're listening to this right now, like just roll up your sleeves and start building it into your process because if you don't you wait for some iteration of Chad GPT17 or Gemini 12 or whatever you you're not going to have the experience. It takes reps. It takes like you have to kind of do trial and error because by the way there are some pretty atrocious results obtainable with AI as well. you know I you know so in this framework in the AI equity analyst I've tested each prompt and some of the versions I had to go through to get to a what I think is a working version they were atrocious you know and some of the use cases AI is just not ready for but you're only going to find that out if you start doing it now but and if you do you're going to start layering AI into your process and I think that could be different for you it might be know a lot more than me or a lot less. But I think start doing it is the main message. I would say even more important than some clever prompt that prompts AI just the right way or something like that. >> I'm really glad you said that. Yeah, I'm I'm I'm glad you said that too because the conversations we've been having lately or even I've been just been thinking about about AI now now compared to say even a year or two years ago when it was what is this how is this going to change the world and take everybody's job now some people still you know think that last part but um the conversation now is just more practical like what you're saying you could have more practical conversation about how does this actually work uh what are the benefits etc if you start trying it is the key thing and so I'm glad you brought that point up to the table >> yeah no absolutely and I would say again there's a lot of focus on prompt engineering and as I just mentioned it can be important like it can definitely make a you know a bad use case good but I think and this is my you know I don't have any data to prove it but in my subjective experience so far like having the right use case so asking the right question is like 80%. And then figuring out how exactly to prompt AI for that question is data 20 because I think that you know like some of my best aha moments with AI in the last six months and I've literally spent like over a 100 hours building this out um because it's the return on that time is just so huge. Like I invested, you know, 100 200 hours but I'm going to save hundreds of hours per year. So the ROI is incredible. But like some of my biggest aha moments were like, "Wait, what if I use it for this?" Uh, and and then I would have like this blocking. Oh no, no, it's going to be terrible at this. No, I'm not going to do it. Well, let me try. And maybe at first it's terrible or maybe it's 80% terrible, but there's a 20% useful component. And and this is another very important point. I think people like talk about AI hallucinating, right? or you know AI giving you the wrong numbers and then they use it as a reason to not use it. I think that that is actually like completely backwards. You have to figure out where is it okay to have hallucinations like where is it okay to have wrong answers. Let me give you a very straightforward example. So at the top of my investing funnel you know I have idea generation right? So, I'm looking for potentially attractive investment candidates. And it actually does not matter if the list AI generates uh and I have two prompts in the framework, you know, special situations and a kind of a Philisher compounder prompt that I put together and I kind of test it out. It doesn't matter if gives if it gives me 20 ideas and 16 of them are terrible. Like what matters is are there four good ones in there, right? Because I'm not saying you want to delegate everything to AI and go to the beach and come back and see how the portfolio is doing. That's not it, you know. I'm saying you use it as a tool. You stay engaged with it and you basically go and say, "Okay, now I have these 20 candidates that AI generated that I would not have had." And again, using both my own, you know, knowledge and also other AI tools saying, "Okay, go through one by one. Garbage, garbage, garbage. Oh, interesting. garbage interesting interesting maybe you know okay and now I have generated something I wouldn't have had even though m most of the output might have been not great so that's an important point I would say >> yeah it's a it's a learning to use it is a process of trial and error that I agree and it's worth it and there's no other way to go about it you're riding a bicycle you're going to get on the thing and you might fall and that's the way it is um I'm glad you said that it's I'm glad you emphasized early in early in what you just said um that asking questions and and you know querying prompting is is really important because finally um it there's finally a place in the world for somebody who might not be able to write code and might not be a highly technical individual but who has a lot of questions about what they're doing and and let's face it if you're looking as we are, as you are um in the in our somewhat different roles, investment ideas, you have a lot of questions all the time and you want to know, you know, the best company in this or that industry or the company that has the least debt in this, you know, industry or anything like that and much more that's much more complicated than that. Um, you know, it it's a it's just a wonderful it's a wonderful thing I think that's happened to us that we have so you know this resource that we can just ask questions to without needing a degree in in you know computer science to do it. Um, >> yeah. No, absolutely. And I think it's funny you mention it because I have my my teing here. My I do have computer science degree from back a quarter century ago, but I don't think it's useful because it teaches me how to code something. it I think you know and I think this is an important insight right people think that okay AI maybe will flatten kind of completely level the playing field and the return to skill will go away because everything will everyone will have access to these like you know genius AI systems that they can just replicate that >> no that's not what we're saying at all >> yeah in the driver's seat >> I think the return on the real skill and insight is going to get magnified by AI AI, right? Exactly. Because >> if you can add, you know, what what is going to go away is the grunt work required or at least a chunk of the grunt work required to execute you know uh on those insights. But if you imagine if you could you know think about you know yourself as okay you have the true value adding things you do whatever they are might again different for you than for me and different for you to listen to them for me as well but you know you have them and then there's all the blocking and tackling to actually make those those insights useful and make them get into your investment portfolio right and so now you can just leverage those insights so much more And you you can skip some of those steps, but again, you're not and this I call it when I wrote the framework, you know, there's like a big orange warning on like page two or something of the PDF, which is this is not a shortcut. This is not meant if you're lazy and you just want again that made magic genie in a bottle, this is not it. What this is is it forces you to be explicit about your process and allows you to do more quality work faster, not just cut corners and hope for the best, which I think is an important distinction. >> Right. Right. So, I'm glad we agree that, you know, the the the human being is in the driver's seat. This idea that AI makes, you know, our humanity like redundant is ridiculous. All right. So, shall we? We should get into this a little bit. And um I'm sure our listeners like, "Okay, what how does how does Gary really use this? What is he doing? What's the secret?" And you have a you have a um a document that you've created that that tells this information, but but we'll, you know, we'll we'll introduce the listener to that. Okay. So, where where would you where should we begin? Where where do you begin? You said idea generation. Um >> yeah I mean no absolutely and I think idea generation is a great use case because when you think about what is AI good at one of the things it's good at is synthesizing a lot of information right rapidly and so I think that so I have two prompts you know so in that PDF you mentioned I have 13 well tested prompts that that work you know and um you know two of those prompts are idea generation prompts so you start at the top of your beginning of your research funnel Right? And you say, "Okay, what candidates should I consider?" And one of them is a Phil Fischer prompt. So Phil Fischer, famous, you know, growth investor, you know, I would think of him as a just a long-term intrinsic value investor with a focus on growing companies, but he's known as a growth investor. So I'm not going to quibble on that. And you basically tell AI, okay, act like Phil Fischer. You know this is not the exact wrong but go and find ideas within certain parameters. Let's say I want certain countries certain market cap ranges what have you and have that and and give me a list of things that fit. And Bill Fischer by the way in his book common stocks uncommon profits had these 15 points kind of a checklist. So you're basically having AI run through the entire market or some subset of the market and have it judge each of these companies on on that those 15 points. Um you're going to get some list. It's going to take a long time. By the way, this is not like, you know, you're going to like you're going to go, you know, put it in. >> Yeah. >> Yeah. you it's going to be it's not going to be instantaneous but you're going to get a list and you're going to go through that list and you may find some hidden gems and I don't know New Zealand that you never would have heard of otherwise so that's qu a very cool use case um again a lot of human oversight you don't want to assume that it's right some of them you'll be like huh really and there's the I the intelligence in AI wait that doesn't like this can't possibly be a fit for what I'm looking for but ignore that don't worry about it because you only care about the ones that are interesting, not the mistakes. And a completely different approach, you know, like let's say you are like into just deep value. You want, you know, special situations, things like that, which I know I love, a lot of my investor friends do. And people like there are shops that have a full-time special sits analyst. All they do is find and kind of prep these ideas for the PM to look at, right? Well, you can go and have AI do that for you. like find all the special situations, spin-offs, you know, restructurings, whatever. You know, we don't need to get into the the nitty-gritty, but the point is, you know, come up and run that monthly. You can run or you can even create, you know, a task to have it update you for uh when new ones come. So, I think at the top of the process, the sky is the limit, right? And I think I think that is just a terrific use case where again, not saying that that's the only way you should generate ideas. I think there are other ways as well, but I I found these two ways to be pretty pretty cool to be honest. >> I find it pretty cool, too, man. [laughter] >> So, I agree. That's good. >> I'm glad you mentioned Phil Fischer, too. >> Yeah. And and that's my point, too, is that this approach is really agnostic to how you invest. Like, it really doesn't matter if you're looking for, you know, completely different set of things that make an investment attractive to you than would make it attractive to me. Like that's what I mean make AI your own like it's there to serve you like execute your process not my process you know and not someone else's process but another question is okay so you find this list and what do you do with this you know okay and you can go hardcore go to the library put on your favorite music and your headphones and start reading 10ks but that is and there's a place for that I think still but I think it's too early you're getting this giant list you're pairing it down to some subset And now you want to go through it and figure out which companies are actually worth the effort. And so I have a whole bunch of prompts from, you know, basic company history and description to like put, you know, your Michael Hoer structural analysis head on and analyze the quality of the business. All kinds of stuff that the main purpose is for you to get to a quick note, right? because you get this giant list and you don't really want to spend your manual labor hours equally on every company on that list. That would be foolish. So you what you do is you essentially use AI to come give you enough analysis to say no no no maybe ooh this is interesting. And now when you say, "Oo, this is interesting." This is not, "Oo, I'm going to buy some." This is, "Ooh, that's when I'm willing to spend the human research uh time and go deeper." >> That I want I want our listeners to know that that moment right there that Gary just described is it's the one I don't know how he feels about it, but I feel strongly about it. That is the moment you pray for every day. Oh, this is interesting because so much of the time you say no, no, no, too much debt, commodity business, no moat, whatever, you know, whatever it is you're looking for. Like you said, it can be any any investment style. But [clears throat] that moment when you go, oh, this is worth looking for. There's something happens where you're confident that even if you wind up saying no, the next hour or two of your time is going to be really well spent. >> Yeah. And and by the way, again, I've done this for a quarter of a century now. And even though I'm a process guy and I'm very like rational, methodical, I've just learned to trust my gut a little bit. It's okay. Sure. And you get this feeling like you have to put every at least I do. I have to put everything else aside and start researching this company. That's a good feeling. That's how you know. >> Yeah. Yeah. >> That's what I'm talking about, man. It's like finally I because otherwise >> um >> I don't know how you feel about this but but I hate the feeling of being unfocused. Sometimes I just push back from my desk and say, "Whoa, whoa, I need to I need to step away." Then I sit down and I write with a pen on a piece of paper what I'm doing next to kind of refocus because that focus, >> man, focusing on something and really putting your aiming your your power at aiming your highowered rifle of your mind at it. Um hopefully with a good accurate scope and [laughter] a good marksmanship um is is what we really live for. Um we analysts, we equity analysts. Um so beyond so so here we are, we are we have you know we've gotten into idea generation. We're now at that moment where we say hey this is interesting. Um and and then you're talking about trusting your gut and you know reading the 10Ks and all that. So what you know what's left? What's left for an AI? >> So there's one more step. There's one more step, you know. Yeah. And I think there's a tool. I'm sure you've heard of it, but you know, I don't know how many people have actually used it. Seriously. And I honestly think it's the most powerful free AI tool like bar none. Like, and it's not Chad GPT. It's not Claude. Um, it's not even Gemini. It's Google's notebook LM. >> And if you >> You're the second person to tell me that. >> Yes. I'll introduce you to the first. I want to get to know them. you know, >> it's our our research director, Matt Wine Shank, told me. >> There we go. >> But I think it's so important because I've had all these companies pitch me on their AI products and, you know, I have nothing bad to say about any of them. I might use some of them down the road. Um, but notebook LM totally free. Number one. Number two, it it basically So, if you don't know how AI works, right, just like a step back, there is something called the context, right? An analogy. So AI the reason like a lot of AI so large language models hallucinate is they have this this basically they sucked in hoovered up the internet for some portion thereof into their context and then when you type in you know you adding context and they're basically kind of combining those two and you know finding the answer for you I'm simplifying there other steps there's thinking models and all that so nom is like a blank slate the only context it has are the source sources that you upload and you can upload up to 300 sources and so I'll tell you exactly how I use it. So for example, you uh I start with downloading you know let's say 20 annual reports if they're available PDFs are fine. You throw them into your notebook. So nobook alm you create a new notebook again it it has not sucked in the internet that its knowledge is limited starts at zero literally starts at zero and then you pop you tell it what you wanted to learn. So now rather than prompt engineering, you're context engineering. you're engineering what context you want the LLM to assess and you let's say you put these 10 annual reports or 20 annual reports and then you ask it okay first you can just have it say tell me the story you know tell me a quick story you know of how this company you know evolved over 20 years right by the way if you really want to save time you can create a podcast you know speaking of podcasts right you can create a a podcast with two hosts uh using just information that you've uploaded and go to the gym, take a walk, whatever, and then come back and you'll have no know learn a bunch of stuff about this company. There's another level to the game here, which is I mentioned earlier on that there are some of these category 2 uses like things that AI might be able to do that you and I can't, right? No matter how much time and money we throw at it. And so what I've done is, you know, so I use expert interviews as part of my research process. So as I go deeper into a company, I don't want to take management's word for it. I want to go ahead and say, okay, what are other people saying, customers, former employees, suppliers, and so forth, right? So let's say you have access to an expert network, and you can download the transcripts, and you add those to notebook. And I have a really cool prompt in that AI equity analyst framework that I mentioned which actually I think is a kind of a category two prompt in the sense that what's happening there is you're asking it to find common themes and trends among the you know among among the interviews and let's say if you are had a 500 you know 50 transcripts or something like that it would take you a long time but even then you you might not connect the dots like oh this happened here and that happened here. This former employee said that and the customer said that. So that amazing use case. And so when I was a young analyst at Fidelity, they brought in CIA former interrogators and they kind of try to teach us how do you interview uh people? How do you detect if someone is lying to you? And one of the things that always kind stuck in my mind is that you know they told us when people lie they don't actually tell you a complete falsehood usually in in relation to uh the question you're asking. What they frequently do is they will answer a slightly different question. So let me give you an example. Let's say you're interviewing a former you know employee of a company you're researching and you say hey how's the culture? Right? and and the person says something like, I don't know, you know, I really like the guys. We really enjoy going out for beers after work, right? So, and you might just move on and feel like no cognitive dissonance whatsoever because you feel like you gotten an answer and you got a good answer. But if you pause for a little bit, you might realize, wait, that that wasn't a answer to my question. I didn't ask you if you like guys the guys you work with or if you went out for beers. I asked you how the culture is because maybe you're going out for beers with the guys you like to and moan about like how bad the culture is, right? You know, you know, and so AI has this amazing way of, you know, and the prompt literally tells AI find things that were unsaid that are literally between the lines. And I think it's it I've seen it tease out some pretty neat things in my experiments. Um, and I think it's an awesome use case that's perfect for the tools that are available right now. Oh yeah, I love that idea. I love the idea of, you know, feeding it conference call transcripts and what aren't they telling us? >> Yep. Exactly. Omission. I mean, so I mean, I think om omission uh m uh stuff is also a big way people lie, right? You know, they they tell you part of the truth but not the whole truth, you know. And uh I mean man, you know, I was just talking to a former I IR officer of a a large company that we've known each other for a while. And I'm not going to mention the company, but the things that the person told me, you know, the things you don't know as on the outside of a public company that are going on on the inside sometime sometimes that's a big it's a big category of things, you know, let let me tell you it's and so I think that especially look as a concentrated investor which I am, you know, if you have like 20 basis points in each idea, maybe you don't care, maybe you investing based on no I don't know quality factors, that's a different game than what I do. But if you're a concentrated long-term investor like myself, then having like a big blow up can be very costly. And I'm not saying this will prevent all of them, but if you can reduce them, that is very valuable, you know, at least to me. >> Absolutely. Um, [clears throat] so like how is there a point in your process, Gary, where um beyond which AI just simply, you know, you've you've still got to, you know, tough it out the old school way. >> Yeah, absolutely. I mean, I don't want AI to do no the thinking for me. So once I kind of sunk my teeth into an idea, I do still do all the deep research. I I do read now some I do faster, you know, uh but I still do it. um I still come up with my own range of values. I still make the decision and there's actually a uh so all of that is me and in that you know in that framework I literally have a flowchart which has steps in the process and uh you know blue steps are AI blue and orange steps are combination of human AI and orange is purely human and what you'll see as you follow go further and further through the process is that at the beginning of the process there's a lot more blue or blue and orange towards the end it's mostly orange it's mostly me right but that way I can do that mostly orange the human part on a lot more ideas >> so this is a very interesting topic because I've seen studies where um medical diagnoses were done with AI only human only and then the combination and the AI outperformed the other two so you know um a medical diagnosis is not an investment decision you I get the But uh you know you understand why it's an interesting question at least right? >> No for sure. I mean and by the way listen maybe that's where we're going right maybe in some number of years you know AI will be hiring me you know and AI will be say Gary this is what I need you to do today. I will make I will I got this right. So you know and by the way I've seen some LinkedIn posts where like they show the org chart of a company and all the seuite is AI you know CMO is an agent you know you know right so you've probably seen those too but I think today we're not there I also frankly I'm a little bit old school like I know what the studies might say but I don't know when there'll be some glitch when there will be some issue when there will be something right um and so may I know again I you know see this these gray hairs you know I've gotten you know through the not hard knocks you know in the markets over the last quarter century and I just feel like when the people are trusting me with their money they're not trusting me to go to the beach and turn on some AI algorithm they want me to actually make the decision so I think it's super important to I mean different people will come out come out at different places on this but for me I bring actually AI in the end and let me tell you how I do it because I think I think it's a very important. So you probably know that I'm a big fan of behavioral financing. My substack is be called behavioral value investor for that reason. I think like essentially in investing we are frequently our own worst enemies. So there's two more steps at the end that you can do. No. So you you've done all that you know early stuff then you've done the d uh deep digging and then you end up um you're going to pull the trigger right? Well no wrong. You not you don't pull the trigger. You do two things. number one. And so you write up your thesis uh and you put it into AI and you have it check it like you know check the logic check I'm not talking about spellch checkcking or grammar here you know talking about you know are there mistakes of omission you know is it internally consistent for instance maybe you say at the beginning I think investment XYZ is great because of A B and C but then you only go and show A and B or maybe there is evidence in your own report that C is untrue right So things like that I think is a very easy and useful way um to find to have AI point out mistakes before you actually put money to work behind them. Right? The other thing is something I call the devil's advocate. And this is an idea I had back in the day. We got a group of, you know, grizzled investors together, friends of mine, and I got them together and I said, "Okay guys, we, you know, we all know behavioral biases are real. We all know we make mistakes. So, how about we kind of create a pack, a little club where once a year you'll be asked to spend a couple of weeks, you know, coming up with a serious opposite case, the devil's advocate case based on the information someone sends you on the company and and your and your work buys you that once a year you can ask, you know, uh, someone else to do that for you. And it worked probably like six months, right? And then people got busy. I mean, listen, it's a heavy lift to ask someone to spend a couple of weeks on an idea they don't truly care about, right? It's like we're human, right? You know, it's like I talk to college students, what I teach at BAPS and I talk to other groups about investing. I always have the these slides in my deck where, you know, about comparing investing to dieting, right? So, you have, you know, I have one slide with this really fit young woman and with healthy foods and exercise and which is what we know we should be doing. And then there is the next slide which is this big fat guy chomping on a donut which is what happens in reality right and so like investing is a little bit like that like there's what we wish we were doing ideally and then there's like the reality hopefully it's not as bad as that but you know you I think you get the gist and so now I have this AI devil's advocate prompt where its goal now is not to stay limited to just my report its goal is to go and be this hunter killer, you know, of my thesis and help me find things that are missing like, you know, obvious common bias called confirmation bias. What is it? You kind of seek out things that agree with your already predetermined conclusion and you kind of ignore things that go against it, right? We all do it except for me. No, I'm just kidding. We all do it. That's the whole point, >> you know? So, so I think AI is amazing at that. So you start with AI, you use it a little bit less as you go deeper and deeper into your idea and then you do the pure human part at the tail end. You kind of use AI to help you avoid mistakes. So that's kind of my process. I mean I'm, you know, if I'm sure there are people who have, you know, found things I haven't. Again, it's the wild west, but that's part of the reason I'm sharing this because it's so exciting and I think there's so much we have to gain from each other in terms of collaborating and looking at how different people are doing this because we can only get better together. That's the only outcome. >> Maybe before we go any further, we should we should tell our listener how they can get to this document of yours because, you know, you want everybody to read it, right? >> Yeah. No, I mean, I think I think it's worth I think it's a it's a it's a good one. You know, I think I've had a lot of positive feedback and I've had people kind of, you know, give me some good push back on some of these things and at some point there'll be a version two that will be even better because of it. But I'll I'll share a link with you that you'll uh you'll have. It's just a, you know, it's a 33page PDF where I kind of go through everything step by step. And also, you know, for some reason you can't find the link, if you go to the behavioral value investor Substack, it's uh there was an article a couple of weeks ago that I wrote which describes the overall framework where you and then you have a link to download the um the PDF. >> Sounds good. Sounds good. Thank you for that. Um where do we go from here, Gary? I mean, um, it sounds like, you know, this is sounds like early, you put a lot of work into it, but this is early days like are you you you're going to get feedback from, you know, perhaps dozens, hundreds, I don't know, of people and then you're going to come up with a 2.0 and you are just to be I just want to be very crystal clear in case this isn't clear to anybody. You are using this to allocate real money for your investors. >> Yeah. No, absolutely. Like I was I just had my annual meeting for the partnership uh on Friday and I I was talking about this with my partners and I literally you know pulled up the screen of a notebook I have with a gazillion sources of a British company I'm researching and you know showed some of the kind of live stuff like it's great. It's not it's not like a oh let's design some theoretical thing and see you know throw it out there on the internet and get some feedback. It's like actually what I'm doing understanding that you still have to iterate and what I'm doing right now will probably be better when I do it in 6 months and 12 months and 18 months and so forth. But it's absolutely um a useful tool today and it will only get better. I mean I think one you know so you know I think you and I met in Vale right the friend friends friends bali's um conference for the first you know bunch of years back right and I was in veil um this summer and I gave a talk on AI there and you know and so one of the things I had you know you know you know how these things go it's a short talk and then you do Q&A from other experienced investors and and so right where I was supposed to take questions I said, "Wait, actually this time we're not going to do that." And the reason is I already asked AI what questions you guys are going to ask ask and here they are you know and you know so we had we all had a good laugh and of course I took questions but I kind of you know it was interesting to see what AI thought the objections were or the challenges were with using AI and one of the a few things that stood out. One was, hey, is this somehow going to dull your ability to do research? Like you're going to atrophy your like primary research muscles. And I think the answer is could if you use AI incorrectly, but if you use it correctly, you're actually going to get more reps on the things that are the most value adding. And you're going to spend less time going, let me put it this way, you're going to spend less time going to the gym and setting up the weights and doing all this other stuff and changing into your clothes and more stuff on the bench, you know, you know, uh, pushing at the weights. So, I think that's how I view it. Um, >> I was going to use the analogy of a carpenter who gets better with with his tools over time and becomes more creative. >> No, absolutely. The other thing AI point, and this is the million-dollar question, or maybe the billion dollar question is [laughter] AI, one of the questions was like, hey, it's awesome that there's all these tools and you can do all this stuff, but is there any evidence that AI actually improve return improves returns, right? And the honest answer is we just don't know. Um I think the head of uh Citadel was on Bloomberg a couple of months ago saying ah nope you know it's nice making junior analysts more efficient but it's not going to add alpha and maybe he's right. I don't know. Um and I think he doesn't know to be honest it's a it's still being kind of worked out and decided. I can tell you with near certainty though is if you don't use AI you'll be at a competitive disadvantage. So maybe you, you know, if you use AI, well, it's not going to magically give you 300 uh basis points of alpha extra per year, but if you insist on doing things kind of the old way uh completely, I think chances are you'll be at a disadvantage. And I don't think I don't think you need to do that because it's it's not that hard to do it. Well, >> right. And if we're if we're um just sort of uh how does one say analogizing about about how this works um or might work out in light of the Citadel guys comments. The first thought that came to my mind when you were talking was Renaissance technologies. I mean if you would have told me before that happened that a bunch of mathematicians and physicists were going to get together and go through a period I think it was 20 years of 80% annualized returns. 80 80 um something truly insane. It might even be 60 or 80, but any of those numbers is just off the charts insane. Um if you had told me that, I would have said, "Oh, that's silly. They're not Warren Buffett." You know what I'm saying? It just would have been crazy. But that's exactly what they did. So, who who knows? This is a tool just like the mathematical tools they were using. maybe we could say um or similar enough to and and maybe there's a you know an AI renaissance out there that's going to just you know shoot the lights out. >> Although you know I'm going to you know like just push you back just a little bit because I think that there's a here >> you know in that like I'm a process guy. I'm a long-term guy, but there's so many hypeers out there, and I know it's not you or any any of your listeners, but like, you know, you ever watch YouTube and get these like commercials like, "Hey, bro, get my trading system and make a billion bucks from your couch, right?" And I feel like AI and the hype surrounding it just gives more ammunition to those like you know shysters bas you know like people who are uh you know selling hope and like it's a confidence scheme right but and I think this is why like no I'm not out here saying hey I have this magic AI tool go buy it for $300 and you will get amazing returns. I'm saying I have basically essentially a very useful public white paper that's completely free and that you can go and do whatever you want with even if you get one idea from it. It's it's going to be helpful. But I think there's going to be a lot of other people who are going to try to make money not investing using AI but taking advantage of the gullible by selling them some you know pipe dream um that AI is going to be a magic solution. And I think you have to be careful. You know, you don't want to fall for that. >> Yeah. Human beings are vulnerable to that type of an appeal. They want to hear that there's a button they can press and and and they and and if you're um if you're a novice, let's just say, or just really don't know anything about how investing works, you may think that there is some way to just kind of um figure it all out so that you never ever lose money and you're never wrong and you're always making lots of money and that just doesn't exist. >> Yeah. [laughter] >> And that's how you lose money. Yeah. >> Yeah. No, >> thinking that is how you That's right. Yeah. [snorts] >> And I, you know, it's funny. I actually saw a cartoon post on LinkedIn a few weeks ago, you know, and and in the first like it's like kind of two scenes, right? In the first kind of panel, uh this guy, you know, P gives the interviewee the pen says, "Sell me this pen." You know, you probably all seen those like, "Sell me, uh uh this pen and you know, kind of tests, right?" And the se and the second p uh p p p p p p p p p p p p p p p p p p p p panel the guy just uh the interviewer says it's aic you know and it's such a you know you know for those you know like you know it's a right the reference to like the hype around AI agents and all right so uh you know so it used to be social proof it used to be I don't know Matt Damon you used it to sign his latest contract you that that used to be like the right answer now it's agentic right you know is like the the punch line and and it's so true right there's so much hype and so much I mean I think it's good that we are experimenting as a society and as a community and investors and even beyond investors um but I think that only a small fraction of the things we're working on are going to actually work out and you know it's important to be realistic about what AI can and cannot do at given point in time. >> It is an exciting time isn't it? Um, on the one hand, uh, lots of people like you and and and us are excited about how to use this new tool. And on the other hand, um given the business that we're that we're in, you know, you and me and and overwhelmingly every guest we ever interview, um we can't help noticing that um lots of people are investing a lot of capital in um data centers that are um I previously represented them as being highly utilized. the load factors, the power usage is is great, but the server utilization is not mostly, as far as I can tell, under 20%. So, the moment strikes me as more similar to the buildout of fiber in the late 1990s, early 2000s than I ever thought, you know, with when there was a lot of dark so-called dark fiber. So there's a lot of dark dark server capacity and and no return in sight on an investments approaching trillions. You know, there there's going to be this trillion dollar multi-t trillion dollar asset out there if they keep going the way they're going and building the way they're building. Um and [clears throat] it's uh it's just I don't know how to think. Like we just spoke with Ben Hunt and we were saying, "Boy, Ben, it sounds like you don't think 2026 is going to be a very good year." and and um it's I don't know do do you just now that we've talked about AI there you know let's talk about it as investors like do you do you have a macro view do you have a view on how this is going to play out in the next 6 12 18 24 months do you indulge that sort of thinking >> yeah no I hear you I I'm a much more of a bottomup investor but listen I think about these things as well but and I think like the Gartner hype cycle kind of framework is perfect for this, right? Because I think like we've seen this so many times where there is a promising technology. It is capable of doing a lot of new things, but then the expectations for what it actually can do and how quickly it can get those things done, you know, is know it's go it they're way over overhyped, right? And I think that the some of the studies I've seen like the amount of extra revenues you need um you know to actually generate a return on the trillions of investments that are going into AI. Those are huge numbers. And like and I know like Elon Musk is out there saying there's going to be an army of robots and uh 10 years from now, you know, no one's going to need to work or something like that. But you know, it's Elon. He likes he like he's an amazing creator of wealth but he's also likes to fantasize about the future and you know a lot of his forecasts don't quite come true. So I don't know. I mean I also know the reason I'm sure you've been following the Michael Bur um saga with you know his you know related to your capex question is that these companies are like manipulating their earnings like they're of balance sheet debt their depreciation is you know insufficient. So there's a lot of froth in the market. I think that there's a lot of gullible people who frankly you know the last 15 years we haven't had a real bare market like we had many corrections where the government quickly stepped in and flooded the market with liquidity and so there's a whole generation of investors who like they know the stocks can go down but they in the back of their minds they expect that dip buyers will come and they will buy all the stocks and it'll be all okay and that might be true but as a student of financial history and markets like we you and I both know there are like decades where things were not that great right and we just happened to be exiting a decade and a half where the markets have been great and very benign and forgiving so I think you combine that with the AI hype and despite the fact that AI is quite real as I I think we spend most of our conversation talking about I think at the end of the day I think there's a lot of danger in the markets and I am very nervous that asset prices almost across the board are very frothy uh or at least very full maybe not frothy in every you know corner of the market but certainly not a lot of distress or cheapness going around and expectations are built for perfection but you know what do I know at least that's my view they are built for perfection yeah um that's but it's funny though because given the um generally you know higher market multiples and things that have prevailed really this century this century really post you know post 2000 um you you could have said that and I did say it I have said it many times you know so it's I guess I'm kind of I want my listener to know [clears throat] and and if you disagree by all means weigh in like noting that something is priced for perfection is not a timing call Gary is not calling the top. >> That's why I hed in the beginning like valuation is probably the worst, you know, indicator of what's going to happen in the next 6 to 12 months. >> Yeah. I just want to >> land on long-term expectations. That's for sure. And I think that if you if you're listening to this and you think, "Oh, I heard markets return 10% per year." And by the way, recently that's been more than that. And you think that that's like, you know, guaranteed from this starting point. I would I would go and recheck your assumptions because you know that I don't think that that's likely to happen for a while. >> Yeah. Long long term um from this moment lots of people have studied it. You know if you go back in history from this kind of a price for perfection moment um the S&P 500 has generally done really poorly and been you know flat or even down over many years like a decade 12 years numbers like that. And again, I'm not forecasting that, but I am cautious. I can tell you as a bottom up investor, I'm having trouble finding new ideas. You know, it's not easy. And I know I'm not supposed to say that because >> old investor, you know, most investors are like, "Oh, I can find amazing ideas in any market. You know, just give me some money." I rather be honest and say it's tough. If you want quality and if you want a good price and you want a business you understand and you want there to be something misunderstood, mispriced, then that that's not big opportunity set right now. At least not one that I'm finding anyway. >> Right. We um in in our extreme value newsletter, Mike Barrett and I have done really well over the past five years. Um but you know, just like everybody else, right? >> Um wow. >> And and I mean much better than the market, you know. >> Uh many many you know tens of basis points more just blow it out but we feel like we're struggling every month now it's it's really we're going ohh god you know can we really recommend this with a straight face yeah we can because of this and this and I wonder when we're going to just say can't do it we we just can't do it one more month you know [laughter] but it is difficult >> I I think your honesty is refreshing right because I think Again, I think it's important to be honest both with yourself and with others. And I think look, I always tell my partners, you know, it's not about how the portfolio, how attractive the portfolio is today because what I'm doing is I'm sitting there day in and out and I'm hunting for opportunities and maybe to bring it back full circle, you know, using both AI and other methods as well. I haven't given up on, you know, you know, the the good old just, you know, hard work and elbow grease, right? Um, and I think the opportunity set will change. And I think if you're being properly cautious right now in how you're investing, that's going to pay dividends when you are able to take advantage of the more attractive opportunity when a lot of people maybe they're not. >> All right, Gary, it's time for our final question, which is the same for every guest, no matter what the topic. Um, you've answered it before. I kind of hope you forgot because I think it works better that way. [laughter] >> No, I'm good. No, that's good. No, don't be nervous. I I'll I'll tell you. The question is this. It's it's for our listener's sake um to give them a takeaway. So, if there's one thought or one takeaway that you would like our listener to have, um what would that be? >> Yeah. Well, since it's uh since the main topic of the conversation is AI, I would say whether you download my uh framework or not or do whatever you do, start using it uh for investing now, at least part of your process and start learning how you can build that for you in a way that's right for you. So, if you uh walk away with nothing else is that just getting the reps in on building AI into your investing is going to be something you're going to be grateful for for years to come. >> Thank you for that and thanks for being here, Gary. Uh it was really great to talk to you. >> Thank you for having me back. I appreciate it. Folks, the White House is on a stock buying spree that is comparable to a hedge fund that has raised billions. And the stocks in question are surging as a direct result. 200% within 24 hours in one case, 90% in another, and that's just the tip of the iceberg. Which is why we're hosting an urgent financial summit, the stocks that save America, on Tuesday, November 18th. That's tomorrow. This event will feature a man you may know well, 50-year investing titan Rick Rule, along with my guest today, Nick Hajj. Together, they'll reveal why the success of some stocks could now be a matter of national security, how a select group of US companies could soon play a historic role in rebuilding the nation's industrial backbone and the wealth of regular people, and the name and ticker of a free stock recommendation they believe could soar as this story accelerates. A stock I personally love, by the way. If you enjoyed today's episode, this briefing will be a mustsee followup. Now, head over to www.whousetocks.com to learn more and reserve your spot. Again, that's www.whoustocks.com. Don't delay. Well, it was great to talk with Gary and it was great to hear somebody sort of get into it um with using AI in an investing process and I'm so glad that he has this PDF. I have not read it yet. He just gave us the link like shortly before we did the uh did the recording here and I have not gotten into it. I I am going to tear that thing apart because I'm struggling right now to sort of figure out AI in my investment process. I use I use a lot of queries, but they tend to be kind of general and I wind up having to do a lot of work to get the garbage out of it. Um, but that was really, you know, as always, Gary's a very thoughtful processoriented guy. So, that was perfect, I thought, and perfect for this moment. >> Yeah, very uh a great resource uh that he's put together. It seems like I I haven't uh checked it out myself yet either, but I will. And it's um just reminds me of kind of how I'm trying to think about all these AI tools that are out there now. Um about really looking at it as like an assistant or um uh like a kind of a entrylevel uh assistant job uh sort of role. uh which is which is good news if you have a job already and are kind of mid or or upper you know career you could you can afford to do this and put these uh let these tools help you. We're seeing right now also that's a problem for people who are trying to find kind of entry level jobs in kind of the knowledge uh economy uh you know maybe junior level analysts and and those sorts of things. Um, but you know, there will be I there's got to be ways to to use it uh from an like entry level perspective as well too. Um, and that'll work itself out over time. But, uh, what what what Gary was talking about is super interesting practical ways of putting this technology to work. I loved his his thing about if you if you didn't know anything that he was saying his points about prompts and asking the the models questions and then notebook LM um just kind of a tool where you could dump a ton of information like the information that you want and uh it's like a second brain you know it's like all the stuff you you wanted to ever remember uh you could dump into like a single folder essentially and have the AI you know pull pieces of information or analysis from it. So, um those two things alone are super valuable to do if you're trying to figure this stuff out, >> right? Um I just realized as you're speaking, you and I, um you of course you write the Stanbury Digest, you know, virtually every day and I do it once a week usually. Um, and at Stanbury generally, even in our newsletters when we make recommendations, we're mostly targeting uh, and we're overwhelmingly our subscribers are self-directed individuals um, who are managing their own money. There are plenty of investment professionals running other people's money among our subscribers, but overwhelmingly um, you know, it's the self-directed individual. So we do a lot of storytelling and we try to get our analysis um you know we we don't just throw a lot of numbers at people. We try to do the analysis and then [clears throat] make it far easier to understand the thesis than a Wall Street analyst might make it. And that storytelling we do I think is really important. I have found that I use AI and I do like general, you know, throw it out to the internet kind of queries to hunt for this stuff. I'm using it to find the story almost more than to do analysis more definitely more at this point. And I'm wondering you you you do a lot of that same kind of storytelling every day of the week just about. Uh, and I'm wondering what your, you know, specific use case looks like. >> That's interesting you say that. I I I haven't really used it at all to to to find stories or narratives. I did early on. This is probably at least over a year ago. Early on, I was like, "All right, let me see what this if how I could incorporate this to daily, you know, the daily work." And >> actually, let me let me um qualify. It's not like I use it to find the story or the narrative, but I [clears throat] definitely use it to hunt down as many sort of facts and details, right, >> to amplify the narrative, right? So, just to just to clarify. >> Yeah. Yeah. Yeah. I get it. And and early on, you know, I was I was like curious, okay, how how great can this can uh you know, this a AI search be essentially? So, you know, I'd say, "What are the top what are the top five stories people are talking about in the stock market today?" And, you know, it'd give you some vague representation of it. Uh, which wasn't very helpful for me at all. Um now so so I use it more as from like a writing editing perspective uh on as like I was saying as like an assistant like um you know you would have we have editors too um and we will but to uh have something written and and then have an AI look at it uh and may be able to fill in some gaps or tell you if you're missing something or um did you think about this kind of what Gary was saying about you know towards the end uh you just kind of yeah use it as a filter. Um so um that's really mostly um you know from from a daily perspective it's different because or for me I feel like because there's so much happening every day that a lot of times for me it's just easier I'm still at the point where it's easier for me to think with my own brain and just put it on the page and instead of that the interstep of going to the going to a technology to tell me what's going on rather than me just doing it myself. You know, I'm still at that point with a lot of a lot of this. Um, if that makes any sense. But, um, I definitely see the the the value in adding it to a workflow um to, uh, just kind of do do the do things that you're not necessarily mundane tasks, but stuff that's like been routine um, over time. And just once you start using this thing, I think you even start realizing, oh, I don't need to do that anymore. And that's okay. Like I'll I don't need I don't need it anymore. And and then you just do other stuff. And then you'll just do other things. You won't even be realizing it. >> Interesting. Okay. Well, I just wanted to see where you were. Um and I'm sure it'll be different a year from now, right? >> Oh, yeah. Yeah. Probably be different in a couple days, but [laughter] >> Yeah, that's right. Yeah. Me, too. Um, but one thing I definitely am going to do is is learn how to use Notebook LM. Um, which I really haven't haven't truly done yet. Um, to focus just, you know, so much of what we've done certainly in Extreme Value and in other publications focus on focuses on really high quality sources, 10Ks, company reports, you know, on the ground due diligence, etc., etc. And um doing that faster is like a dream come true and that's what you could do with with Notebook LM. Anyway, that was awesome. I hope everyone else got lots out of it and took some good notes and we'll download Gary's PDF and take a look at it. Um it was a really great fun interview and a great episode. I hope you enjoyed it as much as we really really truly did. Opinions expressed on this program are solely those of the contributor and do not necessarily reflect the opinions of Stanbury Research, its parent company or affiliates.