Transcription
So, friends, hello everyone. My name is Nick, and you are on the NCK channel. And we haven't seen each other for a long time. And today I want to tell you about an AI tool that has radically changed my work with AI. And from the title, you have probably already understood that it is Perplexity AI. And if you are not using it yet, then after this video you will definitely start using it. [music] So, let's start, perhaps, with the basics. All current models use quite outdated data, because they are trained on them for some time, then these and then the models, come out into the world and, accordingly, the training of the current version of the models ends. For example, Gemini has knowledge up to January 2025. That's already 8 months of new data that the model doesn't know. And ChatGPT is generally September 2024. And that's a year, a year, essentially the last year. The neural network knows nothing. And Claude too, for example, it's March 2025, so 7 months. And in this sense, I always wondered: "What if I need some current information? What if I need to check several sources simultaneously?" Now, many models indeed have such a possibility, as, in fact, to do a search. Here. But in reality, it doesn't always work relevantly, and some models are guilty of using built-in sources. Perplexity AI radically solves these problems in this regard. And the most surprising thing is that you can get access to this tool for absolute pennies. Let's figure out how it works and why it's cool. And the first such small life hack is that Perplexity Pro can be bought on various resale platforms and done for literally, uh, 500-600 rubles, which is, in fact, amazing. I will not name any specific resale platforms. You just type into Google "buy Perplexity" and, uh, choose the best platform for yourself. For this money, you get access to the best AI models on the planet. And this is simply incredible. Let's go into Perplexity, click the model selection button, and, uh, we see almost all the best closed models at the moment. And just think about it, all this for literally 400-500 rubles a year. This is both Claude 3, and Gemini 2.5 Pro, and GPT-4, and Grok 4. And these are the best models on the market. And they cost a lot of money. Let's now, uh, quickly go through the current models and tell you which model I use and for what. My main model is Claude 3. Accordingly, I use it in 90% of cases. For me, it is the best for several reasons. Firstly, because it is excellent at logic. And secondly, in my extremely subjective opinion, this model tries the least to please the user, which for me is an important factor. Gemini 2.5 Pro is a super model, also a thinking one, but I mainly use it to generate some pictures. I use GPT-4 mainly in thinking mode. It copes excellently with various logical tasks, with some mathematics. And I also often use it for generating some images. Claude 3, unfortunately, for some reason, this is already an outdated model from October 1st. I don't know, outdated, apparently, it will be removed from here. A pity. I often use it for writing some creative posts, creative texts, and all that. It is the most human-like, in my opinion. And Grok 4. To be honest, I didn't understand it. I tried to poke around, test it, but I didn't understand it. To be honest, I have some task in my backlog to try Grok 4 more, because it performs excellently on benchmarks, but, we all know that benchmarks do not reflect the truth. And, the coolest thing in this regard is that the models are updated almost in real-time. That is, for example, before, there was Grok 3 here, and as soon as Grok released an update, released the fourth Grok, it appeared here, literally after half a day. And this is an excellent opportunity to test various models. And also, one interesting fact, I have never reached the usage limits, although I use Perplexity quite a lot and super often. Well, and accordingly, we all know that the main superpower of Perplexity is that with every answer, practically with every answer, the neural network goes to the internet and searches only for current information. In my experience, in 95% of cases, I get exactly what I need. The models are not limited by the knowledge of the year they were trained on. And the answer is always supported by fresh data from the internet. And for such money, it's simply incredible. Also, Perplexity has completely replaced Google for me in this regard. If before I used to Google something, try to look at sources, look inside, study the information, and this whole process took a significant amount of time, then now Perplexity completely removes all this, you know, headache. There is a cool feature here, thanks to which you can choose sources. By default, web search is always enabled, but you can also disable it, for example, if you are using some model, for example, for coding, or you don't need internet access, but by default it is always enabled. Also, there are academic, various scientific articles, social, that's Reddit, Hubber X, and financial sources. In most cases, I use a combination of web plus social sources in order to get the most relevant solutions for my question. Because usually on forums, the number of experts who discuss new things, discuss problems, is concentrated there, and you can get a snapshot of current information there. So I highly recommend enabling social sources. If we talk about additional features, then Perplexity has, for example, Deep Search. In my opinion, not the strongest deep search that I have seen. And in general, it cannot be called strong, to be honest. As for me, the strongest deep search is from GPT, of course. I also really like Deep Search, for example, from Google. Here, Deep Search ends at 3-4 minutes, it reviews about 30-35 sources, and on average, if you don't have a very complex question, then it's quite usable. Also, in research, their personal model Sonar is used. You cannot use any third-party models. The next feature is, of course, Labs or laboratory. For me, this is deep research. I use laboratories whenever I need to conduct some large research, study the market, study some issue. Also, by the way, from experience, I can say that laboratories write some texts excellently, deal with marketing, and generally study the target audience. Also, the feature of laboratories is that when a research is created, you get a complete ready-made project. You can get your answer in the format of a web application, so that it is convenient for you to read it, or some tables. Also, the model can draw diagrams and visualize certain options of information visualization during its reasoning. Let's take a look. I just before recording the video made one request. Let's see how the laboratories from Perplexity handled this request. The request was as follows. I asked Perplexity to analyze the AI startup market in Russia and show current trends. And I also said that the answer should be given to me in the form of a web application. It made such a web application. Let's launch the web application in full screen and see a quite interesting and beautiful picture. What is most interesting in my opinion, why I love Perplexity so much and especially the LAPS mode, is the beauty and simplicity of perception of information. I think this is very important. We see such a graph. Everything is clickable, everything can be viewed. By the way, we can see that according to the dynamics, the peak of the AI market is predicted for 2028 and will start to decline by 2029. Interesting. The market volume is currently 305 billion, by 2029 it will be 516 billion. Venture investment share is 30%. There is venture in Russia. Interesting. Okay. And, accordingly, it has even highlighted key players. Sber, Yandex, VK, MTS. Clear. Top companies by revenue and startups. Rubl, Just AI. Super. Excellent. Investment structure, large deals, key venture funds, super great. And technological trends. Computer vision, AI agents, finance, marketing. Excellent. Excellent. And, well, friends, I think comments are unnecessary here. You see it all for yourself. You see for yourself how information has become more convenient and understandable to receive. Also, Perplexity has, space, that is, by space. This is an analogue of projects in GPT and, for example, Claude. What is it? You create a space, where, roughly speaking, you add additional information for the neural network so that it is always in the context of your particular activity. Here, in particular, I have a space that I created to work faster and more conveniently and ask questions specifically regarding Python. Let's go to the settings. I have a somewhat awkward prompt, but in this regard, it works excellently, because the main task is simply to tell it that I need answers specifically regarding Python. And, I also embedded in it the specifics of writing Python code in Python, because there are some quirks with syntax, these JSON objects, and all that. And therefore, so as not to constantly tell it to go to the internet and see how to correctly write Python in Python, I immediately embedded it, and it works excellently. Here you also have the option, for example, to add some files so that they are always attached and at any convenient or inconvenient opportunity, the neural network can access the files. And you can also attach links, for example, a link to some important documentation, if you need it. And here is a small tip, which is that if you see that you are repeatedly writing the same lines in the prompt, where you explain that I work in Python, it has such a version, NA is located on a self-hosted server, everything is there, and you realize that you type this every time and spend time on it, then you can embed all this into a space, thereby, собственно, speed up your work with artificial intelligence. Next is Discover. I admit honestly, I don't really like news. I am 26 and not from the TV generation, but the Discover tab has won me over, to be honest. In short, Perplexity collects personalized news for you here. And these are not just news, not just some kind of article summary, you know, like in Google or Yandex. These are excerpts from dozens of articles that tell about a particular piece of news from different angles. For example, the CEO of Walmart warns that artificial ellipsis. Well, they intrigued, they intrigued. Published 51 minutes ago. Accordingly, if a piece of news interests you, you go in, look, see, read the information and see that the article is compiled based on thirty-seven sources. Amazing. And also, as an additional feature, you can also ask an additional question about a piece of news. For example, if something did not suit you or something remained unclear, you can, for example, ask a question. Here, let's. And we get a personalized answer, which also draws information not only from open sources, but also from this article. That is, it can go to the specific sources indicated in the article during the answer. Well, at least that's how it works for me in most cases. So, the magic of editing. Hop. Even the magic of editing was not needed. It studied 20 sources and gave this answer. The CEO of Walmart. Doug McMillon, what an interesting surname he has. Recently made one of the most categorical statements, blah blah blah. And here we get a list of vulnerable professions. Translators, historians, writers, salespeople, web developers, I don't understand, and brokerage clerks, radio hosts. Super. Here. And this is actually the last thing I wanted to tell you. This is Discover. A super interesting and, in my opinion, useful feature. And let's summarize. For 500-600 rubles, you get access to the five best AI models at the moment, current information from the internet, personal spaces for work, in-depth research, personalized news, and generation of code, diagrams, websites, well, and in principle, anything. And against the backdrop of all this, I don't see alternatives, to be honest. It seems to me that Perplexity has a strategy for market capture, considering the prices, considering their dumping. And I think Perplexity should succeed. In 2025, information noise will only grow. The theory of the dead internet, AI, AI content, fake news, and all that. And, in my opinion, tools are needed that will allow filtering and structuring information. And Perplexity is the best thing I know in this regard today. So, if the video was useful for you, then be sure to like it, subscribe to this channel. Also, subscribe to my Telegram, the QR code will be here now. There I share my personal experience, share cases, share experience, talk about what I am developing, how and thanks to what tools. Improve your brains, automate. See you soon. Bye. [music]