Transcription
I have saved a full 1,460 hours on learning, using a combination of Obsidian, Gemini, and NotebookLM. If before for the same information I spent a whole month studying a new technology, I needed to watch full videos, read tons of articles, posts, then now, using my new approach with the same combination of neural networks, I get the same result, but not in months, but in 10-15 minutes. At the same time, I get an identical result as if I had done it all manually. This is precisely the point of optimization. Few people think about it, but traditional learning without using this chain has many very serious problems. If you don't solve them, then in the era of neural networks, it will be very difficult for you to compete with those who use these same neural networks. But what exactly are these problems? The first problem is when you study a new technology. To do this, you watch a lot of YouTube videos, then read articles on websites, read separate books and posts in Telegram. This is exactly how I used to get different tricks, then in the process of work I gained new experience, and in the end the technology was learned, but this worked until the appearance of this connection. And in this scheme, I noticed one big flaw. I'm not talking about drugs, but about the problem of the approach in general. I think you've noticed this too. You open 20 YouTube videos on the same topic to get different knowledge. In the end, all authors repeat the same thing. From each video, you extract literally 20% of the information. Constantly the same information. In the end, it is enough to remove duplicates, remove the water and extract the core essence from the videos. You get the same thing, but in a shorter time. And this is exactly what I solved. Next, the second problem. I read about 60 books a year and noticed a peculiar feature. At least five or even 10 books will be maximally introductory, and it turns out that I already know all this information. As a result, I lost about 70 hours reading absolutely useless books. Again, I note that they are useless for me, because I already know this. But for beginners, this may not be water. The same book about digital minimalism will be useful for someone who has not read such books at all. But, perhaps, it will not be useful for me at all, because I have read dozens of such books. So the idea is as follows. There should be a neural network that can check if I know this information or not. And only if I don't know it, it gives me this information. If the neural network understands that there is no useful information in this book specifically for me, then it simply writes directly: "There is no need to read this book, because you already know it. It is in your Obsidian, in your knowledge base." As a result, this was also solved. I have already removed as many as 10 books that do not need to be read, because I already know this. And as a result, I saved about 150 or 200 hours. I will also show this later. And the third problem. You are choosing a new smartphone, laptop, or mini PC. You have clear requirements, but there is one problem. There are simply millions of models. You then need to go and search for videos on YouTube, study each model yourself. As a result, I also solved this. Before, it could take a whole week, or even a month, but with my neural network connection. You simply find the most necessary models in 5 minutes, then studying the price and characteristics, it says that here I have a couple of models. The most profitable in terms of price and quality are these. If you suddenly need more RAM, then here are two more models. If you need a reserve for 5 years, then this model. These are just some of the problems. I will show even cooler features. As a result, by applying this, you get results in 15 minutes, while those who don't use it spend weeks and months. I will show my real combat tasks where I applied all this, and I will show the speed before and after. Moreover, all this extract of ready-made knowledge can be added to Obsidian. I will also show how to add Obsidian to this neural network connection. So, I'll tell you the very first example. Before this scheme, I solved this task for a full 3 months, and now, using the neural network connection, it can be repeated in 15 minutes. So, what was the task? Not long ago, I was choosing a country to live in. I had about eight options. Each country needs to be studied, a separate note made, and everything written down point by point. As a result, I go to YouTube and start watching videos one after another. If there is any fact about a country that is useful to me, I add it to the necessary Obsidian note. As a result, in this way, I get all the advantages and disadvantages of each country. Then you can compare everything, filter out bad options and leave only the most adequate ones. As a result, this created such a headache that you simply cannot imagine. I spent a full 3 months watching tons of videos about different countries and constantly stumbled upon the same information. I had to constantly rewind integrations, the author's stories about kebabs. As a result, considering only the working time and the length of these videos, approximately 360 hours were spent. And before that, I also studied other countries. And there you can easily add another 360, or even more. And in the end, when I did the same thing in NotebookLM in 15 minutes, my mind was blown. Let me show you how I did it. So, let's go into NotebookLM. Initially, there are ready-made notebooks. You can click "Show all" and get ready-made notebooks on various topics. Scrolling down, we get recent notebooks and can create a new one. This is, roughly speaking, just different chats. So, let's create a notebook and let's name it. Let's call it relocation. Then we go to the left column. We can add sources directly, i.e., links to YouTube videos, articles, Telegram posts, websites, and so on. You can make NotebookLM search for sources itself. If you want it to find websites itself, then just select "Depressarch" here. And then you need to enter a query. Let's say, we enter this query and click search. It will find all the necessary videos, websites, and so on. And then you can ask questions. But personally, I used a different approach. Initially, for convenience, download this extension. It allows you to add YouTube videos directly. Moreover, it allows you to add an entire channel. As a result, if the channel is entirely about Armenia, then we get all the information about the country by entering just one query. As a result, here I found an example channel. Scroll to the very top, click here. This appeared after installing the extension. Then you can either create a new notebook or select an existing one. So, click and select relocation. As a result, all the videos went into processing. And opening a separate video, you can also click on NotebookLM and add it separately. Or, click "Add source," upload file, select file on disk, paste the desired text fragment, or click on websites and paste the necessary links here. As a result, the videos are loaded and we insert this prompt. Again, initially I wrote it in my own words, then I pasted this prompt into Gemini, wrote, like, rephrase the prompt for NotebookLM. Then comes the instruction for analysis. It studies all the necessary topics, then highlights the pros and cons, indicating information about medicine, transport, internet, climate, and so on. Then the cost of rent, approximate prices, and the result is in Markdown. Each point is a separate thought. And in the end, it provides the source. This is what distinguishes it from GPT and Gemini, because it does not fantasize, but takes information strictly from sources. And as a result, we get all the information. Scrolling down, here is information about Bulgaria. Minimum salary, taxes, cost of living, real estate, and so on. Then separately Poland, then Czech Republic, and so on. We throw all this into Obsidian. Let's say, here we get such notes. I immediately understand what the approximate prices are, what the rent is, what the climate is, and so on. If I had studied all this manually, how much time would I have spent studying these videos? It would have been many times longer. Moreover, it also added a comparison. Then I can use this chat to ask questions based on this information. Let's say, like this. I give an additional request. As a result, it takes all these videos and the chat and gives me an answer. As a result, here we get a rating of all these necessary countries. Also, each of its points is accompanied by a source. We get to the very text. That is, it did not invent it, it took it from the video. And below is the very video. Moreover, based on all this information, you can create an audio, video summary, flashcards for Anki, then separate tests, infographics, or a presentation. For example. Here I made a Mindmap. This is the main branch. Then let's look at the cons. Open. And here are the cons. These are stray animals, poor road quality, economic problems, low salaries. And in general, this is all true, because I studied Bulgaria manually. Then we get the rent range, approximate food costs, and so on. I think this is a real blast. It even found approximate prices for studios. So, next, the same, but already in the form of an infographic. This is, roughly speaking, a cheat sheet, which has rental prices, housing prices, comparison with Turkey, and so on. Also separately, such a table. I don't understand at all how I used to do all this manually. Moreover, here are separate flashcards for learning, I added a new technology here and you study. Then you click download and get such a file for Anki. Then separately created tests. Select an option and get an explanation. Also separately, it creates an audio summary and below a video summary. Let's say, you are going home, it's inconvenient for you to read. Just turn on the audio. It will give you the most concise summary. Removing all sorts of filler words, profanity, and other junk, you get a perfectly ready file. If someone needs to read 20 PDF files with reports or with the same technical specification, then you just upload it here, get the audio, and it will tell you everything very briefly. Also, as you know, all this time I was giving the Python OOP course and a private channel as a gift. As a result, now I am working on a Linux course, and as a result, the private channel will need to be made a separate product. Right now, by purchasing the AOP course, you get it as a gift. You also get eternal updates and eternal access to both. And those who manage to sign up before I introduce a separate private channel, for them everything remains the same. They also get all updates, get a very detailed AOP course and just all the basics, all my tricks, research, real optimization cases. I show my methods of saving memory so that you can make your programs better. And in the private channel, I am already creating a separate video that I don't upload to YouTube. Let's say, on the same NotebookLM. I also have such tricks that I will publish only in the private channel, because I don't want to create competitors for myself and publish all this to a huge audience. There is information that needs to be shared only within a limited circle of people. This is exactly what I do in the private channel, I show various working cases, share my experience, show my protection tricks. I also make videos about Linux, publish tools, mini-courses, and so on. When I make the private course separately, the probability is that it will be the same. Separately for the course and the same for the private channel, because updates are constant, access is eternal. Therefore, while I haven't made the private channel separately, you can get the AOP course and the private channel for one price. To do this, leave an application and I look forward to seeing you in the private channel. As a result, if we sum up all the time spent, all similar tasks, then the figure will go far beyond what I indicated in the title. But that's not all. Let's recall the optimization of books. Let's create a new notebook specifically for this task. Let's take a very popular book, "Rich Dad Poor Dad." We can add it directly or initially research reviews. Turn on Depressarch, enter a query to search for reviews. Then click, and it starts searching. As a result, at the review stage, you can already understand if the book is good or not. But you also need to find the PDF file of the book and upload it here. I added the PDF separately, click "Add source." Then upload file. There are certain formats that it supports. It doesn't support Papago, so through Papago or online services, you can convert it to PDF, or to text. Also, this is only a preliminary analysis, but already from it I understand whether to read the book or not. Here, even your Obsidian base can be added as a separate file. Only the entire base needs to be converted into one PDF file so that it can then search through this PDF based on the book and reviews, and understand if this book is useful for you or not. Again, how to export the entire Obsidian to PDF? You can either search for ready-made plugins or write your own script. I am currently doing it this way, but the idea of comparing with Obsidian itself is very cool, however, I need to remove some information from it to then export it all to PDF. I will probably export it using a regular Python script. It's very simple. We iterate through all the files and add them as text to the PDF. Moreover, it doesn't even have to be a PDF, but just a text file. This will be even easier. Then add it here and get a comparison with Obsidian. I will also show another way to create one file for the base. Open this script. Specify the path to the base itself. Then the result file. Then separately ignore directories and files, because here I have pictures. The information here is not particularly needed. Then it iterates through all the folders and adds everything to the result. As a result, we run it and get a separate file. Here is the number of lines. We upload this file to NotebookLM. So, into those very files and ask a question. Based on my Obsidian base, namely result. Analyze all sources and also the book separately. It takes my financial notes and understands that the book is useless for me, because I read it back in 2018 and already know everything. As a result, it analyzes and says that most likely it should not be read. Then it describes why, and then come the reviews that it took from various sites. These are real reviews from some Tresh and so on, not that it invented it all like part of GPT. As a result, in this way, I have already filtered out 10 books. This is approximately 150 or 200 hours. Moreover, you can ask for analysis by chapters or ask which specific chapters will be useful to you, and specifically based on your Obsidian base. And as a result, Obsidian is a real reflection of your brain. By filling it, you get short, small notes that can be reread later. Moreover, this is already a whole weapon for neural networks. We, roughly speaking, have connected our brain to a neural network. If you are in doubt that a book might be useful, just enter a query. That is, ask to make a summary. Then I will make an audio summary, then a separate presentation, also an infographic and a mind map. A mind map is really a very easy way to learn information. As a result, we make a request, and then click on the desired points. And also the time that we saved can now be spent on some complex, voluminous book that will really bring profit and is not in our base. How much time can be saved this way in a year or in 5 years, calculate it yourself. There will be tens of thousands of hours. Also, if NotebookLM is more like a source of information that extracts information very precisely, then writing a full script, generating a website based on information, or generating some document, or discussing sources, it cannot do completely. Therefore, let's go to Gemini, click here and select NotebookLM. Don't pay attention to the language, because I am currently in Poland. Then we take the prompt and add it. We ask to write a script for a video reviewing a book. NotebookLM has a lot of sources, and it doesn't cope very well with this task. The formatting really suffers, but we add all this to Gemini, send it, and get the result. You can do the same with studying countries, different technologies, and so on. That is, you extract precise information from NotebookLM, and then based on this precise information, you use Gemini, which knows how to tell, generate something, and so on. As a result, here is the script. And judging by the book, it really highlighted the necessary theses. This is exactly what is written in the book. If you think that's all, then you are very mistaken, because there are so many cases and tricks for NotebookLM that you can study it for weeks. I also have tons of optimizations, but they simply won't fit into this video. Therefore, if you need those very new features, ways to speed up learning and work, which I have personally tested and which I personally use, then give it a like and write about it in the comments. I will show even cooler ways to save your time. I will also make a separate video for the private channel, as promised.