Как Создать и Продать бизнес с помощью нейросети

An interview with a guy who wrote an extension for chrome in 10 hours with the help of a neuroset without technical knowledge. A week later he already had 1,000 paid orders, and a month later he sold this startup in this video. Let's go. Tell me, what do you do? Well, in general, what is your profession, what did you do before that? I am an entrepreneur. I work exclusively on web projects. Hmm, before I started working on this full-time, I worked for years at a startup accelerator, where I saw all possible project cases and how they start, develop, close, raise investments, spend these investments, and so on and so forth. I've seen it all in six years. And in 2017, I decided to go work on my own projects, on my own ideas. And since then I've been doing this full-time. I am not raising any investments. I work exclusively either alone or in teams of two people: myself and a technical partner. We have several projects that we have done together with my technical partners. And my philosophy is to build something every day. to build something new. I'm not interested in working on the same thing for years at all. I get bored very quickly. And that's why I can have three to five projects at completely different stages at the same time. Well, if I do them with partners, then accordingly we create a roadmap and move along it in parallel. They do their part, I do my part. And when did you learn about advanced intelligence and from whom? Well, my main source of information is Twitter. I have two accounts. One Ukrainian-language, very old, which I started a long time ago, and one English-language, which I started specifically to simply communicate with the global community, entrepreneurs, indie makers, just creators, it doesn't matter what they call themselves, but with the English-speaking community. And for me, Twitter is the source of all, uh, all the news. So I knew about AI for a long time. Back when I was working at the accelerator, I saw some projects that talked about this, that were trying to do something there with machine learning and so on. a-a specifically about the possibility of using so-called artificial intelligence, because it is actually, yes, not artificial intelligence at all, that's why I'm talking about so-called. Through the API, I found out when Open AI released their API, that is, they gave developers the opportunity to interact with their API or test it through their playground, which they have on their website. Well, that's why it happened when they released it publicly. I immediately applied, got access through whitelist, got access probably a month or a few weeks later. And since then I have tried to work with it, using it in various of my projects. That is, if I'm not mistaken, it was probably in 2021. Dadada. Well, that is, not on hype. I saw on Twitter that you called the GPT chat your CTO without programming skills, and you started creating a project. Tell us a little about it. How the idea came, what he faced, what difficulties, how he solved them. On the same Twitter, I read that people are starting to use GPT chat in some of their daily tasks, including learning to program or using it in their work as such an assistant. Well, I was actually interested in this topic, because programming is not something completely incomprehensible to me. I have a high-level understanding of how the code works. Well, just sitting there programming, of course, is out of my reach at such a professional level. That's actually why I decided to try it out, and see if the GPT chat could really not just answer questions there, but actually help me create something from scratch. So I decided to try it on a real case. The best way to learn is when you take a project idea and try to implement it gradually through iterations, improving the internal functionality. So I just opened the GPT chat and wrote: "Help me create a Chrome Extension. " And from that time on, if we started

interacting with him. That's why I called him my CTO, because he basically wrote all the code. Oh, what did I do? I analyzed the functionality, tried, tested, set new tasks, and copy-pasted the code. Back and forth, back and forth. For 10 hours, until I had a working version. In reality, in 10 hours, in a 10-hour workday, I actually started doing it in the morning. In the evening I had a working version of the mumbled. A very simple project, of course, very simple, but it worked and at the end of the day I was able to record a demo video of how this project actually works. And how did you find the first clients? I quickly made a landing page. Well, I made this demo video and just wrote about it on my English-language Twitter account. Yes, we actually managed to get some kind of first audience, there, the first validation and so on, because I, uh, took pre-orders from people. That is, they paid before they even saw the Chrome Extension, how they installed it, and so on. That is, I only had a demo video of how it works on my computer. In order to publish it, I had to tinker with it for a few more days, because I had to fill out an application with Google to get it into the Chrome extensions marketplace, and so on. That is, I had to do even more work to make it publicly available, to make it installable from the marketplace, than I spent time writing it at all. And if there are any difficulties with Google? That is, you say, the complexity was much greater than Yes. That is, they have some requirements. That is, you actually wrote an extension with the GPT chat, which Chrome later adopted. That is, this is very cool. Yes. They have requirements. They consist in the fact that you must follow a certain algorithm in creating this Chrome extension. And since Chrome Extension is a program that runs inside Chrome, it has access, for example, to tabs. Therefore, you need to explain to them very well why you need access to tabs, access to URLs, access to inputs. I created a Chrome extension that had one simple functionality. In any, uh, text window on any, any website that you open, you could write sAI, then some query, and press Enter. And, actually, this request that you wrote after AI, I sent it via Chrome Extension to the Open AI API, just as a request to AI and received a response and inserted this response here, in this box where you call it. The idea was very simple. And I wanted, for example, to check one of my tweets for grammar. That is, I wrote CAII before my tweet, pressed Enter, wrote to it something like check grammar, pressed Enter and received a response with the grammar already checked there for this comment that I want to leave, or tweet, because the idea was for this extension to work in any inbox where we write text online. And so I had to explain in great detail why I needed access to websites, to these little windows where you can write text and so on. But the GPT chat coped with this task perfectly, because it served as a formalizer of all these ideas. That is, the point was that I simply copied all the fields from the application into the Chrome extension, copied all the fields with descriptions, and described the task to him, that here I have a Chrome extension that you and I wrote together. He does things like that, that's how he works. So he needs such access. Fill out this application, answer all these questions, explaining in detail why we need access to the website, access to input boxes, and so on. he made an application. And this application itself, what was most interesting, passed moderation the first time. That is, Google didn't send it back to me for revision at all. That is, they accepted it completely from the first time. The whole difficulty was that Google simply took several days to review this application. This is where the human factor came into play in this case. Yes, that is, JPT RМ and I created something in 10 hours, he wrote the entire application in, uh, a minute, and then we waited three or four days for Google to verify it, approve it, for someone to look at it and, uh, allow it to be published.

This is absolutely incredible. So tell me, please, from this entire history, everyone can do it, or is there something so complicated here? Why don't others do it there? Why are others afraid to create this? I don't know why others are afraid. Can everyone do this? I don't think everyone can do it. In order to work with GPT at this level, you need to understand what a product is and what components it consists of. how can it work. That is, in order for the GPT chat to write some code, it needs to explain what you want, in sufficient detail. In order to explain this to him, you first need to map out how the product will work. Therefore, just any person cannot sit down and write GPT chat instructions simply because they do not have enough experience interacting with products at the level of their creation. That's all. But if this person spends a little time, understands how any products they interact with every day work, breaks them down into components, and understands that there are parts that are responsible, for example, for authorization. That is, if we log in or sign up on a website, it means that this is the authorization stage. And if you break down all the interaction with the products you love into these components, then you can write specific instructions for the JPT chat based on these components so that it writes this code. Also, this code won't always actually work. This is the next stage of problems, because expecting chat GPT to write perfect code that will work the first time, and also remembering that we have several components. That is, just because we ask GPT for advice on how to write this part, and this part in the chat, doesn't mean that they will agree, that they will work together. That's why there's so much complexity here. Well, I wouldn't say we're at the stage where you can describe the requirements in great detail and immediately get a cool working product. No, we are not at this stage. But specifically, if we are talking about GPT chat, then it can be a very good assistant, as it was in my first case and as it is, for example, now. I am currently working on a new product. And this time I decided to learn to program. And GPT chat helps me with this. I'm writing a very simple app in PHP for a number of reasons, including because PHP is a very old, ancient language. A lot has already been written on it. And we will publish the chat GPT, as a model trained on content, online, including on GitHub and on ST stack Overflow. chat GPT can write PHP code well, so it can test it, of course. And because Open AI increased the context window just yesterday, if I'm not mistaken, or the day before yesterday, they increased the context window from 4,000 tokens to 16,000 tokens. Ugh. This means that my PHP app, which I create in one file, I have a single PHP app, not a single file PHP app, and every time I need to make changes or, for example, optimize the code, I can simply upload all my code directly into the chat. And the bigger the context window I have, the bigger the code I can, of course, send a chat to GPT and ask it for help with a certain, let's say, part of this code. Here. And this is the next level, really. Wow, very interesting. You have already answered the three following questions directly to me. That's why it turns out, you had Twitter, how many subscribers did you have initially? You said that you just showed a demo of your product, and you already had people sign up for it. How many subscribers were there at that time? Well, it wasn't enough, I don't remember 300 either, maybe somewhere around that. And, that is, it was registered with 300, or this post just somehow got viral. Yes, yes. Precisely because, look, if there were, let's say, several interesting stories would come together here. First, I created a very simple, but understandable product. So I was able to demonstrate it. Secondly, the story with GPT chat itself in the role of CTO. Thirdly, I'm still a first year, not my first year in marketing, so I know how it can all add up. Thanks to this, I was able to tell this story correctly. It, of course, caught the attention of the audience and went viral. This brought the first pre-orders. I then used this story to share my

experience a few more times, which also went viral. And that's how, actually, this case got people's attention. And the icing on the cake was the sale of this project three weeks after its launch, which I also used as an interesting marketing story. And it actually got the most reach there, meaning there was a tweet with 3 million impressions. Here. And all of this together, in just a month, increased my audience 10 times. Yes, and how much did you earn before selling on that site? I know this Andrei. Oh, he performed at our University when I was studying there. I know that there should be some sales. During the first day of pre-orders, there were about 1,000 dollars in value, I do n't remember, I think it was 12. Well, I raised the price there 12, 15, 17, 19. That is, it's always a dynamic process. It needs to be tested in real time. That is, in the first day, yes, somewhere around a thousand, about 1,000 dollars in pre-sales. That is, there was no product. That is, I didn't give it to anyone at that moment, because it wasn't on Google or this web market. Well, yes. And for how much, if it's not a secret, did you manage to sell it? Well, I'm not disclosing this information because the new owner asked in the document not to disclose it. Okay. And who is the new owner? American. I am not disclosing this information either. Okay. Only the person I bought it from can speak about it. That is, this is basic, basic. Well, I didn't know that it was a secret, so I'm sorry. No, it's not a secret. There are just people who, for example, on the contrary, turn it into history, right? That is, I mean those who buy. That is, they simply say: "Okay, there was a case, I'm buying it, I'll continue to develop it there, and so on and so forth. " There are people who don't want to do this. They are uncomfortable doing this. They, well, don't want any publicity. They are interested in a specific business, not storytelling. and that's absolutely their business. That is, I cannot influence it. Okay. Can you please share what neurons you still use in addition to the GPT chat? That's what you've been doing for all this time, they've proven to you that they really help in your work, they're effective, they're cool. You can share with the audience. Yes, everything, everything is very simple, you understand. I start from the tasks I have and accordingly automate what I actually have. I always keep my finger on the pulse. I'm curious. I research everything that comes up. I'm specifically subscribed to several newsletters that are sent out every day, with links to what's new and so on. Even at the development stage, that is, there are literally projects on GitHub that don't even have a web interface. That is, they didn't even launch the demo. I test all these things, watch them, and monitor them. As a product, I'm interested in what it looks like, what the interaction is in certain use cases, how I can learn something, and so on. As an entrepreneur, I'm interested in what new business models are emerging around AI and so on. That is, I always research it, but I don't use any other models simply because I don't have problems in which I can use them. Here are the specific tasks I told you about. That is, for me, the GPT chat is the main resource that I use every day, the web interface for accessing the API. In addition, my partner and I made a new app for the company Setup. This is a platform that allows you to publish applications on Mac. And we made an app that also interacts with the Open AI API and helps with math problems. That is, this is working with text in any application. It's trivial that you have texts in your notes, you can rewrite them, personalize them, and so on. And whether to interact with PDF files that you have on your computer. That is, well, just a universal app that allows you to work with texts. This is its main task. What about security? That is, he has access to all your files on the computer? No, he doesn't have access to your files. You only call it when you need it, by highlighting text. That is, for example, if I want to look at the summary of some contract that I have in PDF on my computer, I will open the PDF, select the text via Command A, and then launch the app so that it summarizes this specific text for me. All. That is, I just dictate to her the text that I highlighted. It does not have access to any other files or programs.

And what do you think about Auto? I think he has access to a computer. Well, at least he can do a lot of things. Well, Auto GTP now works in the browser, so yes, you can give it access there. These products, AI agents, so- called, they are in the very, very early stages. I also test them all, see how they work. Everything is in the very, very early stages right now. That means, of course, that they have some kind of future. Well, it's a very interesting niche, but the quality there is mediocre, well, it works for a long time. That is, if I really need some information here and now, I will look for it using other methods. If I really need to scrape the web, write some kind of report, or something like that, then yes. Yes, you can use that too. Well, then vopros takoy, well, I will deviate a little from the topic, yes, here is this startup. What do you think about the development of Iya in the next 2-3 years? What will it be like? No idea. I don't think anything about it specifically for one simple reason. I'm interested in building something from what I have now. And I will always be in the context of building some kind of product, technology development. As these technologies develop and I already have some experience interacting with them, at the developer level, at the user level, I will come up with other products and create other products, and so on. I don't want to act as a futurologist in these cases, because I'm, well, not interested in it. This is not my role. My role is to build. So I calmly sit at my computer and build. Okay. And what can you advise people who want to build something with the help of this, maybe some life hacks, advice from yourself that you, for example, did not understand before you created the project, but now you can see it very well. Now I see very clearly, for example, that, firstly, in order to create something, you need to write a lot of text. If people watch this video and, well, I don't know, read its title and there's something to write, there GPT chat will write a program for you, an application Chrome extension, it doesn't matter what, I'm sure a picture will appear in their head where they write: "Create me a program that does it, press Enter and get the code of the entire program, and also at the end a button to download it to the Play Market. " This will not happen. You will have to write a lot of text, write, rewrite it, clarify it, and so on and so forth. So be prepared to structure your requests. My experience interacting with people shows that people are not able to express their thoughts in a structured way. And this is their biggest problem when they interact with GPT chat or any other AI. Because you can only get a quality answer if you create a quality request. And in order to create a quality query, you need systems thinking, you need to understand, as I said at the very beginning, how products work, what components they consist of, and so on. You need to take an A4 sheet of paper, draw, make diagrams of how it might work, break it down into components. Then try to describe each component in words, try to describe what exactly you want at the output, how exactly it should work, the sequence, and so on. That is, it is very important to understand that in order to get the maximum possible response, you need the highest possible input from someone who wants to create a product or even just get a good text as an output. Thank you very much for the interview and good luck with the project. I will share your link. I think it will be very interesting for the guys to see what you write about new projects on your Twitter. I know that you are currently building another project. You have a hat on Twitter. I came in, something interesting, but it is not clear what. Therefore, I think you are creating intrigue. No, not really. This is exactly what I have already done, I have already done it. That is, this is a project for the soul. And the essence of it is precisely that this is a project that will teach people to think systematically. It is designed for this. And it is precisely this systems thinking that I was talking about, which is needed to interact with AI in a quality way, that if we summarize in general what people will gain in the future when interacting with AI, those who learn to think systems will win. Here's who wins. And that's exactly what mental walking is designed for. This is a simulator that

uh, helps you learn mental models in specific situations that arise in your life. That is, you can use it to think about a certain situation through the prism of mental models. And mental models are aggregated, let's say, views of the world in a certain way. For example, thinkers use mental models in their work. Watch an interview with Naval Ravikant, or Ray Dalio, or Elon Musk. They all talk about the mental models they use. Moreover, any book you read has a certain mental model embedded in it. And the goal of this book is to teach you this mental model so that you can then use it for your life. And, uh, there are hundreds, maybe thousands, of mental models. And each of us already uses them at our basic level. What are you, what am I? You may not even realize that you are using a particular mental model, but it is a specific way of thinking about a situation. How do you think in a given situation in order to, for example, make a decision? Very often, we, as humans, experience anxiety and confusion in situations that we haven't thought about, for which we don't have mental models. And the task of the mental block is to help people learn at least a few basic mental models for work, for life, for, I don't know, relationships. There are a lot of them for every area of ​​life. So the AI ​​in this is a very good simulator. Describe the situation to him. Just come in after our conversation and describe to him some situation that, well, is going on in your background somewhere and you don't fully understand how to resolve it or what decision to make in this situation. Describe mentalwalk in three to five sentences. It's better to be five, because the better you describe the situation, the better mental model he will choose for you. And you will see how it works. He will offer you a certain way of thinking. He won't make decisions for you. He will not say: "Do this or that. " He will say: "Look, based on what category this situation is, I suggest you look at it through the prism of, for example, the Eisenhower matrix and tell you how to apply it step by step to your situation. And you will be able to apply it yourself, get the result in the end and see three more questions. I structured the answer in such a way that you get three more questions that allow you to look at this situation from a different angle. Here. And this is the best simulator that I use all the time. And I just made it to use for myself. Here. Well, I just shared it with people, because I say, this is a project for the soul. You just intrigued me. I'll go to those now. And your data is saved. Now you will see that I ask him and find out what problems I have there. That's exactly why nothing is saved. Therefore, keep in mind that everything you correspond with him is stored only in your browser. And even if you go to another browser, you will see this story, because everything is stored only locally in the browser, but you as the creator. Question. Well, I understand. It is possible, of course. Don't take my word for it, but I promise you that no data is stored. I promise on record. And thank you very much. So far, nothing is stored on June 15, 2023. Yeah. This story once again shows us that now, with the help of neural networks, the opportunity to start your own business and do what you always wanted has never been so close. Therefore, if you are just starting out, believe in yourself. Perseverance, neural networks. will lead you to your result even faster than you thought. Like, subscribe, be inspired, watch, learn on my channel with the help of neural networks.