Latest / The Dutch Investors / #79 | How to Use AI as a Long-Term Investing Tool w/ Compound with AI

Transcript

#79 | How to Use AI as a Long-Term Investing Tool w/ Compound with AI · The Dutch Investors

• 0:08
  SPEAKER_00: You need to take time, learn the tool and try it, test and learn and progress.
• 0:13
  SPEAKER_00: And at the end of the day, you will acquire a skill that you will use into your investing process.
• 0:20
  SPEAKER_01: The tool is only as good as the person using it, right?
• 0:23
  SPEAKER_01: It's exactly the art it's the archer, not the arrows.
• 0:26
  SPEAKER_00: Exactly.
• 0:30
  SPEAKER_01: Do you think Munger, if he was still here, would he be using AI?
• 0:34
  SPEAKER_01: Do you think?
• 0:34
  SPEAKER_00: I don't think that he played with the tools.
• 0:37
  SPEAKER_00: He didn't have the age, but he will be.
• 0:39
  SPEAKER_00: I imagine if he had uh 30 or 40 years old, he would be happy to use it.
• 0:47
  SPEAKER_00: If you ask the right question, you can learn a lot.
• 0:50
  SPEAKER_00: And we all know that uh Mangar was a learning machine.
• 1:15
  SPEAKER_01: Welcome on the show, the Dutch Investor Show.
• 1:17
  SPEAKER_01: How are you doing?
• 1:18
  SPEAKER_01: I'm excited to talk to you today.
• 1:20
  SPEAKER_00: Thanks and welcome for and thanks for inviting me.
• 1:23
  SPEAKER_00: I'm doing good.
• 1:24
  SPEAKER_00: Thank you.
• 1:25
  SPEAKER_01: You told me you're from uh you told me you're from France, and uh well, so we're close neighbors.
• 1:30
  SPEAKER_00: Yes, I'm based in France and we have the same time zone also, it's easier to set up things.
• 1:35
  SPEAKER_01: Yeah, I know, right?
• 1:36
  SPEAKER_01: Yeah, that's always a hassle.
• 1:38
  SPEAKER_01: So there is so much going on right now.
• 1:41
  SPEAKER_01: Um, and I feel like AI is the buzzword and has been the buzzword for like two or three years now.
• 1:47
  SPEAKER_01: Um the the hype, uh whether we can call it a hype, we'll we'll get to later, but it's still the word of the year, I feel like.
• 1:56
  SPEAKER_01: Uh, if it's not Trump, it's gotta be AI.
• 1:59
  SPEAKER_01: So I'd love to know what made you so interested in AI.
• 2:03
  SPEAKER_01: So maybe just to set the stage, um briefly introduce yourself, uh what you do on a on a daily basis basis, and you know what is it about AI that makes it such a unique and powerful tool, you think?
• 2:17
  SPEAKER_00: So I'm Mustafa from the Compound with EI newsletter.
• 2:22
  SPEAKER_00: So I'm a long-term investor, I'm an engineer by training, and I have been teaching people in my newsletter how to use AI for investing because I deeply believe that AI can make us go 10 times faster and also 10 times deeper in the investing research.
• 2:43
  SPEAKER_00: So I have been investing since 2018-19.
• 2:49
  SPEAKER_00: Um Buffett Mangor investor style, started as a value investor and then progressed more to a GARP growth at a reasonable price investing style.
• 3:01
  SPEAKER_00: And uh since the first Chat GPT version, I have started using AI, and I had always in my mind investing as a way to go faster and understand deeply things.
• 3:15
  SPEAKER_00: And with time, with the new version, each time the models kept getting better and they had better results.
• 3:23
  SPEAKER_00: And when I started having full process automated to go faster and deeper, I started sharing it in my newsletter, compounded with AI, and now we are 21,000 investors inside the newsletter, and I'm always happy to share in my knowledge and my expertise with people.
• 3:43
  SPEAKER_01: Awesome.
• 3:44
  SPEAKER_01: I think uh we have the similarity that we both are.
• 3:47
  SPEAKER_01: We see Buffett and Munger as uh as idols, as someone who inspired us for sure.
• 3:53
  SPEAKER_01: Do you think Munger, if he was still here, would he be using AI?
• 3:57
  SPEAKER_01: Do you think?
• 3:58
  SPEAKER_00: Of course.
• 3:59
  SPEAKER_00: I think he will be someone who will be using a lot AI.
• 4:03
  SPEAKER_00: You know, Munger had something special that his peak years were uh when he had uh 90 from 95 to uh around 100.
• 4:14
  SPEAKER_00: He died uh one month before turning 100 years.
• 4:18
  SPEAKER_00: And uh that's where that was his peak.
• 4:21
  SPEAKER_00: He has meeting with a lot of people, and uh he will meet a lot of people, sharing his knowledge, and I have heard him uh have heard people talking with him about AI, and he said that he understood it well that it will be a revolution.
• 4:37
  SPEAKER_00: I don't think that he played with the tools, uh he he didn't have the age, but he will be uh I imagine if he had uh 30 or 40 years old, he will be uh happy to use it because AI is uh a tool that will read, understand, summarize, and if you ask the right question, you can learn a lot.
• 5:00
  SPEAKER_00: And we all know that uh Mangar was a learning machine.
• 5:03
  SPEAKER_00: Uh that's Manish Pabra, he was talking when he visits him.
• 5:08
  SPEAKER_00: He sees books, and uh he said uh Mangar is like a machine.
• 5:12
  SPEAKER_00: He has books, he learns, and then he progress.
• 5:16
  SPEAKER_00: And the next book and the next one, and he do this all the day.
• 5:19
  SPEAKER_00: I think he would be using using it a lot.
• 5:23
  SPEAKER_01: When it comes to AI, it feels like there are two groups like the believers and the disbelievers.
• 5:30
  SPEAKER_01: Do you sit in one of these groups?
• 5:32
  SPEAKER_01: Perhaps the question is a bit unnecessary and we already know the answer, right?
• 5:36
  SPEAKER_01: But I would just love to hear your thoughts on you know the power of AI.
• 5:41
  SPEAKER_01: Do you believe AI is here to stay?
• 5:43
  SPEAKER_01: Will it be a revolutionary new technology, or do you think it slowly fades away in a couple years and it'll be here, but it won't be as exciting as it is today?
• 5:56
  SPEAKER_00: You know, as humans, we tend to try to simplify things, and that's something that that's good because we need to in our daily life, we need to simplify things to go faster.
• 6:07
  SPEAKER_00: You don't need to understand the detail of each subject and anything you have in your daily life, so it makes us go faster, but that's a mental model that we should not use when we have such big subjects.
• 6:21
  SPEAKER_00: There is a lot of people who say, I see AI, it's a bubble, or AI is useful or not useful, and that's it.
• 6:29
  SPEAKER_00: My approach is different.
• 6:30
  SPEAKER_00: Like I said, I'm an engineer by training, so it took time to understand what's AI and to apply it in different use cases.
• 6:40
  SPEAKER_00: I've tried for my work, for my personal life, for my project, for uh raising my child, for a lot of projects that I had, for renovation, renovating my home and for investing also.
• 6:53
  SPEAKER_00: And the idea is that for each use case, you need to use AI differently.
• 6:59
  SPEAKER_00: If for some use cases, AI is not the best thing to do today because things progress and it goes fast.
• 7:07
  SPEAKER_00: But for most of the subjects, AI can help a lot.
• 7:11
  SPEAKER_00: And my uh my answer is you need to know how to use AI for each use case.
• 7:17
  SPEAKER_00: If you are not using it correctly, it will be garbage.
• 7:20
  SPEAKER_00: If you are using it correctly, it can be helpful, and depending on the task you are trying to do, uh AI can do 80% of the work or 20% of the work.
• 7:32
  SPEAKER_00: So, for this, you need to take use case, try to learn about AI, apply it, and then have a decision, have a conclusion about it.
• 7:42
  SPEAKER_00: For me, for investing, I'm sure that it's it's helping to go faster and deeper, and we can go into details to share with you use cases.
• 7:49
  SPEAKER_00: And um, have I have been sharing this in my news that it was 20,000 21,000 people.
• 7:55
  SPEAKER_00: I receive every day uh messages from people and DMs telling me that uh they are going faster, understanding a lot of things with it.
• 8:04
  SPEAKER_00: I have talked, I have been talking to a friend a couple of days ago, and he told me that he didn't invest in stock because he ran one of my prompts about how to identify the red flags, and he knew very well the sector, but he didn't think about that point.
• 8:19
  SPEAKER_00: So AI was pointing something that he didn't think about, and uh at the end of the day, it's the human who takes the decision.
• 8:27
  SPEAKER_00: But uh AI can make us go faster, and we can go into the details, structure things to help um uh people understand the basics.
• 8:36
  SPEAKER_00: Like you said, it's a hype, and uh from investing point of view, it's a bubble, and they don't invest in AI stocks.
• 8:43
  SPEAKER_00: But uh since hype is always there is always part of truth and a part of hype, like internet in the tech bubble.
• 8:52
  SPEAKER_00: Of course, internet did change the world, but all the websites that they were created with no revenue model, they didn't do anything, but the technology was revolutionary.
• 9:03
  SPEAKER_00: We are in the same situation.
• 9:05
  SPEAKER_00: AI did already change the world and will change the world, and we need to accept it.
• 9:11
  SPEAKER_00: We need to learn how to use it and to use it correctly.
• 9:16
  SPEAKER_01: That makes sense.
• 9:17
  SPEAKER_01: I think that's a good comparison.
• 9:19
  SPEAKER_01: So before we go into the nitty-gritty of AI, you know there are three types of edges, and you'll probably uh know them: like the analytical edge, right?
• 9:28
  SPEAKER_01: Being able to look at things differently, analyze things better than someone else, the informational edge, knowing something better or something that someone else doesn't know, and the behavioral edge and uh having a better mindset, for example, more patience, whatever.
• 9:45
  SPEAKER_01: Now, I'd argue that for most investors, the behavioral edge is the best one for us.
• 9:53
  SPEAKER_01: Uh you know, today everyone has access to the same tools, ChatGPT, Gemini, perhaps Notebook LM, YouTube, Fiscal AI, whatever it may be.
• 10:03
  SPEAKER_01: Do you even think that the informational edge still has any use case?
• 10:07
  SPEAKER_01: Where is the edge for investors in today's day and age, do you think?
• 10:12
  SPEAKER_00: Let me answer you with a simple example.
• 10:15
  SPEAKER_00: If you take an experienced investor and you give him an analyst, will he help him or not go in faster?
• 10:23
  SPEAKER_00: I think that everybody will say that he will help him.
• 10:27
  SPEAKER_00: How this analyst, junior analyst or an intern will help an experienced investor.
• 10:32
  SPEAKER_00: He will not think instead of him, he will not help him emotionally to manage a 30% drop in stock price.
• 10:41
  SPEAKER_00: But he will make him go faster.
• 10:44
  SPEAKER_00: Let's take an example.
• 10:46
  SPEAKER_00: I have a lot of WhatsApp groups we were talking before, and uh a lot of uh my friends share stock ideas.
• 10:53
  SPEAKER_00: What I do, uh what I have done in the past is that I I take all of these ideas and I will say when I will have time, I will analyze these stocks.
• 11:02
  SPEAKER_00: So I start with the first step to understand what's the business model, what's the industry, how it works, and then say, okay, do I want to go deeper or not?
• 11:12
  SPEAKER_00: Because sometimes I don't want to uh to invest in some industries that I don't understand, or the business model seems uh very far from my investing style, or I have share a searching uh stock with moat and I see that there is no mode, so I I stop.
• 11:29
  SPEAKER_00: But that's part of understanding business model.
• 11:32
  SPEAKER_00: Usually it could take two, three, or four hours, depending on how well you know that industry.
• 11:39
  SPEAKER_00: But now, when one of my friends shared with me an idea, I have a process with AI that makes me understand the business model in 10-15 minutes.
• 11:49
  SPEAKER_00: What does it mean?
• 11:50
  SPEAKER_00: Is that I can analyze and understand 10 to 20 more stocks in the same timeline.
• 11:59
  SPEAKER_00: So I'm turning more rocks.
• 12:01
  SPEAKER_00: So that's the idea.
• 12:02
  SPEAKER_00: The first thing with AI is that like I'm having an analyst who is summarizing things and giving me things to read to go faster.
• 12:09
  SPEAKER_00: That's my first use case with AI.
• 12:11
  SPEAKER_00: So every investor can use it if he has the right process, he understanding the right prompts.
• 12:18
  SPEAKER_00: He will have someone working for him to make him turn in more rocks.
• 12:21
  SPEAKER_00: At the end of the day, it's the decision of the investor, but uh at least help us to see more stocks, and like Peter Lynch said, the investor who will be not who is the one who is turning more rocks.
• 12:34
  SPEAKER_00: So I deeply believe that uh at least for for that use case, that's only the first use case.
• 12:39
  SPEAKER_00: AI is helping us to see more stocks, and it's up to us later to go deeper or not.
• 12:45
  SPEAKER_01: So increasing like the discoverability, turning over more rocks, which means you increase the chances of finding these potentially high-quality companies by turning over uh more rocks.
• 12:56
  SPEAKER_01: That makes a lot of sense.
• 12:58
  SPEAKER_01: And I think you already briefly touched on understanding the business model, so potentially finding out if a company is within your circle of competence before doing any research.
• 13:09
  SPEAKER_01: So uh I think that's uh pretty applicable.
• 13:12
  SPEAKER_01: Before we get into your workflow, which I'm very interested in hearing, maybe just touch on the toolkit.
• 13:18
  SPEAKER_01: So, what to use.
• 13:20
  SPEAKER_01: I mean, there are a million tools out there.
• 13:23
  SPEAKER_01: So, what tools?
• 13:24
  SPEAKER_01: Maybe you can give us some examples.
• 13:26
  SPEAKER_01: Do you use personally?
• 13:27
  SPEAKER_01: And would you recommend people testing and trying out for someone starting today?
• 13:33
  SPEAKER_00: Uh I'm really an AI adapter, so I'm using AI in my personal work and investing.
• 13:39
  SPEAKER_00: Let's just focus on investing part.
• 13:42
  SPEAKER_00: Yes, perfect.
• 13:42
  SPEAKER_00: Yeah.
• 13:43
  SPEAKER_00: Yeah.
• 13:43
  SPEAKER_00: To make it faster.
• 13:45
  SPEAKER_00: For investing, we need to understand two things that we can use AI in two different ways.
• 13:53
  SPEAKER_00: The first set of tools are closed loop.
• 13:56
  SPEAKER_00: I call them a closed loop.
• 13:58
  SPEAKER_00: When you will you want to master your your sources.
• 14:01
  SPEAKER_00: Let's say I want AI to work only on these sources and to give me an output.
• 14:09
  SPEAKER_00: The idea is that in this use case, I want to be responsible on the what I have, what I put in as in uh what I have as an input, and I ask the right questions, and AI will give me outputs that already only link to what I asked to do.
• 14:27
  SPEAKER_00: So if I will put it differently, the first use case, the first setup set of tools are um are of closed loop tools where the investor control all the inputs.
• 14:41
  SPEAKER_00: So, and I'm talking about Notebook LM, that's a Google tool where you will upload all your sources, you will ask a lot some questions and you have prompts and you will have answers.
• 14:52
  SPEAKER_00: And in this use case, you will have only answers linked to your sources.
• 14:57
  SPEAKER_00: So, for example, if you in your sources you didn't upload any report of the company of 2019, for example, if you ask what was the revenue for 2019, notebook LM will not know the answer because he will not go outside and search for information.
• 15:14
  SPEAKER_00: That's the first set of tools, and the best tool is easy, it's Notebook LM because that's mainly the only tool that works like this.
• 15:22
  SPEAKER_00: Because Google they have created to help students to learn, and they want them to work only on specific documents, books, and academic things that they will upload.
• 15:35
  SPEAKER_00: First use case.
• 15:36
  SPEAKER_00: If you want to work in a closed loop with the sources that you choose yourself, you need to use notebook LM, don't bother using anything else.
• 15:48
  SPEAKER_00: The second type of tools are the open loop, so where you will search information in internet and in uh in anything on in the internet, and there we were talking about all LLMs that we know, so ChatGPT, Gemini, uh, Cloud, Preplixity is not an LLM, but let's say Perplexity also.
• 16:08
  SPEAKER_00: And in this case, you can either talk to them directly in chat mode, like uh everybody uses it.
• 16:16
  SPEAKER_00: So go and type a question, or you could you can use a specific function what's called the deep research.
• 16:23
  SPEAKER_00: Deep research is not using the LLM, the AI directly, you are using an agent who is using AI.
• 16:32
  SPEAKER_00: So let's think about it like someone who will control the AI.
• 16:37
  SPEAKER_00: So let's take an example, you want to understand the ADN, for example, the payment processing company, which is a good stuff.
• 16:45
  SPEAKER_00: Uh you want to understand Adian, you can go to ChatGPT or your favorite LLM and ask him what's the business model of Adian.
• 16:53
  SPEAKER_00: He will go into internet and the search information samurai gives an answer.
• 16:58
  SPEAKER_00: That's the chat mode.
• 16:59
  SPEAKER_00: If you choose option D pre-search inside ChatGPT or another LLM, what happens is that there is another layer that will be created.
• 17:09
  SPEAKER_00: He will take your prompt, let's say it was only help me understand the ADN.
• 17:15
  SPEAKER_00: He will say, Okay, to understand ADN, what he's requesting, we want to understand the payment under C.
• 17:21
  SPEAKER_00: We want to understand what are the margins, we want to understand what's the growth.
• 17:27
  SPEAKER_00: And then he will go search for all information related to industry, all information related to the growth, all information related to margins, and he will go and search for around 100 to 200 sources and then give you an answer.
• 17:44
  SPEAKER_00: And that's really different from the chat mode.
• 17:46
  SPEAKER_00: But for this, I always share that we need to use the deep research mode.
• 17:52
  SPEAKER_00: So to summarize, if we you want to work in closed loop model, we need to use deep research, and then I can we can discuss later which tool and how to use it correctly.
• 18:05
  SPEAKER_00: Is it clear or you you want to double-click or on one of the points?
• 18:09
  SPEAKER_01: No, no, no, I think that makes sense.
• 18:11
  SPEAKER_01: You know, I I'm I'm pretty I know quite a lot about AI, but I think most listeners probably know more than me, so I think I'm a good reference here.
• 18:20
  SPEAKER_01: So closed loop tools only use the sources you upload.
• 18:23
  SPEAKER_01: So if you decide to only use the uh investor relation sources in notebook LM, it can only answer questions uh containing the information inside those sources.
• 18:35
  SPEAKER_01: And open loop tools, the Gemini ChatGPT, use the internet, the LLMs, to come up with an answer.
• 18:42
  SPEAKER_01: And the danger here is, I guess, is hallucinations, they want to answer your question, which is most likely, whether they're true or not, it's not really doesn't really matter, right?
• 18:53
  SPEAKER_01: It's zeros and ones, and they just want to yeah.
• 18:56
  SPEAKER_00: And hallucinations are are a feature of EI.
• 19:00
  SPEAKER_00: That's not a bag, that's something that we need to accept.
• 19:04
  SPEAKER_00: If you are in open loop mode, AI will hallucinate.
• 19:09
  SPEAKER_00: Our job is to put the right processes and the right prompts, use it correctly to reduce hallucinations.
• 19:16
  SPEAKER_00: I've written already an article about this: how to reduce hallucinations.
• 19:20
  SPEAKER_00: Each time I uh I explain it to people, they understand deeply how how to use it.
• 19:27
  SPEAKER_00: Let's take an example.
• 19:28
  SPEAKER_00: The first thing is that you can use to reduce hallucinations is first to use the right uh mode.
• 19:35
  SPEAKER_00: If you are using chat mode instead of deep research mode, the difference can be uh 100% in terms of hallucinations.
• 19:43
  SPEAKER_00: Let's uh you can take the same company's same prompt and put it in uh standard LLM mode, chat mode, and you put it in deep research, the difference will be uh will be big.
• 19:56
  SPEAKER_00: So that's at least the first step to reduce hallucination.
• 19:59
  SPEAKER_00: The second step is the prompts itself.
• 20:01
  SPEAKER_00: The prompt is only the instruction to ask the AI what to do inside the prompt.
• 20:06
  SPEAKER_00: You can say use only sources from the investor relationship.
• 20:10
  SPEAKER_00: You you can name even some industry expert, and AI will mostly go to only sources that you asked.
• 20:18
  SPEAKER_00: You and ask, you can ask for specific structure, you can ask him to double check.
• 20:24
  SPEAKER_00: And if you apply all my check or the checklist for to reduce hallucination, honestly, you have uh you will not have zero errors, but the errors that you will have will be 0001%.
• 20:38
  SPEAKER_00: And at the end of the day, you are not asking AI to buy the stock, you will you will take that decision.
• 20:44
  SPEAKER_00: So for this, I don't want people to say AI has hallucination, I will not use it.
• 20:49
  SPEAKER_00: That's not like saying car can have an accident, I will never drive.
• 20:54
  SPEAKER_00: That's not the best approach.
• 20:57
  SPEAKER_01: And people can hallucinate too, right?
• 20:59
  SPEAKER_01: If you uh grew up with the wrong rights, so yeah, yeah.
• 21:03
  SPEAKER_01: Um, so before we go into your workflow, do you use a specific platform or LLM?
• 21:10
  SPEAKER_01: Or rather, do you prefer one over the other?
• 21:12
  SPEAKER_01: So my question is do you switch between models?
• 21:16
  SPEAKER_01: Do you have like a specific rule where okay, I'm reading a 10K now, so I use notebook LM, but for deep research I use Gemini, but for brainstorming I use ChatGPT, or how do you how does that depend on the how do you know which LLM to use?
• 21:33
  SPEAKER_00: That's the right question.
• 21:34
  SPEAKER_00: We need to understand which LLM to use and for which use case.
• 21:39
  SPEAKER_00: That's very important.
• 21:42
  SPEAKER_00: Let's say we have three use cases.
• 21:44
  SPEAKER_00: First one is only to understand specific things.
• 21:47
  SPEAKER_00: For example, if you are not expert in uh accounting, you have basic questions about accounting, you want to learn about it.
• 21:56
  SPEAKER_00: That's something that you can use in any LLM, whatever.
• 22:00
  SPEAKER_00: Chat GPT, Gemini, Cloud, whatever LLM, and ask your question and go into mode to follow up question.
• 22:06
  SPEAKER_00: So you ask the question, and then let's suppose that you don't know what week return on invested capital.
• 22:14
  SPEAKER_00: You ask the question, you can you tell him, give me an example, and you are learning.
• 22:17
  SPEAKER_00: That's the easiest thing.
• 22:19
  SPEAKER_00: You can use whatever LLM, you will have similar results.
• 22:22
  SPEAKER_00: The second use case will be when I want to understand the global idea on how an industry works or how stock works, and then especially the you should do use the research.
• 22:36
  SPEAKER_00: That's that's non-negotiable.
• 22:38
  SPEAKER_00: And I explained why.
• 22:39
  SPEAKER_00: The second thing, which LLM were to use it today in January 2026, the best one is Gemini, and I the paid uh always talking about the paid version that the difference is uh is bigger, and it's 20 uh euro, 20 dollars, so it's not a big investment.
• 22:59
  SPEAKER_00: I'm using Gemini.
• 23:00
  SPEAKER_00: The why is that at least once a month I take all my prompts, uh, I test them on the same stocks in different LLMs.
• 23:08
  SPEAKER_00: So I go into ChatGPT, I go to Perplexity Cloud, Gemini, and test them and analyze the results.
• 23:15
  SPEAKER_00: And I usually usually use companies that I know well that I have been investing owning for three, four years, that I know them very well, and compare results.
• 23:24
  SPEAKER_00: And they compare them on different levels.
• 23:26
  SPEAKER_00: For example, uh, is the result more precise on the numbers, on explaining what I'm asking?
• 23:34
  SPEAKER_00: Uh, is it uh is it there is there a lot of fluff, so non-necessary words?
• 23:39
  SPEAKER_00: And for the moment, for the last four or five months, Gemini uh is very far to from uh from others in in this results.
• 23:49
  SPEAKER_00: The last use case will be when you want really to go deeper specifically on specific points, upload your documents into Notebook LM and ask your questions or generate custom reports.
• 24:03
  SPEAKER_00: In this case, when you know you want to really go deeper into all segments of products and uh check in all numbers and the details of uh of any business, go into Notebook LM.
• 24:15
  SPEAKER_00: And today, January 2026, AI is going fast and this can change in the future.
• 24:21
  SPEAKER_00: For this, I will continue testing and uh sharing what are the best tools for the moment.
• 24:26
  SPEAKER_01: Yeah, I think it's uh useful to follow someone like yourself that is so interested in AI, and it helps to follow certain people that test these models because it it does change very rapidly.
• 24:40
  SPEAKER_01: I mean, a year ago we wouldn't even have imagined that Gemini would grow so fast and be so much better.
• 24:46
  SPEAKER_01: ChatGPT was the buzzword and uh was the LLM to use.
• 24:50
  SPEAKER_01: We actually switched to Gemini ourselves as well a couple months ago because it's just currently the best one, but who knows, right?
• 24:58
  SPEAKER_01: 10 years from now, five years from now, or maybe next month it goes so fast.
• 25:02
  SPEAKER_01: But this also brings us to another issue regarding information.
• 25:07
  SPEAKER_01: There is so much information, and one could argue there's way too much.
• 25:12
  SPEAKER_01: How do you organize your your thoughts, your files?
• 25:15
  SPEAKER_01: Do you keep a massive database of PDFs in your Google Drive or do you grab fresh ones every time you start doing new research on a company?
• 25:27
  SPEAKER_01: Is there something that is there like a strategy you follow to organize all the information?
• 25:35
  SPEAKER_00: My strategy mirrors how I think about investing.
• 25:40
  SPEAKER_00: I really think about investing like owning businesses, and and that's not a buzzword, like people sometimes don't understand what it means.
• 25:49
  SPEAKER_00: They hear people saying, okay, when you are buying stock, you are buying the business.
• 25:54
  SPEAKER_00: But at the end of the day, they are not asking the right question.
• 25:56
  SPEAKER_00: For me, when I try to understand the business, even before AI, I want to understand deeply.
• 26:04
  SPEAKER_00: It's like I'm creating it.
• 26:05
  SPEAKER_00: For example, if you will go and open a restaurant, you will understand everything.
• 26:09
  SPEAKER_00: What are margin, you are check every uh every everything, who are competitors around, what are the legislations and everything before opening that restaurant.
• 26:20
  SPEAKER_00: And that's only simple business.
• 26:22
  SPEAKER_00: But people buy big businesses without understanding the details.
• 26:26
  SPEAKER_00: So for me, it's very, very important to understand really good a business and a sector before going into putting money on that stock.
• 26:37
  SPEAKER_00: So my AI process mirrors that belief.
• 26:41
  SPEAKER_00: And so, what I will do is that when I hear about a stock, I will try to understand and have an over you overview of that stock.
• 26:50
  SPEAKER_00: So, what I will do is that I have a prompt library when I store where I store my prompts.
• 26:55
  SPEAKER_00: So the prompts is are the instructions.
• 26:58
  SPEAKER_00: So if prompt is very well done, the result will be good.
• 27:02
  SPEAKER_00: If the prompt is not good, the result of the AI will not be good.
• 27:06
  SPEAKER_00: So, what I will do is that I have worked on understanding what are the best practices and how to write the good prompts.
• 27:14
  SPEAKER_00: I put my investing expertise to ask the right question.
• 27:17
  SPEAKER_00: For example, how I think about business model.
• 27:20
  SPEAKER_00: I want to understand in one sentence how they make money, who are the customers, why the customers buy, when they buy, what's making them taking the decision?
• 27:29
  SPEAKER_00: Is it the brand?
• 27:30
  SPEAKER_00: Is it a word of mouth?
• 27:31
  SPEAKER_00: Is it a paid acquisition channel?
• 27:34
  SPEAKER_00: I want to understand everything.
• 27:36
  SPEAKER_00: And that questions, I put them inside the prompt, and then I take the prompt and I test it on 10, 15, 20 business.
• 27:44
  SPEAKER_00: And then when I'm satisfied about the result, I validate that prompt and I will put in my library.
• 27:50
  SPEAKER_00: That's a prompt I will use next time when I want to understand the business.
• 27:54
  SPEAKER_00: That's that's one example, and they do this for all things.
• 27:58
  SPEAKER_00: When I want to go deeper into a return on investment, I have a prompt.
• 28:01
  SPEAKER_00: When I want to go to understand the management, I have something, and they have categories with validated prompts that I created, tested based on my experience.
• 28:11
  SPEAKER_01: So instead of having like a massive storage of all kinds of different files, you have like a library of prompts.
• 28:22
  SPEAKER_01: And if you want to know something, you use that tested prompt to get the information you want.
• 28:28
  SPEAKER_00: Exactly.
• 28:29
  SPEAKER_00: And uh to cover all investing subjects, you need around 20, 25, a maximum 30 prompts that can cover the biggest questions that um about investing research process.
• 28:44
  SPEAKER_00: What I will do is that I will, for example, I I hear about in your stock, I run the first a first prompt who is an overview.
• 28:51
  SPEAKER_00: We will talk about how it works to explain how the business model works.
• 28:57
  SPEAKER_00: And then I have a trick.
• 29:00
  SPEAKER_00: I I like to hear also, I like podcasts.
• 29:02
  SPEAKER_00: When I will be training, I want to always to hear to listen to a podcast.
• 29:07
  SPEAKER_00: So sometimes I will take that result, the result of that prompt.
• 29:12
  SPEAKER_00: Let me say it again, it will be easier to understand that.
• 29:15
  SPEAKER_00: So let's say I hear a stock name.
• 29:19
  SPEAKER_00: I I hear about a business, I want to understand it.
• 29:22
  SPEAKER_00: I will go to my prompt library, copy-paste my prompt, tested, validated about how to understand the business model.
• 29:30
  SPEAKER_00: I will run it in Gemini in deep research mode.
• 29:33
  SPEAKER_00: It will take minutes and the result will be a report with 20 to 15 pages detailed explaining the business model.
• 29:40
  SPEAKER_00: I can read it if I'm in front of my computer, or sometimes I don't have time.
• 29:46
  SPEAKER_00: I will transform it to podcast in Notebook LM, and then I will listen to that podcast while training or while going uh to work with in my car.
• 29:59
  SPEAKER_00: And then, like this, always I have two or three businesses to discover in a podcast mode.
• 30:04
  SPEAKER_00: But I'm people I'm someone who likes to listen to things, so it's helpful for me.
• 30:09
  SPEAKER_00: Other people want to read the 10 to 15 pages quickly, and that's the first use case when you get uh uh familiar with the business model in the fastest ways.
• 30:20
  SPEAKER_00: It used to take hours to do this.
• 30:22
  SPEAKER_00: Here, you are not trying to understand every little detail.
• 30:27
  SPEAKER_00: That's not yet the moment, it's to understand how the business works.
• 30:31
  SPEAKER_01: Makes sense.
• 30:32
  SPEAKER_01: I think that that's a good segment to uh transition to your to your workflow, so how to do it.
• 30:39
  SPEAKER_01: So let's assume the listeners, whether they're beginner investors or very experienced, doesn't really matter, but you don't have a company to analyze yet.
• 30:48
  SPEAKER_01: You have uh you have a blank slate.
• 30:50
  SPEAKER_01: Walk us through your investment process from finding a company to you know learning all about the company and figuring it out if it's worth an investment.
• 31:02
  SPEAKER_01: So maybe just from A to Z, how you how you think about the workflow.
• 31:08
  SPEAKER_00: You you want me to go with you have a stock name, or you want also to include the let's let's use Starbucks, uh, for example.
• 31:16
  SPEAKER_01: You know, you found a company, you used a screener, you found something interesting.
• 31:20
  SPEAKER_01: Starbucks is down 60%.
• 31:22
  SPEAKER_01: You want to figure out if it's a good investment or not using AI.
• 31:26
  SPEAKER_01: So, how do you think about that?
• 31:27
  SPEAKER_01: What do you do?
• 31:29
  SPEAKER_00: So let's suppose that you don't know what uh Starbucks does as a business model, but in this case, you are not living in uh maybe I'll choose a difficult company.
• 31:40
  SPEAKER_01: No, no, that's okay.
• 31:42
  SPEAKER_01: I have no clue what the company does now.
• 31:43
  SPEAKER_00: Let's take Palantir, for example.
• 31:46
  SPEAKER_00: We need to understand that you can never understand one business in one shot.
• 31:52
  SPEAKER_00: That's a wrong assumption.
• 31:53
  SPEAKER_00: You will not be an expert in a stock only in one day using AI or or without AI.
• 31:59
  SPEAKER_00: What we will do is that first we want to understand the sector.
• 32:04
  SPEAKER_00: If you are not familiar at all with that industry, I will recommend what I will do is that to understand that that industry.
• 32:14
  SPEAKER_00: So, what I will do is that personally, I will go to my prompt library, I will go to my category how to understand an industry, and I have created prompts.
• 32:25
  SPEAKER_00: The first one, I'm looking at my screen, is the industry overview prompt.
• 32:29
  SPEAKER_00: I'm opening it just to read to read um what's inside.
• 32:33
  SPEAKER_00: The prompt is you are an industry analyst writing for long-term investors to analyze the industry X.
• 32:40
  SPEAKER_00: The purpose is to produce a concise factual overview, and that's a prompt with uh two pages long.
• 32:47
  SPEAKER_00: Asking the questions that are important for me to understand in the industry, like what's the value chain, who are the competitors, how the industry um evoluted in time, what's the growth drivers.
• 33:02
  SPEAKER_00: I will run that prompt into Gemini Deep Research, or if you are paying for another LLM to code works also, but today we prefer Gemini.
• 33:10
  SPEAKER_00: You will have a report that will explain the industry.
• 33:13
  SPEAKER_00: Why we start with this?
• 33:15
  SPEAKER_00: Because it will give me an understanding of how that industry works.
• 33:21
  SPEAKER_00: That's my first step.
• 33:22
  SPEAKER_00: My second step will go to the specific name.
• 33:25
  SPEAKER_00: So Palantir, okay.
• 33:27
  SPEAKER_00: I understand a little bit the industry, and here you have a choice.
• 33:31
  SPEAKER_00: If you want to go hard mode code mode and go into notebook LM, upload all your resources and then generate reports that will summarize every segment of that uh of that uh business model.
• 33:46
  SPEAKER_00: You can do this for me.
• 33:47
  SPEAKER_00: The easiest way for now is to go and choose a specific prompt.
• 33:52
  SPEAKER_00: A prompt, for example, talking about the industry, I'm I'm checking my library.
• 33:58
  SPEAKER_00: Is I have the business overview report that we talked about.
• 34:01
  SPEAKER_00: He will explain how the stock will make money.
• 34:04
  SPEAKER_00: The second one is the growth and margin drivers, and this is focused on explaining what are the growth drivers for the business.
• 34:13
  SPEAKER_00: We understand them for the industry, now it's for the business, and what are drivers for the margins.
• 34:20
  SPEAKER_00: I have all my questions.
• 34:21
  SPEAKER_00: Question I am reading about cyclicality, sensitivity to economics, um to economy, and then you will understand also this part.
• 34:31
  SPEAKER_00: And running three or four prompts, you will have a good understanding of the business, and then you can say, Okay, I want to go deeper or not, and until today, I'll always open the annual report and read them.
• 34:43
  SPEAKER_00: But I will go a lot faster because I understand the context, I understand the industry.
• 34:49
  SPEAKER_00: And let's say Palantir is a good example.
• 34:52
  SPEAKER_00: I will not invest on the valuation.
• 34:54
  SPEAKER_00: I'm the value investor at the at heart, so I will not do it.
• 34:58
  SPEAKER_00: But at the end of the day, if you will not buy because valuation is not good, you have learned a lot of things.
• 35:04
  SPEAKER_00: Investing is not only about uh stock picking, it's about learning and progressing.
• 35:10
  SPEAKER_00: And using the this approach, you will each time learn and the progress, and sometimes your valuation is good and you will invest.
• 35:18
  SPEAKER_00: Yeah, at the end of the day, I don't use AI today to do the valuation work, I do it to understand the stock.
• 35:25
  SPEAKER_00: And that's an example.
• 35:26
  SPEAKER_00: I'm not going into the full for workflow because we'll have a specific prompt for each step.
• 35:31
  SPEAKER_00: Makes sense, but only to give an idea how it works.
• 35:34
  SPEAKER_00: Is it clear?
• 35:35
  SPEAKER_01: Yeah, I think uh so your chain linking, even if you just want to understand the context of where a company operates and whether the company is investable or not, doesn't really matter.
• 35:46
  SPEAKER_01: You always learn something about the industry, about the company, about a competitor, and that will help you with a future research report or company.
• 35:54
  SPEAKER_01: I think that makes sense, but I found it interesting that you start with the industry and not the company itself.
• 36:02
  SPEAKER_01: Do you do that because you want to understand where they're operating and then see why the company operates in that specific niche or segment?
• 36:14
  SPEAKER_01: Or why do you start with the industry?
• 36:16
  SPEAKER_00: That's also taken from my uh investing expertise.
• 36:20
  SPEAKER_00: Let's say AI is only a tool and it mirrors your expertise and how you think about investing.
• 36:26
  SPEAKER_00: From my side, I think that most of the returns come from the industry.
• 36:32
  SPEAKER_00: There is good industries and bad ones.
• 36:34
  SPEAKER_00: There is industries that are who needs a lot of uh investments, they are uh cyclical, and the returns are not exceptional if you check on long-term horizon.
• 36:44
  SPEAKER_00: And for this, I always start from an industry to understand it because whatever the stock, 80% of the return will come from the industry, and the 20% if you have from the stock uh name.
• 36:58
  SPEAKER_00: At the end of the day, I'm talking about long-term investing.
• 37:01
  SPEAKER_00: Of course, you can invest in a shitty industry, and I have did this, and uh and it works also, but that's uh that's trade in multiples, that's not um investing in the future of stock.
• 37:13
  SPEAKER_00: So for this, I believe that understanding is important.
• 37:17
  SPEAKER_00: So I start like this, so it mirrors my investing style.
• 37:22
  SPEAKER_01: So when do you figure out if a company is within your circle of competence or not?
• 37:26
  SPEAKER_01: After the listening to the industry and the business model, do you figure it out then or do you go a little deeper or when when do you cut it off?
• 37:35
  SPEAKER_00: If it's okay for me, it's a business that I understand, it looks interesting, it's not 100 times sales.
• 37:42
  SPEAKER_00: I would not buy that and I will go deeper.
• 37:44
  SPEAKER_00: I will download on your report and try to read it like I do, but I go a lot faster because I I have understood a lot of things, and what I will do is that I upload it into Notebook LM and ask deep questions.
• 37:57
  SPEAKER_00: For example, I will use I will read reports sometimes.
• 38:01
  SPEAKER_00: Let's say let's be honest, we don't have time.
• 38:03
  SPEAKER_00: You will not read the 10 last reports.
• 38:07
  SPEAKER_00: Buffit does that honestly.
• 38:10
  SPEAKER_00: I read the reports, the last one, the two, but I really have time to read the 10 last reports.
• 38:17
  SPEAKER_00: Of course, I see the data, but I don't read them from A to Z, the last 10th, even before EI.
• 38:23
  SPEAKER_00: So, what I will do is that I will give them to an outbook LM and ask questions about timeline.
• 38:29
  SPEAKER_00: I will give him the last 10th and ask questions like what did management said from the last 10 years and for each year's, what did happen?
• 38:38
  SPEAKER_00: For example, if management in 2019 said we will have 10% growth for the next three years, check fact-check that what did happen in reality, and also what was the explanation?
• 38:51
  SPEAKER_00: If I see in 2019 they promised 15% growth, and they had in 20, let's say there is no COVID because it changed all the all the things, let's say that uh it's normal year, and they had uh 10% and they explain it with economy is not good, and next year also they made the same promise and they explain it economy is not good.
• 39:12
  SPEAKER_00: I know that I can't trust the the the management while talking about this, and that's kind of things that AI make it very easy and quick to see that will take us a lot of time manually.
• 39:24
  SPEAKER_00: Other example also is to say to see the evolution of the the the products and the segments inside the categories inside the business model.
• 39:35
  SPEAKER_00: You can give 10 years and the notebook or another LLM will summarize this in a very visual way and easy way that took us a lot of time.
• 39:45
  SPEAKER_00: I will never say AI will do all the work.
• 39:48
  SPEAKER_00: We need to keep all the work and the valuation also part is from our side because it's easy.
• 39:54
  SPEAKER_00: I don't like DCF, I don't do DCAF, I use uh free cash flow yield plus the growth to estimate the return, so I don't need a computer to help me to do that.
• 40:04
  SPEAKER_01: Yeah, yeah, I think uh that's one of the things Buffett always said.
• 40:07
  SPEAKER_01: I don't I just need a napkin, right?
• 40:09
  SPEAKER_01: To uh to come up with a with a fair value.
• 40:12
  SPEAKER_01: Uh I would like to go a little deeper on prompting uh in a second, but before we we go there, something that came up while you were explaining um you know, understanding the company and uh you you're sort of explaining the calibration where you test LLMs and new ideas on companies you already know, so you know where it's hallucinating or doesn't work properly.
• 40:38
  SPEAKER_01: But if everyone has access to these tools and eventually learns how to use them properly, it means everybody finds the same information, right?
• 40:48
  SPEAKER_01: So to generate alpha in the stock market, you gotta be right, while others gotta be somewhat wrong.
• 40:55
  SPEAKER_01: If everybody is right all the time, the stock prices will always reflect the perfect current valuation, so you won't generate excess returns.
• 41:05
  SPEAKER_01: So, how do you think about that?
• 41:07
  SPEAKER_01: You're sort of shooting yourself in the foot, Mustafa, by sharing all your prompting, everybody finds good information.
• 41:14
  SPEAKER_00: No, no.
• 41:15
  SPEAKER_00: The beauty of investing is that first it's not uh some zero gain.
• 41:21
  SPEAKER_01: Agreed.
• 41:22
  SPEAKER_00: Everybody can have good returns if the companies are growing profitably, of course.
• 41:28
  SPEAKER_00: But the main idea is that first we are using AI.
• 41:33
  SPEAKER_00: We think that everybody is using AI.
• 41:35
  SPEAKER_00: Factually, that's not true.
• 41:37
  SPEAKER_00: A lot of people believe that's not useful and they're not using so for the moment we have an edge to go in faster with AI, but that will not last uh more than two two years, let's say.
• 41:50
  SPEAKER_00: But at the end of the day, you need to master this tool to go fast like everybody.
• 41:56
  SPEAKER_00: But the second thing, and the most important in investing the timeline horizon is the biggest edge for me.
• 42:07
  SPEAKER_00: Sometimes we know that stock will drop to this year because the there is a specific element.
• 42:14
  SPEAKER_00: For example, uh they are working on restructuring the business, they will have a lot of cost, and so the margin will be bad this year.
• 42:22
  SPEAKER_00: All institutional investors that represent 80 or 90 percent of the market will not buy that stock, they can't because they have to feed their children, they have uh they have work to do, and they prefer going having say market return instead of taking a big contrarian bet that may work or may not.
• 42:45
  SPEAKER_00: But us as investors that have a long-time horizon, we can understand business with AI and then take a decision to say, okay, I know that this year will not be okay, but that's an opportunity for me to invest in it.
• 42:59
  SPEAKER_00: Also, it's our analytical capacities.
• 43:04
  SPEAKER_00: Let's say, for example, AI will explain you all the regulation risk.
• 43:09
  SPEAKER_00: For example, of a company X or Y, you have risk of regulation.
• 43:13
  SPEAKER_00: It's up to us to say, for me, maybe I will say, Okay, but with the valuation, we are five time free cash flows.
• 43:20
  SPEAKER_00: I accept to take the risk for that, and you will say, Okay, no, for me, that's also that I will not take that risk.
• 43:27
  SPEAKER_00: And at the end of the day, in three years, maybe I will have good returns or not.
• 43:32
  SPEAKER_00: And that the decision was because I said that I will accept that risk.
• 43:37
  SPEAKER_00: So at the end of the day, AI is a tool to grow fast and uh everybody will use it in two or three years.
• 43:43
  SPEAKER_00: That's not the case today, and a lot of people are using it wrongly, also.
• 43:47
  SPEAKER_00: But even if uh in five years everybody will be using it, the returns will come from our expertise and how we think about things, and also our time horizon.
• 43:57
  SPEAKER_00: Uh, if if uh someone Take all of my inputs and use AI for everything that is even the best expert in the world.
• 44:06
  SPEAKER_00: But he thinks in one year time horizon in investing, you cannot uh work like this.
• 44:13
  SPEAKER_00: I don't believe in trading and such things, so sorry guys, but uh it will not work.
• 44:21
  SPEAKER_01: Yeah, you know, I agree that investing is not a uh a zero sum game, but there is a big, there is a there is a but.
• 44:29
  SPEAKER_01: You know, it's a positive sum game because as companies innovate and grow, the entire economic pie gets bigger and we can all win together.
• 44:39
  SPEAKER_01: But if we are trying to beat the market, that is mathematically a zero sum.
• 44:46
  SPEAKER_01: Because for you to outperform the average, someone else must underperform.
• 44:51
  SPEAKER_01: So that's what I meant with um um with finding the right companies, and if everybody has the perfect process.
• 45:00
  SPEAKER_01: Yeah, I wonder how that works, but it's unlikely since not everybody will have the same process or beliefs, and there's always people investing differently and not having the behavioral edge, right?
• 45:12
  SPEAKER_00: So uh and I don't remember who who said who say this, but maybe Boffett or another investor.
• 45:18
  SPEAKER_00: The stock market is a machine to transfer money from the the patient to uh from people with short uh with the from people with the short-term horizon to people with long-term horizon, and whatever the tools, we will have some people that will only focus on the next quarter six months.
• 45:34
  SPEAKER_00: Yeah, that's true.
• 45:35
  SPEAKER_00: So this we will always win if we are long-term investors.
• 45:38
  SPEAKER_01: Yeah, agreed, agreed.
• 45:39
  SPEAKER_01: So let's talk about prompting.
• 45:41
  SPEAKER_00: Yeah.
• 45:43
  SPEAKER_01: You know, is it correct for me to say that prompting is the next valuable skill?
• 45:53
  SPEAKER_00: Prompting is a valuable skill for sure.
• 45:57
  SPEAKER_00: Is it the ultimate one?
• 45:59
  SPEAKER_00: No, no.
• 46:00
  SPEAKER_00: The best skill with AI is to be able to think correctly with a paper and pen.
• 46:05
  SPEAKER_00: When I was explaining the investment process why I use AI in that structure, because I took time to rewrite my investing process, all the questions that I have.
• 46:16
  SPEAKER_00: When I'm inside my prompt, I have all questions that I put because I thought about it.
• 46:22
  SPEAKER_00: Okay, Mustafa.
• 46:23
  SPEAKER_00: Last time that we I worked about my this stock, how what questions did it ask?
• 46:28
  SPEAKER_00: What do I have inside my checklist?
• 46:30
  SPEAKER_00: What I what type of question I ask in that NST or another one?
• 46:34
  SPEAKER_00: And I write them.
• 46:35
  SPEAKER_00: Even if you have the best practices of prompting, but not the right questions that came from your experience, from you taking time to think about it, prompting will not add an added value.
• 46:46
  SPEAKER_00: But once you have done the basic work of putting your knowledge on a pen and with on a paper with a pen, later the prompting is very important because he will make you get the most of your knowledge with AI.
• 47:01
  SPEAKER_00: And the best practices are things to follow to have the right prompt.
• 47:05
  SPEAKER_00: For example, if if we go into some examples, you need always to have one prompt, one job.
• 47:13
  SPEAKER_00: Never ask AI to have all details answered on everything in one session, because there is limits in technically in token and things like that, but there is technical limits, let's say first, and the second is AI is will mix a lot of things and will not go deeply into things.
• 47:32
  SPEAKER_00: So keep in mind one prompt, one major jump joke.
• 47:36
  SPEAKER_00: For example, I will not say I want to understand the business model, the management, and the valuation in one shot.
• 47:43
  SPEAKER_00: No.
• 47:43
  SPEAKER_00: If I want an overview, that's one prompt.
• 47:46
  SPEAKER_00: If I want management, that's one prompt.
• 47:48
  SPEAKER_00: If I want to go deeper in one one of the my checklist points, I will I will have one prompt.
• 47:56
  SPEAKER_00: One of the best practices is also to give a role.
• 48:00
  SPEAKER_00: So AI has been trade on everything.
• 48:03
  SPEAKER_00: Let's say AI is a good writer, good investor, and also a good uh mathematician.
• 48:10
  SPEAKER_00: When you tell him you are a mathematician, he will use the sources linked to the mathematics and not the others.
• 48:17
  SPEAKER_00: That would that help a lot.
• 48:19
  SPEAKER_00: There is a lot of things to take into account, like structure, asking how to use AI and applying these best practices of prompting, will make most of our knowledge that we put uh in uh on on a paper from AI.
• 48:35
  SPEAKER_00: But at the end of the day, you need to write your process, you ask questions, learn how to prompt, and then test also.
• 48:44
  SPEAKER_00: Because while testing, you will see sometimes uh AI tend to, for example, that's a real example.
• 48:50
  SPEAKER_00: I have been using Notable LM a lot, and sometimes he proposed automatic report to summarize business based on the all the the reports that you put without a specific prompt.
• 49:01
  SPEAKER_00: And without specific prompt, for example, I had a business that's uh software who who does 80% of his revenue from one segment and 20% from the other.
• 49:13
  SPEAKER_00: But AI kept talking like it's the same value for both segments.
• 49:18
  SPEAKER_00: At the end of the day, I don't give the same value for both because most sort of money comes from segment one.
• 49:25
  SPEAKER_00: So in my prompt, I precise if there is a lot of products, focus more on the ones with the bigger revenues or margins, and like this while iterating and testing, you have something that works better.
• 49:41
  SPEAKER_01: So, for the listeners wanting to test prompting, what makes uh you mentioned using a role, for example, being specific, so not asking too much in one question, maybe give a couple short examples or what do investors need to try for using a good prompt versus a bad prompt?
• 50:04
  SPEAKER_00: Yeah, let's say, for example, you are talking about uh about uh let's say SML, for example, ASML, semiconductors.
• 50:14
  SPEAKER_00: If you know that you are working on this sector, on the role, you can say you are semiconductor equipment analyst.
• 50:20
  SPEAKER_00: It will be more focused than you are an analyst.
• 50:24
  SPEAKER_00: That that's one example.
• 50:26
  SPEAKER_00: Also, for for example, for the sources you say inside your prompt, use only uh official sources and do not use uh Reddit or other blogs, for example, because you know that you can have uh it's up to you.
• 50:42
  SPEAKER_00: I don't want to use ideas for from Reddit, for example.
• 50:45
  SPEAKER_00: Also, talking about the structure of the output, I always say give me an output detailed report in this format, start with an executive summary because I want to, when I start reading, I will have uh uh some couple of sentences summarying everything in and then explain for me.
• 51:05
  SPEAKER_00: So the let's say the overview, explain the business model and then how they make money, who are customers, who take the decision, is it B2B, B2C, and give all the structure of the output, and this will help us get something similar each time, and like this we can compare and progress.
• 51:22
  SPEAKER_00: So, what I recommend is that I I share a lot of prompts for free uh in my sub stack.
• 51:29
  SPEAKER_00: Take one of the examples, try to analyze it and reverse engineer it because there is a lot of best practices, but there is also invest in experience and tests that makes good prompts.
• 51:42
  SPEAKER_00: But I'm always sharing free prompts um with in on my uh page in my notes, in my free newsletters also.
• 51:49
  SPEAKER_00: Of course, I share others on paid well, but uh, but you can the value is there also.
• 51:53
  SPEAKER_00: You can you can use them reverse engineer and learn.
• 51:56
  SPEAKER_01: Yeah, yeah, and I think that's uh that's an excellent tip because uh your investing style is not the same as uh everybody else's, so uh you gotta tweak it and play around with uh whatever style of investing or sort of companies you look for.
• 52:11
  SPEAKER_01: To invert the question, let's uh Charlie monger it for a second.
• 52:15
  SPEAKER_01: What is something you don't use AI for?
• 52:18
  SPEAKER_00: Numbers today.
• 52:20
  SPEAKER_00: AI is not best, they are quite good, getting good, especially with cloud.
• 52:25
  SPEAKER_00: But uh, if I want to do calculation um deep, I don't trust AI with numbers today.
• 52:32
  SPEAKER_00: I'm not talking about uh gross margin to say gross margin is 10%, uh is 80%, uh, 10% I like big modes, so I'll say that you have big numbers and also about um all simple numbers that will be okay.
• 52:47
  SPEAKER_00: But if you want all complex calculation DCF things, I will not trust AI with that.
• 52:53
  SPEAKER_00: Today they are getting better on numbers, but they are not very good yet.
• 52:57
  SPEAKER_00: But the good news, as investors, we need only to know how to do simple additions and and a couple of percentages.
• 53:04
  SPEAKER_00: We don't need deep mathematical calculation.
• 53:08
  SPEAKER_00: My analyst friends will not be happy with with my what I'm saying.
• 53:11
  SPEAKER_00: But sorry, sorry guys, you don't need uh complex mathematic for for investing.
• 53:17
  SPEAKER_01: You need some basic understanding of the numbers, yeah.
• 53:20
  SPEAKER_00: Yeah, but that's not that's not but I'm an engineer also, so maybe.
• 53:26
  SPEAKER_00: And that's one example.
• 53:28
  SPEAKER_00: The numbers, I will not use AI for it.
• 53:30
  SPEAKER_00: I will not use AI to take decisions that will be never thing that you need, you never trust AI taking decisions, you take the decision.
• 53:39
  SPEAKER_00: It's like an intern, it's like an analyst that you are not paying much and doing a lot of work, but that's not the idea, is not to take decisions.
• 53:49
  SPEAKER_00: I will not use it uh to replace part of my processes.
• 53:55
  SPEAKER_00: For example, if I you always try to understand the business, then go into sector business, so go into uh my checklist and go the steps.
• 54:04
  SPEAKER_00: I will not outsource completely a step to AI.
• 54:08
  SPEAKER_00: For example, I will never say, okay, AI did summarize the business model, I will go directly into the next step because I have have automated all of this step.
• 54:18
  SPEAKER_00: No, I will review that and check it also.
• 54:22
  SPEAKER_00: And we say it in put always human in the loop because AI can make small mistakes and to note compiled errors, try to always check the results and and validate them.
• 54:35
  SPEAKER_00: That's a couple of uh of uh points where I will not use AI today, but things go faster that this can change uh in next week.
• 54:43
  SPEAKER_00: We don't know.
• 54:45
  SPEAKER_01: Can you uh use AI to analyze intangibles like management quality, maybe things like brand strength or company culture?
• 54:56
  SPEAKER_00: Ironically, that's one of the best use cases of AI.
• 55:00
  SPEAKER_00: Management, for example, the track record, how they have uh worked in the past years, you with the right prompt or right question, you can do that.
• 55:08
  SPEAKER_00: Let's take an example.
• 55:10
  SPEAKER_00: First thing I want two things of the in management.
• 55:13
  SPEAKER_00: I want them to have a skin in the game.
• 55:15
  SPEAKER_00: That's easy to check.
• 55:16
  SPEAKER_00: If they have stock or not, are they selling or buying?
• 55:19
  SPEAKER_00: That's easy to check.
• 55:20
  SPEAKER_00: The second thing is that are they long-term investors, long-term mindset or not?
• 55:27
  SPEAKER_00: And that's based on their decisions.
• 55:29
  SPEAKER_00: I will ask AI to track all decisions and compare and analyze if it is uh for it has taken into long-term strategy, thinking about long-term, or only to improve uh to improve uh also about the brand.
• 55:43
  SPEAKER_00: Let's take an example.
• 55:46
  SPEAKER_00: Uh, let's let's take Starbucks, for example.
• 55:50
  SPEAKER_00: We talked about it.
• 55:51
  SPEAKER_00: You want to to have an idea about uh what's the brand image today.
• 55:58
  SPEAKER_00: You can ask your friends, you will have uh 10-15 friends, maybe if you are too popular, 50, not more.
• 56:06
  SPEAKER_00: So the your statistically, you will not have a good understanding of the bread image only with asking your friends.
• 56:13
  SPEAKER_00: What you can ask AI to do is go into now you can go into Reddit onto Facebook and to all discussions with the where people talk about the ideas, and you will have more statistically significant numbers, and you will see 1000, 2000 people talking about a brand, and you will see that's is what does people think about that brand, and that could be early indicators because even if today numbers are good, if people start complaining and a lot of people start complaining about uh a brand, maybe it's an early indicator that's uh that's not something to do.
• 56:52
  SPEAKER_00: Also, I like to check if the company treats well their employees.
• 57:00
  SPEAKER_00: Maybe I live in not a realistic uh world, but I deeply believe that most of businesses are good because people work in them there are good.
• 57:09
  SPEAKER_00: The capital knowledge, how people are feeling in the day-to-day operations is very important.
• 57:16
  SPEAKER_00: And I want to understand the company culture, and that can be done with AI.
• 57:20
  SPEAKER_00: Check in the GlossDor and all the website specialists to re the the get the uh to understand the culture of the business, and I can do three on um whatever LLM in deep research mode to go and search for all of things.
• 57:37
  SPEAKER_00: Of course, that's what I'm trying to do.
• 57:40
  SPEAKER_00: If you can see the pattern, is go in faster, automating things.
• 57:44
  SPEAKER_00: But at the end of the day, the AI will give me culture, give me management, give me information, and it's up to me to say the culture is not good, but that's not a big problem because let's say we are not we are in a monopoly uh company and we don't share market, and I don't know that it's not very important, but in this case, no brand is important, so I will not invest in this stock or its risk.
• 58:09
  SPEAKER_00: At least I know it's a risk.
• 58:11
  SPEAKER_00: But it's up to the investor to take the decision.
• 58:15
  SPEAKER_01: What is the biggest mistake you see new users adopting AI fall into when they uh start applying AI to their investing process?
• 58:25
  SPEAKER_00: First one is not taking time to learn.
• 58:28
  SPEAKER_00: Sometimes people say it's working or it's not working in five minutes.
• 58:34
  SPEAKER_00: That's too arrogant, let's say, to take decisions and say, I have a friend, he tried ChatGPT in 2023, and until today he will tell me, no, AI is not useful.
• 58:48
  SPEAKER_00: Bro, you have used it in 2023, everything did change.
• 58:51
  SPEAKER_00: And last time I heard someone saying to me, But AI doesn't have access into internet.
• 58:56
  SPEAKER_00: Yes, the first model didn't have access to into the internet because they weren't trained until uh year 2023, for example, and then they don't have real-time access.
• 59:07
  SPEAKER_00: But that was in 2023.
• 59:09
  SPEAKER_00: Try to always understand, and that's not only about AI, it's about all other aspects of life, always accept learning and checking things.
• 59:21
  SPEAKER_00: I will never say AI is not good for something without trying it.
• 59:26
  SPEAKER_00: And first reflex if I can't make AI work for something, maybe I don't have the right process, maybe I don't have the right prompt.
• 59:34
  SPEAKER_00: If I try a lot and doesn't work, okay, today AI is not ready to do this.
• 59:39
  SPEAKER_00: But it will be decision based on facts and not a feeling and an idea that I had in um 2023 or uh because things are are moving very fast, and we can have there is a lot of mistakes, that's the biggest one, not trying to learn.
• 59:55
  SPEAKER_00: Second one, and I I see it also with my uh investors with more experience.
• 1:00:00
  SPEAKER_00: I have around nine years experience because I'm younger, I'm uh 33 years old and uh happily married, and I have a daughter of three years.
• 1:00:12
  SPEAKER_00: And uh I have some friends that are who are older, they are been investing in 20, 30 years in markets.
• 1:00:20
  SPEAKER_00: But for them, it using AI is feels like cheating, and they will tell me uh I have experience, that's not for me.
• 1:00:28
  SPEAKER_00: For me, it's like um saying I will not use calculator because I'm I'm good at mental math.
• 1:00:35
  SPEAKER_00: Yeah, okay, but at the end of the day, you need to go faster.
• 1:00:39
  SPEAKER_00: So that's kind of mindset errors where people don't want to go or they have they have some faulty beliefs that are blocking them.
• 1:00:49
  SPEAKER_00: But once that's okay, you need to build your skill and learn basics, understand the basics of AI because one of the big errors is that if you don't understand how AI works, you cannot trust it.
• 1:01:03
  SPEAKER_00: You need to try to take time, understand the basics, and don't go directly.
• 1:01:08
  SPEAKER_00: One of the most errors is that someone will go into uh he will copy prompt, run it, and okay, you will not learn like this.
• 1:01:16
  SPEAKER_00: You need at least to take some time to understand the basics.
• 1:01:19
  SPEAKER_00: One of the errors also is to to say AI is magic and expect AI to do everything, and then be disappointed when you don't have the best results in one shot, in one prompt.
• 1:01:32
  SPEAKER_00: Understand what AI is capable of doing.
• 1:01:35
  SPEAKER_00: That's what we have been talking, and thanks for uh giving the opportunity for people to understand the these ideas and then use it correctly.
• 1:01:44
  SPEAKER_00: So the main errors will be more uh behaviorals and not technical, because technical things you will learn them while using and trying, if you want, of course.
• 1:01:56
  SPEAKER_01: Yeah, you're basically talking about a growth and a fixed mindset, right?
• 1:01:59
  SPEAKER_01: So being a lifelong learner, and that's something we aspire to be.
• 1:02:04
  SPEAKER_01: Beautiful.
• 1:02:04
  SPEAKER_01: So let's wrap this up.
• 1:02:06
  SPEAKER_01: But before we leave, I always ask uh the question for our listeners to give their single biggest tip, idea, advice related to investing.
• 1:02:15
  SPEAKER_01: It can be anything, something you've learned over the years, a mistake you've made, whatever it may be.
• 1:02:20
  SPEAKER_01: So, what's the one thing you want the listeners to take away from this episode?
• 1:02:26
  SPEAKER_00: Try AI for investing.
• 1:02:28
  SPEAKER_00: It works.
• 1:02:29
  SPEAKER_00: Believe me, it works.
• 1:02:31
  SPEAKER_00: I have been using it, I have a lot of friends that have been using it, and I'm not talking about people who start investing.
• 1:02:37
  SPEAKER_00: A big network of people with long time, long experienced professionals, and they all benefited from AI.
• 1:02:44
  SPEAKER_00: You can do the same, but you need to take time, learn about it, make it link to your investing process, and uh use and learn, try it.
• 1:02:57
  SPEAKER_00: Uh, there is sometimes we see Steve Jobs taking a talk, and you'll see he's very good, or you see someone good at something.
• 1:03:07
  SPEAKER_00: But the first time that that uh Steve Jobs went on stage and talked, he was not that good.
• 1:03:13
  SPEAKER_00: Well, that that's not true.
• 1:03:14
  SPEAKER_00: Have been doing presentation 30-50 times, he has prepared 100 hours before doing that presentation.
• 1:03:21
  SPEAKER_00: For this, he is very good on that.
• 1:03:24
  SPEAKER_00: That's the same thing.
• 1:03:25
  SPEAKER_00: You will see an expert using something, they will say, you will say, he's very good, maybe he has natural talent.
• 1:03:33
  SPEAKER_00: No, you need to take time, learn the tool, and try it, test and learn and progress.
• 1:03:39
  SPEAKER_00: And at the end of the day, you will acquire a skill that you will use into your investing process.
• 1:03:44
  SPEAKER_00: I deeply believe that we, for me, at least in my case, I am an investor because I want to have a better life.
• 1:03:52
  SPEAKER_00: I want to have more time with my family, I want to have uh better quality time with my uh my my uh my family and my uh my children.
• 1:04:02
  SPEAKER_00: And AI will make us to go faster in order to have more time with our families.
• 1:04:08
  SPEAKER_00: That's for me, that's uh maybe I'm starting to be uh older, 33 years old, but that's my thinking.
• 1:04:16
  SPEAKER_00: So at the end of the day, I will be very happy if someone takes action and learn how to use AI for investing and will have more time with his family because and avoid also errors because let's say if we were running a prompt about uh the big red flags and you identify red flag and you don't go and you provide your capital, I will be really, really happy from that.
• 1:04:41
  SPEAKER_01: The tool is only as good as the person using it, right?
• 1:04:44
  SPEAKER_01: It's exactly the art it's the archer, not the arrows.
• 1:04:47
  SPEAKER_01: Exactly.
• 1:04:48
  SPEAKER_01: Yeah, beautiful.
• 1:04:49
  SPEAKER_01: Where can people find more on you, Mastafa?
• 1:04:52
  SPEAKER_01: Where uh can they find more on using AI, uh becoming a better investor?
• 1:04:57
  SPEAKER_01: Uh feel free to share your socials and I'll uh put them in the description as well.
• 1:05:02
  SPEAKER_00: In my newsletter, compound with AI, it's on Substack, but you and I share a lot of notes, free notes with prompts with uh framework.
• 1:05:12
  SPEAKER_00: I share also some free newsletters, and of course I have my Patreon, but I'm talking about the free things where you can learn a lot of things.
• 1:05:21
  SPEAKER_00: And if you have questions, don't hesitate to contact me.
• 1:05:25
  SPEAKER_00: I I'm always in touch with people if I have time to answer things, and honestly, I I will recommend everyone to take time to learn, even if it's not from me, from someone else, just take time to learn AI for investing or for for other aspects of life.
• 1:05:43
  SPEAKER_01: Well, thank you so much for taking the time to share this.
• 1:05:46
  SPEAKER_01: I know you're doing this out of your passion, and uh you're taking time out of your day to to share this with us.
• 1:05:52
  SPEAKER_01: So we appreciate it a lot, and uh I hope we'll uh we'll stay in touch.
• 1:05:56
  SPEAKER_00: Yeah, thank you.
• 1:05:57
  SPEAKER_01: See you in the next one.
• 1:05:58
  SPEAKER_00: Goodbye.