He Creado un Segundo Cerebro con IA… y Es una Locura! — Transcript
Full transcript
- 0:00This is crazy. Imagine for a moment
- 0:03that you could create a second brain, a
- 0:06complete neural network of all the
- 0:09concepts that you want the AI to have
- 0:11and that they feed off each other so
- 0:14that every time the AI needs to work
- 0:17with you, it is an expert who has all
- 0:19the information at its disposal. And
- 0:22this is not a method that I have
- 0:25invented, it is the step-by-step method
- 0:27shared by the employee of Antropic
- 0:30Carpati, who was the co-founder of Open
- 0:33AI and who has provided the method to
- 0:35train the AI and create this second
- 0:38brain. And this has many implications,
- 0:41from being able to have your entire
- 0:43business to helping you understand a
- 0:45company with all its concepts, how they
- 0:48align with each other, all the
- 0:50information that we have available,
- 0:53digested and ready to be used every
- 0:55time we use artificial intelligence.
- 0:57Well, one of the things that I was
- 1:00doing was using the most important
- 1:02academic literature in finance so that
- 1:05it could synthesize it all and create
- 1:07an expert with all the important
- 1:10knowledge of the financial world and
- 1:12how all these concepts are related to
- 1:15each other. Now, academic papers are a
- 1:17bit boring and we are going to try to
- 1:20create a kind of second brain with all
- 1:22the knowledge of the last decades of
- 1:25the best investor of all time and that
- 1:27way we will have all the folders with
- 1:29all the concepts organized that we can
- 1:32read through Obsidian. I am going to
- 1:34explain to you step by step and the
- 1:36concept of Carpaty's method so that you
- 1:38can replicate it in anything you want.
- 1:41Industry analysis, company analysis,
- 1:43anything you need to study, everything
- 1:46to be able to create the right
- 1:48architecture to work with AI. Look, the
- 1:51main problem with how we use artificial
- 1:54intelligence today is that surely you
- 1:56and I and everyone we do is that we end
- 1:58up with the one we use. For example, I
- 2:00usually use the cloud, we have several
- 2:02videos on the channel of practical
- 2:03things that I have done and we give it
- 2:05all the information. For example,
- 2:06imagine that we are analyzing a company
- 2:09and we upload the latest reports of the
- 2:11company, the 10K, the 10Q, an analysis
- 2:13of experts and different things and we
- 2:15ask it to analyze. The problem with
- 2:17this is that the context remains in
- 2:18this conversation and when the
- 2:20conversation ends, that context tends
- 2:22to disappear. It is true that the AI
- 2:24can save memory and recover certain
- 2:26concepts in other conversations, but
- 2:28all those files that you have given it
- 2:30have not been synthesized and,
- 2:31therefore, the next time you want to
- 2:33ask it something you would have to
- 2:35upload these files again so that it
- 2:36could work with them again, which
- 2:38involves a great consumption of tokens
- 2:40and, obviously, repeating everything,
- 2:42uploading the files again; or you could
- 2:44make a project with all the files
- 2:46inside, but each time you ask it about
- 2:48something, the AI is going to have to
- 2:49look at those files over and over again
- 2:51. And that is the problem that Carpati
- 2:54raised a few months ago. He said that
- 2:55the problem is that everyone gives it a
- 2:57collection of files, different things
- 2:59and that this works, but that each time
- 3:01you have to start from scratch. There
- 3:03is no accumulation. Every time you ask
- 3:05it a new question, the AI has to find
- 3:07the relevant pieces and fragments, and
- 3:09here he emphasizes that nothing is
- 3:11being built, but proposes a different
- 3:14idea. Instead of just giving it the raw
- 3:17documents each time, what we can do is
- 3:20create an architecture that constantly
- 3:23builds on top of it and updates pages,
- 3:26reviews topics, looks for contradictory
- 3:29data to fix it or even improve the
- 3:32system. All this sounds a bit technical
- 3:35, but let's bring it down to earth. And
- 3:37the solution that Carpat proposes is
- 3:38the one that he himself is implementing
- 3:40, since he shared this in a tweet as
- 3:42something personal, saying, "Hey, this
- 3:44is how I work." And it went extremely
- 3:46viral. That led Carpaty to share the
- 3:49guide, the structure of how he does it.
- 3:51He published a pattern that is not a
- 3:53product or code. It is a way of working
- 3:56, especially for those of us who work
- 3:58with information, with many documents.
- 4:00Instead of having the AI reread your
- 4:02documents for every question, you build
- 4:04a persistent wiki once, a complete
- 4:06three-layer structure that I'm going to
- 4:08show you now and we're going to do it
- 4:10step by step, and you maintain it
- 4:11forever. It says, for example: what
- 4:14gets interesting is that once your wiki
- 4:17, that brain, is so big (for example,
- 4:19mine has about 100 articles and over
- 4:22400,000 words), you can ask your AI
- 4:24agent any complex question about that
- 4:27subject, because it has access to that
- 4:29entire directory. And this basically
- 4:32works with a three-layer architecture
- 4:34and it may seem very technical, but I'm
- 4:36going to explain it very simply.
- 4:37Observe layer 3, where we would have
- 4:39our sources, our documents, and all of
- 4:41our raw files. Here what I'm going to
- 4:44do is dump all of Warren Buffett's
- 4:47letters. Those are raw files in PDF, in
- 4:51Word, in whatever, that are not
- 4:54modified. That folder is ours, it's
- 4:56where our files are saved. Now, this is
- 4:58where it gets interesting. That local
- 5:01folder will have a cloud.md, MD, which
- 5:03is basically going to be a file that
- 5:06tells Cloud how to work within that
- 5:08folder, within that brain, if you want
- 5:11to call it that. Cloud MD will know
- 5:13exactly your tastes, your way of
- 5:15working, your way of analyzing, in this
- 5:17case, the market, your investment
- 5:19philosophy and will adapt its work to
- 5:21that context to get to know you and
- 5:24know how it can best help you. And in
- 5:26layer two we have the wiki. The wiki
- 5:28belongs to the AI. That's where it
- 5:30creates, updates and maintains. In
- 5:32other words, what the AI is going to do
- 5:34is that each raw file that we give it
- 5:36is going to be ingested, it is going to
- 5:38be synthesized and it is going to be
- 5:40able to create pages that then allow it
- 5:42to have that information always
- 5:44available to it quickly and how that
- 5:46information relates to all the rest.
- 5:48Look, here we have the diagram in a
- 5:49little more detail and now when we do
- 5:51it step by step you are going to
- 5:52understand it much better. Folder one
- 5:53contains all our files. We create this
- 5:56on our computer and we can create the
- 5:58raw files we want. Layer number two
- 6:00belongs to artificial intelligence.
- 6:02This is where the artificial
- 6:03intelligence will create those pages,
- 6:05those concepts and summarize the
- 6:07information. And it will also have an
- 6:09index and a log. And Carpati says this
- 6:10is very useful for the following. The
- 6:12index will basically be a file that is
- 6:14basically like the index of a book. It
- 6:16tells the artificial intelligence
- 6:18everything that is in this layer two
- 6:19and it basically organizes it. Each
- 6:21page is listed with its link, a quick
- 6:23summary of what each page shows,
- 6:25different concepts, sources, etc. What
- 6:27the artificial intelligence will do is
- 6:29first read the index and then search
- 6:31within all the files for the relevant
- 6:33pages. And then the log is basically a
- 6:35file that is kept up to date, which is
- 6:38chronological and where it will put
- 6:39everything that has been done,
- 6:41everything that has been changed within
- 6:43this brain. That way we always have a
- 6:45history of all the changes that have
- 6:47been made. With this, what we achieve,
- 6:49people, is a brutal context, that every
- 6:51time we work with AI, it has at its
- 6:53disposal all that brain, those
- 6:55connections that we have seen to be
- 6:57able to gather information or help us
- 6:59in the best possible way. According to
- 7:01Carpati, what are the applications of
- 7:03use? Well, it can be personal to have
- 7:05your goals, health, personal
- 7:06development topics, you can keep track,
- 7:08for example, if you keep a diary,
- 7:10articles, podcasts. It also helps you
- 7:12with research and this is the most
- 7:14interesting part in my opinion because
- 7:16it allows you to create a brain on a
- 7:18topic that you want to investigate and
- 7:20to be able to gather a lot of
- 7:21information and synthesize it. For
- 7:23example, if you are studying something
- 7:25specific or you are learning about a
- 7:27specific industry, about a sector,
- 7:29about a topic, well, you can create a
- 7:31kind of brain focused on this. The same
- 7:33with books, the same with teamwork or
- 7:36even business. So now we are going to
- 7:38see what application we can give it. As
- 7:40I was telling you, I've been
- 7:42experimenting with academic papers to
- 7:44get the concepts down, to have detailed
- 7:46information from all the academic
- 7:47studies out there and the real
- 7:49foundations of finance, but now we're
- 7:51going to do it with something a little
- 7:53cooler because I know all this is very
- 7:55boring. Let's do it with Warren Buffett
- 7:57. We're going to try to create a second
- 7:59brain that compiles all those years of
- 8:01teachings that Buffett has written in
- 8:03his annual letters. And how do you work
- 8:05? What platforms do you work with? Well
- 8:07, Carpati typically uses Obsidian,
- 8:09which is, at the end of the day, just a
- 8:12reader for MD notes files, isn't it?
- 8:14Markdown. The good thing about it is
- 8:16that if Obsidian ever disappears, those
- 8:18Markdown files are text files that any
- 8:20computer can read, so you could use
- 8:22them in another application. And
- 8:23Obsidian is basically just that note
- 8:26file reader that also makes that
- 8:28connection map for us that we can then
- 8:30have. But Obsidian basically only reads
- 8:33Bolt, which is this architecture. And
- 8:35then with Cloud Code, personally,
- 8:36that's who I'm going to work with to be
- 8:38able to update all this. To use Cloud
- 8:41Code, you need at least a Pro account.
- 8:43So I'm going to look for the Versai
- 8:45Hataway shareholder letters. I'm going
- 8:46to go to the Versai Hataway page. And
- 8:48as you can see, here we have all the
- 8:51years. If I click, it takes me directly
- 8:53to the PDF and I can download it with
- 8:55the current year. And here, as you can
- 8:57see, each letter with Warren Buffett's
- 8:59lessons for that year. This is
- 9:01obviously going to take us some time,
- 9:02guys, it's not something we can do
- 9:04quickly and that's it. We have to have
- 9:05it well structured, well organized, but
- 9:07once we have it, I think we can give it
- 9:09many uses. Okay, so I've already
- 9:11downloaded the letters from 2024 to
- 9:132008. For now, we're going to stop
- 9:14there and then I'll continue feeding it
- 9:16, because you can continue feeding it
- 9:18with many more raw files later. That's
- 9:20the great thing, that you scale this
- 9:22structure up more and more. Then, I
- 9:24could add the missing letters from
- 9:27previous years. Okay, for now we're
- 9:29going to leave these files here and
- 9:30what we're going to do is go to Cloud
- 9:32Code and then to use Cloud Code you can
- 9:34do it in different ways. You can do it
- 9:36through your computer's terminal if you
- 9:37want, you can do it through the
- 9:39application, which I think I'm going to
- 9:41do with that in this video because it
- 9:42would be the easiest. Or you can
- 9:44install, for example, a tool like
- 9:46Visual Studio Code, which is the one
- 9:48many people use, and there integrate
- 9:50Cloud Code and use it through this tool
- 9:52. In my case, I'm going to use the
- 9:54cloud application on my computer, where
- 9:56I'm going to go directly, here we have
- 9:57chats, cowork and code. I'm going to go
- 9:59to the code section and here we would
- 10:01be in cloud code. What I also have to
- 10:03do is create a new folder on my
- 10:05computer. That's going to be the folder
- 10:06where we're going to create the entire
- 10:07structure. For example, I'm going to
- 10:09call it brain buffet. And now what I'm
- 10:11going to do is select within Cloud Code
- 10:13, give access to this folder that we
- 10:15just created. Basically here you can
- 10:17select by opening a folder and select
- 10:20the folder that we created, brain
- 10:22buffet. Trust the workspace, we click
- 10:23on it and we would already have the
- 10:25folder assigned. Before creating the
- 10:27Cloud MD file under Carpati standards,
- 10:29what we are going to do is create a
- 10:31subfolder and we are going to call it
- 10:33row. And this is where our raw files
- 10:35are going to go. So, here in the row we
- 10:37are going to dump all of Warren
- 10:39Buffett's letters. Basically, I am
- 10:41going to take them from the other
- 10:43folder and dump them here. And now we
- 10:45would have another subfolder inside the
- 10:47brain buffet folder, which is row, with
- 10:49all the letters in PDF. And here we
- 10:51could continue adding files, since this
- 10:53folder belongs to us and these files
- 10:54are not going to be touched. And now
- 10:56inside Cloud Code and with the folder
- 10:58already selected, what I am going to do
- 10:59is give it a little context. I am going
- 11:01to tell you that here below you have
- 11:03Carpaty's LM Wiki pattern. I am going
- 11:05to attach it. Now I am going to tell
- 11:07you where to download it. Build me a
- 11:09second Warren Buffett brain. This is
- 11:11kind of my goal. And you give it a
- 11:13little context. Here you can explain
- 11:15whatever you want. The more precise you
- 11:17are, the better, because that way the
- 11:19cloud MD, which is the file that Cloud
- 11:21is going to use to follow a series of
- 11:22rules within this entire architecture,
- 11:24will obviously be more detailed and
- 11:26will work better. And here I told him,
- 11:28since I am an analyst, my focus is
- 11:30long-term investment, second order,
- 11:32create the cloud.md, the structure and
- 11:34then read all the row files and build a
- 11:36structured and interconnected financial
- 11:39brain on Puffet. My idea is to have at
- 11:41his disposal all the knowledge he used
- 11:43in his letters so that he can use it in
- 11:45my investments. Stop and ask me
- 11:46whatever you need before starting. This
- 11:48is important, okay? So that he can ask
- 11:50us questions about things that he wants
- 11:52to know how we want to do. This is key
- 11:54to putting it. And now down here I am
- 11:56going to copy Carpaty's pattern. For
- 11:59this I am going to leave the website in
- 12:00the description. As you can see, he
- 12:01made it public 3 months ago. And here
- 12:03he basically explains the idea of how
- 12:05this architecture works. This is simply
- 12:08the idea of how it works, it is not
- 12:10code or says how to do anything, simply
- 12:12how to structure the idea. Therefore,
- 12:14if we give the idea to Cloud Code, it
- 12:17will know how to approach this
- 12:19architecture to follow Carpati's steps
- 12:21exactly. Therefore, we are going to
- 12:23click on row here to be able to copy it
- 12:25. And what I am going to do is copy all
- 12:27this. Control C, I go back to cloud and
- 12:30I am going to paste it down here. And
- 12:33so we have all this. I press enter.
- 12:36Here we can select the language model.
- 12:37For now I am going to go with opus,
- 12:39although Fabel is now available again,
- 12:40but for now I am going to go with opus
- 12:42because I do not think it needs a lot
- 12:43of capacity to do this. It is going to
- 12:45ask us for access to the folder, in
- 12:47this case we have created a new one,
- 12:48the Buffett brain folder, which is the
- 12:50one we have given it access to and it
- 12:51is going to start working. Now what it
- 12:53is doing is creating the Cloud MD file,
- 12:54which is going to be the series of
- 12:56guides that it is going to use. Each
- 12:57time Cloud accesses a file within this
- 12:59folder, note, we have only given it
- 13:01access to this folder, it is going to
- 13:02ask us for different permissions. There
- 13:04, if you want, you can enable the
- 13:06bypass so that it doesn't ask for your
- 13:07permission and does it directly, or you
- 13:09can simply confirm permission each time
- 13:11. See? Extract information from the PDF
- 13:13, well, you allow it and that's it. In
- 13:15this case, what it's doing is reading
- 13:16the PDFs. Look, it's already asking us
- 13:18the first questions. In what language
- 13:20do you want to build the wiki? The
- 13:21summary pages, the principles, and the
- 13:24analysis. Do you want it in Spanish
- 13:25with the citations in English? Do you
- 13:27want it in full English? Do you want it
- 13:29in full Belgian? Each page with its
- 13:31version in both languages. In my
- 13:32specific case, I'm going to go with
- 13:33English, but the good thing here is
- 13:35that you can configure it however you
- 13:36want. If you want everything in Spanish
- 13:38, well, in Spanish, I'm going to choose
- 13:40English by pressing two or by clicking,
- 13:42but by selecting two and pressing next.
- 13:44Okay. It tells me, "What should the
- 13:46brain be optimized around?" Choose
- 13:47those that provide most value.
- 13:49Principles and mental models,
- 13:51application layer, actionable
- 13:52checklists, framework, Buffett-style
- 13:55decisions, business cases, transaction
- 13:57and stake records that Buffett analyzes
- 14:00as live case studies, accounting and
- 14:02financial concepts, well, we could tell
- 14:04you a little bit about all of them
- 14:07because they all interest me. Third
- 14:09question, how much freedom do I have
- 14:11regarding sources? Only the letters
- 14:13with citation. Everything stated in the
- 14:15wiki must come from the 17 letters and
- 14:17be cited by year plus topic, maximum
- 14:18rigor and traceability. This way we
- 14:20have traceability of what is being
- 14:22cited from each document. And we have
- 14:23the document in raw file. This would be
- 14:26the cleanest option and it is the one I
- 14:27recommend the most. Then we could use
- 14:29it, allowed to supplement with
- 14:30established knowledge from outside the
- 14:32letters if you want it to use external
- 14:33information. But when we want to work
- 14:36with good information that only we
- 14:38provide, and this is especially for
- 14:40analyzing companies, my recommendation
- 14:42would be to opt for one. That way it
- 14:44will always use the information in the
- 14:46raw file. I'm going to click on one in
- 14:48this case. Next. I only want what the
- 14:50letters say. How do I run the initial
- 14:52build? Everything now. I process the 17
- 14:54cards in one go with subagents. I build
- 14:56the wiki and show you the result. First
- 14:58, I do a pilot. I process two or three
- 14:59cards. I show you and, after your
- 15:01approval, I continue with the rest. My
- 15:03opinion is to first do a pilot to see
- 15:04how it is synthesizing each of the
- 15:06cards. We click send and it will start
- 15:08giving us a pilot in case we want to
- 15:10correct things. Well, there you have it
- 15:12, the pilot is now built. Here is the
- 15:15summary for your review. In this case,
- 15:17it took about 20 minutes because we
- 15:19gave it a lot of files. My
- 15:21recommendation is to first use a
- 15:23slightly lower language model, perhaps,
- 15:25for example, with Sonet it would be
- 15:27enough to consume fewer tokens. And the
- 15:30second thing is that we have to do this
- 15:33progressively. That is to say, if at
- 15:35first we try to give it all the brain
- 15:37files and have it ingest them all at
- 15:39once, in my opinion it will be too much
- 15:41work. We could have told it at the
- 15:43initial prompt to first go step by step
- 15:45, to only create the cloud MD and to
- 15:47only ingest one of the letter files.
- 15:50And each day what we do is that we ask
- 15:52it to ingest more row files that are
- 15:54missing and in this way it would
- 15:56gradually work and we can do it
- 15:57progressively without having to consume
- 15:59our entire session in a single session.
- 16:02Having said that, here it tells us that
- 16:04it has already created the pilot brain,
- 16:06the pages that it has created, all the
- 16:08concepts, all the business cases. In
- 16:10this case it has used a few, well, I
- 16:13have used 33 interconnected pages, okay
- 16:15? There would be many letters that it
- 16:17has not created, but what we are going
- 16:19to do is that then little by little
- 16:21each day we could tell it to ingest
- 16:22some letters. And this will again be
- 16:24built layer by layer by layer. We
- 16:27already have the structure, therefore
- 16:29everything that we add to it and do
- 16:31more will add up. So, let's see how the
- 16:33folder has turned out. Look, here I
- 16:35have Buffett's brain, he created Cloud
- 16:37MD, he didn't touch the row folder
- 16:39because those are our files and he
- 16:41created the wiki folder. And look at
- 16:44everything he's done. He's created the
- 16:46concepts, he's created the cases, he's
- 16:49organized everything by the
- 16:50applications, the people involved in
- 16:52the letters, the principles, the
- 16:54sources he's used, the summary, a lot
- 16:56of things. But of course, reading this
- 16:58this way isn't useful. This is the
- 17:00architecture. With this, every time we
- 17:03work in Cloud Code and have our
- 17:05Buffett's brain folder selected, we can
- 17:08ask him questions or we can add more
- 17:11things or again have all the context
- 17:13because now Cloud can find it very
- 17:16quickly. But now we can use an MD file
- 17:19reader to see a little bit about how
- 17:21this whole brain turned out and we can
- 17:23also have access to those notes. For
- 17:26this we can use the Obsidian
- 17:27application. You have it for Mac and
- 17:30also for other platforms. I think it's
- 17:32available for Windows too. Look, you
- 17:35download Obsidian, it's free, and when
- 17:37you download it, it's going to tell you
- 17:39this. When you open Obsidian, it's
- 17:41going to tell you either to create a
- 17:43new vault or to open a folder as a
- 17:44vault. In this case, since we already
- 17:46have the folder, we already have that
- 17:48brain created, we want to open it. So,
- 17:49we're going to open the Buffet brain
- 17:51folder. When I open the folder, I
- 17:53already have all the subfolders here.
- 17:55First, the row files, which we said,
- 17:57which are all the letters in PDF.
- 18:00Perfect, we're not going to touch those
- 18:01. And then all the wikis that it has
- 18:03created. Look, it has created a
- 18:05checklist to look at the quality of the
- 18:08business that we can use. Hey, it has
- 18:10pricing power, it has incremental
- 18:12returns on capital, the balance sheet,
- 18:14the management, it has created
- 18:16different red flags, important points
- 18:18that we have to review when we analyze
- 18:20red flags, right? When we analyze a
- 18:22company and it shows us everything
- 18:24linked together. Note that each concept
- 18:27and each year is linked so that we can
- 18:29access other notes on these topics.
- 18:32Then he has made us a folder of cases
- 18:34with, for example, some companies, for
- 18:36example, Geiko, the insurance company,
- 18:38where he talks to us about different
- 18:39things, he links us to all the things
- 18:41that we want to have. Siscandis also,
- 18:44the candy company. Then he has also
- 18:46made us the concepts, for example, the
- 18:49concept of economic goodwill, what it
- 18:51means, how the concept of float works,
- 18:53the concept of intrinsic value, who it
- 18:55relates to. And we can click here with
- 18:58Obsidian and it makes us a graph of the
- 19:00concepts that he has created and we can
- 19:02see how everything is linked to each
- 19:04other. For example, Mr. Market
- 19:07redirects us to other important
- 19:08concepts. For example, in 2008 he
- 19:10talked to us about all this, in 2011 he
- 19:12talked to us about all these concepts
- 19:15and we could click and go to the
- 19:16concept that we created. For example,
- 19:19if we create a concept such as the
- 19:20margin of safety, we see everything
- 19:22that is related to the margin of safety
- 19:24. Then it also creates those important
- 19:27pages, that structure that Carpaty
- 19:29mentioned, which will basically always
- 19:31be updated. For example, the index
- 19:33basically tells us how everything is
- 19:35structured. The log tells us the latest
- 19:37changes. Here it tells us that it has
- 19:39created the schema and then it tells us
- 19:41that it has ingested the letters from
- 19:432008, 2011 and 2014 only. So you can
- 19:45see little information here, but look,
- 19:47it is already quite a few concepts and
- 19:48we have only extracted three letters.
- 19:50What I could do now day by day is ask
- 19:52it to ingest more letters so that it
- 19:55gradually creates this entire database,
- 19:57this entire brain that we can then use
- 19:59for the context. And here is the Cloud
- 20:01MD file, which we could review and make
- 20:03changes to. The Cloud MD is basically
- 20:05the entire architecture and all the
- 20:07steps of how Cloud is going to read
- 20:08these files, how it is going to use
- 20:10them, what rules it is going to apply,
- 20:12why it does it, what the context is and
- 20:14we would have everything. So whenever
- 20:16we want, we can upload more files to
- 20:18the row folder, raw files, and have it
- 20:20ingest them and build those wikis,
- 20:22those pages, and all those concepts,
- 20:25and that way, little by little, we're
- 20:27building that brain. For example, let's
- 20:30open a new window in Cloud Code. Let's
- 20:33tap Buffet's brain again and ask it to
- 20:36ingest, let's tell it the charts for
- 20:412020, 2021, and 2022. If I had added a
- 20:44new file, I could have told it directly
- 20:47, "Ingest this file." And that way,
- 20:50little by little, we're building that
- 20:52brain. The utility of this is that
- 20:55every time we use Cloud Code with that
- 20:57folder selected, Cloud Code will
- 20:59already have access to all that context
- 21:01, to that entire database, to all those
- 21:03pages created with all the concepts
- 21:06that are related to each other. Imagine
- 21:08that we are studying to make a video or
- 21:11a script about Warren Buffett, the next
- 21:13time I work on a script about Warren
- 21:15Buffett for my YouTube channel, I could
- 21:18work on this folder so that I have all
- 21:20the references. And then, in the
- 21:22YouTube script, it will cite what he
- 21:24said in the 2012 letter, what he said
- 21:26in the 2010 letter. This is related to
- 21:28the concept that Geiko explained in the
- 21:312020 letter. Do you understand? We are
- 21:33creating a kind of neural network so
- 21:35that every time we want to work on it,
- 21:37we always have the context, we don't
- 21:40have to teach it again. Obviously, at
- 21:42first it is hard to build, it takes
- 21:43time, it requires spending tokens, it
- 21:45requires making the whole structure,
- 21:47dedicating a few minutes to it. As you
- 21:49can see, it took me 20 minutes to make
- 21:50the whole structure. But then little by
- 21:52little, feeding that structure, well,
- 21:54that's what takes the least time and
- 21:56dedicating a little time each week to
- 21:58feeding this kind of brain. I can think
- 22:01of many applications. For example, for
- 22:03those of you who are analysts and cover
- 22:05a company in great detail and
- 22:07constantly monitor this company,
- 22:09analyze all its results, see all the
- 22:11news it publishes, have all the
- 22:13necessary information, you could create
- 22:15an independent brain with all the
- 22:17information from this company, all the
- 22:19raw data, the latest 10K, the annual
- 22:21reports, all the financial data, and
- 22:23have Cloud only use those raw documents
- 22:26. That way it won't have to go out and
- 22:27look for anything to fill in, because
- 22:29sometimes AI has a bit of amnesia and,
- 22:31when it can't find information in a
- 22:33document, it basically gives you
- 22:34information from a context it has from
- 22:36a web page, or from a blog, or from the
- 22:38consensus that currently exists on Wall
- 22:39Street. So, this way of working seems
- 22:41much more correct to me because you're
- 22:43basically going to pull from the files
- 22:45of this database that you're building
- 22:47and then, in addition, you'll have it
- 22:49for yourself so that you can access all
- 22:51that information and have it well
- 22:53organized and well connected to each
- 22:55other. In fact, I have to show you
- 22:57something before I go, and it's the
- 22:58following. The good thing is that we
- 23:00can also feed these Obsidian files with
- 23:02its extension. We can download the
- 23:04Obsidian Chrome extension and, when we
- 23:07are in an article that we think can
- 23:09feed our brain, our architecture, we
- 23:12could go to the extension, click on
- 23:14Obsidian Webclipper and, basically, we
- 23:17could make a clipping of this article
- 23:19and directly add it to Obsidian. We
- 23:22click on open Obsidian and look, it can
- 23:24make a text file of the article note
- 23:27with the source, the author, the date
- 23:29it was published and with the summary
- 23:31of this article so that we also have it
- 23:34available in that architecture, in that
- 23:36database. It is also important that
- 23:38with Cloud, each time we work, we are
- 23:40going to give it context and feedback.
- 23:41And that will mean that the
- 23:43conversations will be adjusted.
- 23:44Everything will be recorded in the log:
- 23:46all the changes we requested, all the
- 23:48feedback we gave, "Hey, this wasn't put
- 23:50so well," or "this concept didn't get
- 23:52the terms right," and you can go about
- 23:54correcting that entire file, that
- 23:56entire architecture, so that we have it
- 23:58much cleaner. Again, this isn't magic,
- 24:01it obviously requires time to maintain,
- 24:04create, and also give it a useful use.
- 24:06So if you liked it, don't forget to
- 24:08support the video. And now, I'll see
- 24:10you in the next one. And if you want to
- 24:12start investing, I'll leave you a
- 24:13complete free course below where I show
- 24:15you the step-by-step instructions to
- 24:17get started. See you in the next video.
About this transcript
This page contains the full transcript of He Creado un Segundo Cerebro con IA… y Es una Locura! by Javier DV - Inversión y Finanzas, generated from the public captions YouTube serves with the video. The transcript has 4,995 words across 708 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.
What you can do with it
Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.
Free YouTube transcript tool
YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.