Transcription
We need to create a prototype from an idea in one sentence within 2 hours. A mini-version of the entire course. Any complex system that works, it originated from a simple system that works. And so always, at the beginning, we orient the LLM, who it should be. The market really needs this. Ready information about competitors. Okay. Well, that's cool, cool. Here's another one. And look at this, it even made some onboarding and some privacy statement appeared. I think you are convinced that this is real. It took a little over an hour and a half. Yes, of course, I talked a lot and distracted you, but overall, within 2 hours, we got a prototype. So, today we are launching our Founder course. What's today? Today we have a very simple, but at the same time very complex task. We need to create a prototype of this idea from an idea in one sentence within 2 hours. And overall, I hope to show you that this is indeed possible, and to show you what key steps need to be taken at this stage. And looking a little ahead, I will say that today's meeting, from the perspective of the entire course, is a mini-version of the entire course, because in subsequent meetings, we will delve deeper into each of the steps in the plan that we will first outline and then execute. Therefore, it might seem to you that at some stage of today's meeting, we have too superficially, for example, analyzed how an idea can be varied. Or we have too superficially generated the requirements, or the technical specification that goes into the model for generating the application. But do not be surprised, this is done consciously, because now our task is to go through the entire process end-to-end, and then we will go deeper and do things better and more accurately, because I globally believe in the approach where, well, which sounds like the law of systems, that any complex system that works, it originated from a simple system that works. And our task is to first create a simple system today. So, what are these steps? The first step is that I will propose an idea in one sentence. And it will be an idea based on one of the projects, unsuccessful projects that we did in the past. And I will tell you how long we worked on it. And today, based on this idea, we will go through the entire prototype, the entire process from idea to prototype. I urge you, either to do the same idea that I will propose, we will get to that. Or you can do it in parallel, if you already have an idea in mind, you can do it in parallel, I will provide all the prompts as we complete each step. And you simply insert your other idea, your own idea, into the first of the prompts, and go through the entire journey with us, but with your own idea, essentially. So the prompts are written in such a way that they can be adapted to your own idea. Therefore, today's meeting is about what is more convenient for you: either the idea that I will propose, or in parallel, your own, and see. In any case, we assume that after this meeting, as a kind of "homework," you will do it with your own and not just one. Second, after we turn the idea into one sentence into a more detailed description, for whom it is, why it is, how we will earn money from it, and so on and so forth, we will do a market analysis. That is, we will try to confirm or refute with numbers from the internet whether the assumptions we have in the idea are indeed valid. We have moved from a detailed description of the idea to a market analysis, which allows us to validate the idea and understand how it might need to be slightly refined. When we start doing it, it will be clearer. The third step is that we have an idea, we have market research, we need to turn it into a technical specification. To turn it into a technical specification, we need to understand what things to include in the first version of the product, or prototype, and what not to. And this is a process, we will go through this process, and I will point out the key moments in this process. Well, and then when we already have the technical specification, we, accordingly, generate the application based on this technical specification and, accordingly, get this whole journey from idea to prototype. What tools will we need and for what? From this four-step process, we will need ChatGPT, as I already wrote. And essentially, ChatGPT's deep research mode will help us, over three steps, to go from an idea in one sentence to a technical specification or Product Requirements Document (PRD), as it is called, which is usually written by product managers for the creation of, well, a technical specification for creating something. And second, today we will get acquainted with one of the popular tools for this kind of application generation. This is Lavable Dev, which will allow us to do the fourth of these steps - to turn the technical specification into a working application. And what is important, perhaps, to say is that you will be doing these actions together with me. And I will warn you right away, since this is generative AI, it may be that even when performing the same idea, even copying the same prompts that I copy, you will get a slightly different result. This is normal, do not be alarmed. This is how such systems work. We will talk about this a bit more at the next meeting. For now, I would like to be as practical as possible so that we go through the process of generating a ready prototype, so that you can try all this before we start delving deeper. So, what idea will we take? And, well, I just tried to choose an idea that would be understandable enough for everyone. Plus, I had some personal experience with it. I will talk more about why this is important later, but globally, as an example, we will take a product that my team and I worked on in 2012-2013. Unfortunately, it was unsuccessful. But the main idea was as follows. At one time, I read about the concept of flow from Mihaly Csikszentmihalyi. And an experiment was described there where people received messages on their phones several times a day, asking them to answer a few questions about their state. And then, after analyzing these answers, it allowed the developers, well, those who conducted the study, to better understand how and what affects the state of flow and how to understand when someone is in a state of flow. We, with the team, started with a similar experiment manually. I did it myself at first, and when I liked the results, I told the team. At some point, someone, we were running the iOS Developer School, the first in Russia. And we needed to come up with some idea for the guys in this school. And we decided, my colleagues came up with it, why not implement a mobile application that would allow people to understand their mood, for the purpose of understanding their mood, understanding what affects their mood, and how to manage it. And the time was, where we chose, there were two drums, you choose energy, that is, how tired or, conversely, energized you are. And in the upper and lower parts, you choose the mood, because you can be tired but in a good mood. You could upload photos, write tags, accordingly, and then you would get graphs where with whom, when you feel good and when you feel bad, for example, this application showed how to make yourself feel good. We will take this project as a test, but again, you can take your own idea. So, how do we formulate the idea? It is an application that allows people to manage their, to understand and manage their mood. Essentially, this is the idea. And, essentially, how we will break it down in the first step. We have a one-liner. We will feed this one-liner into ChatGPT with a prompt. Now we will do all this. And a very important point is that what interests us most now, and you will see that the prompt is written in such a way. We are interested in ChatGPT helping us expand this one sentence into a more understandable description of what we want to do, and perhaps ask us important questions that we should think about. So, let's open ChatGPT. I will open the already prepared one and comment on why the prompt is made this way and so on. And then I will give you 5 minutes to do it for your idea or for this same idea, read the result, and then we will discuss it. Okay? Let's start with the prompt, and I will explain why this prompt and why we are doing it this way. What is the task of this prompt? Why am I, why am I creating prompts in English? You can take this prompt and ask ChatGPT to translate it into Russian. Why am I doing it in English? Firstly, most of the information available on the internet is in English. And when we do some, for example, additional research, market research for step two, or when we formulate things, how LLMs work, when the source material was in English and we formulate the task in English, there is a higher probability and greater accuracy if the original task was also in English. Indeed, there are many studies that say: "You can write in other languages too." But these same studies say that about 20 to 30% of the meaning is lost, because, well, translations are not always correct, and when the LLM builds vectors, it will not always place an English and Russian word next to each other. Therefore, I will show you in English in the basic version, but I will explain clearly in Russian. If, by chance, you have difficulty, then just copy the prompt, ask ChatGPT to translate it into Russian, and, accordingly, you can use it that way. But let's figure it out. Well, first, perhaps the most important thing is that we always say who it is. We must orient ChatGPT, who it is. Well, actually, just so you know, without going into technical details of how LLMs work, I'll just say, imagine you have a lot of information, a lot of knowledge, but if you average this knowledge, the knowledge will be average. But if you create a vector of focus, you say: "You are, for example, a product expert with ten years of experience in Product Market Fit, growth," and so on. Then we give it a kind of focus beam. And when the LLM builds a sentence, it builds it by predicting what word or part of a word should come after the previous one, it will be more likely to choose words closer to the focus you are setting. Do you understand? Therefore, we always, at the beginning, orient the LLM, who it should be, what character we set for it, what focus of attention, what mindset, you know, what frame of attention we set for it, because then it will choose more relevant words. This is the first thing. Therefore, we say that, you are an experienced product strategist in product market fit. Second, what is important is to provide the context. And since the most important thing is the idea in one sentence, we add it here. These two sections are optional, but you will see now that I have filled them in specifically, because, firstly, it is useful for answering these questions yourself. What problem are we solving and for whom, I think, first and foremost, and we will do this in the prompt. This is precisely the task. What I want it to do is to refine the question. We will analyze it in detail using an example. I want it to suggest who the ideal customer for such an idea might be and what problems we are solving for them, what features in this product, how to earn money from it, and so on and so forth. And in the end, we want a concentrated answer. So, here you see, I have, firstly, inserted a one-sentence idea, and here I immediately clarified that I think it is an application for tracking mood for startup founders. And the main task it solves is to help them understand and manage their mood. And the key users, as I think, are startup founders. The better the model, the better the results. And if I ran this prompt on GPT-3, it would first ask me questions and then proceed to generate a detailed description of the idea. But again, this prompt will also work in GPT-4, so it's not a problem if you don't have access. But if you have access to better models, then of course, I recommend using them. If you don't have GPT-3, you can use, for example, Google's model, it's better. Indeed, it is. Let's look further. So, what is the result? Let's go. It clarified some questions for me, I answered, because it was GPT-3 initially, but if you use GPT-4, it will simply generate it automatically. But what we see in the result is the main thing, that, firstly, we have three different potential customer segments. And actually, this will be very important. In our subsequent meetings, I will explain why it is very important to have a good understanding of the persona for whom you are solving a problem, and what job you are solving for them, because this affects the marketing strategy, and so on and so forth. It is visible that there are different personas, different value propositions, a proposed MVP is described. We will analyze it in more detail later. The business model is described, and so on. So, what do we do now? Open ChatGPT, copy the prompt, insert your answers, your idea, for whom it is, what problem to solve. If you don't know the answer to a question, you can delete it. The main thing is that there is a description of the idea. And launch the prompt. When the prompt is ready, send a plus sign in the chat. You most likely won't have any problems with execution here, but these problems may arise at the last stage, and therefore, don't worry if something didn't work out, there will be a separate time where we will do it. But for those who have generated it, read what you got, because we will now do a small debrief and you will need it. Don't use deep research yet. You don't need to do deep research yet. To save time. Just run it without deep research. But globally, when we do market research, there will be deep research, and I will explain why it is good and why it is bad. If you try to answer the question for yourself, what, let's say, unexpected twist or something, you first formulated the idea somehow. What's new, if there was something like that, or what's unexpected, if there was something like that, did you see in the answer that really, well, you would like to note. If it asks questions, answer those questions and let it proceed. If you can't answer all the questions, tell it that I don't know the answers, decide yourself. For me, it was very interesting how it broke down into personas and generally what it highlighted as a persona for solo founders. One of the problems that solo founders have is this feeling of isolation and that your energy fluctuates constantly, because, well, you don't have a second person or a team, and it turns out that you are doing it alone, and this greatly affects the feeling of who needs it, who doesn't. And it's much more likely to quit when you're doing it alone. I've seen many founders who, because they started alone, just quit because it's hard. And I was interested in how it broke it down. Although I didn't provide this information at all. I just said it was for founders. And then we will talk about how important it is to understand much more clearly for whom you are doing it, than just a formulation like, say, startup founders. If you took VC-backed founders as your initial user segment, they are more likely to have money to pay for your service. And if you take solo founders, probably not. And you will see in the monetization section, look, when the business revenue model depends on which persona, which segment you choose. This will affect what monetization is possible. And that, for example, in the case of VC-backed or accelerators, perhaps a B2B solution can even be offered, because they have a task, let's say, to increase the probability, and when we do market research, for these VCs to increase the probability that founders will reach the end and not quit. Well, that is, they are much more interested in this. I think this was a simple enough task. Let's move on. Most people managed. Let's move on. And what most often happens at this stage is that you get some new ideas for monetization. Here you saw that I showed a case where my initial setup didn't even have the idea that it could be a B2B application, but we see that one of the models is indeed B2B SaaS. It's interesting that as a go-to-market strategy, you can package the benefits that an accelerator gives to founders within that accelerator. That is, you can promote the product through accelerators. This could also not have been there initially, but, accordingly, the idea appeared here. These are very interesting ideas. It's cool how you could test, do experiments for validation, and, accordingly, what success criteria can be formulated. Well, and additional risks that may arise. We have a product description, more detailed. There are some ideas emerging. And in my version of the generation, I essentially got three directions. Firstly, strategic directions for such a product. Firstly, the direction is that it can be a companion for solo founders. Here, of course, what's great is the niche. Very simple pricing, but most likely, well, a rather small target segment. I think it will grow, including from the ideas I will share today. But on the other hand, there is a B2B idea. A B2B play, let's call it that. It's to create a portfolio wellness platform and sell it to VCs or accelerators. For founders, it will be free. Here, it's more about B2B sales, very high check, built-in distribution, meaning you don't have to chase founders, they are given this as a benefit. But here, of course, sales will be enterprise-level, and, accordingly, the entire API idea, which is often not in the head at all. But an interesting thought, I recently wrote a post about Sam, about Sam Altman's interview at Sequoia, where he talked about how they came up with the idea for ChatGPT. And he said that sometimes, when you've made a cool technology, you don't know what product it should be, so you give it an API and see what people will do with it. And he talked about how they noticed that people liked to communicate simply with ChatGPT. Well, back then there was no GPT, just with the model. And then the idea for ChatGPT came from that. But anyway, we have a product description. We ideally, given these directions, we should validate and understand which direction is the most interesting and delve a bit deeper into the market. We will talk separately about how to do market research using ChatGPT. For now, we will do it in a very simplified way. Nevertheless, we will see important results. The main thing we need to do now is to launch Market Research. How our prompt and our positioning changes. Now it is important for us to conduct market research, so we set a frame in the prompt that you are specifically an analyst, a market analyst, who, accordingly, uses research, data mining, and so on to get results. We ask that when conclusions are made about our assumptions, because, for example, a hidden assumption in this idea, in the idea above, is that founders generally face the problem of fluctuating mood, right? These hidden assumptions that are within our product, they need to be extracted first, and secondly, confirmation or refutation of these assumptions needs to be sought. Therefore, we ask it to provide links. And as one of the first research tasks, we say: "Take the five most important assumptions underlying our product idea and check them," because sometimes it gives an understanding that we think it's needed, but we can't find confirmation for it in the market. And for each of the assumptions, we ask it to assess the level of confidence and explain why, so that we understand why a particular conclusion was made. Second, we want to know about competitors. Perhaps there is already one company, for example, we decided to make a search engine, and there is one company that controls more than 50% of the market. Essentially, we would like to know about this, right? With Google, we probably know, but with mood applications, I don't think every one of us knows if there is an absolute leader, how many users they have, what their positioning is, what their pricing is, and so on. For this, our deep research will be very useful. No less important is to ask it to think. And here, look, it starts to not only validate us, right? Remember, this stage is called "validate and differentiate," meaning confirm the assumption we have, and recommend how to best position ourselves. Therefore, in the third step, we ask it to recommend to us, from the perspective of six different aspects, where it sees gaps in the market that we could, by positioning ourselves under this guise, well, occupy our place, our white space that is currently available. Don't be surprised that it's so simple. Is it really that easy to find a space? No. But I want to demonstrate the basics. And at the next meeting, when we go into the deep validation phase, I will show, and in the homework assignment, by the way, I will show you how to search for these things more deeply. What else do we want to know? If there are any forums where this problem was discussed, any additional signals, perhaps posts, job postings. For B2B applications, for B2B ideas, job postings are often a good signal that there is a problem. For example, if I am looking for a target customer for our product, where we automate sales, then if I see that a company is hiring a certain type of salesperson, then I understand that most likely there is either a shortage of them, or the performance is insufficient, or they cannot pay enough. In general, there is a problem, and I can diagnose this problem with external signals. Therefore, here there are job postings, because sometimes a good signal, especially in B2B, is hiring a certain person. Well, and the last validation experiment. Again, I will explain this in more detail in the next meetings, but this is what we will need to plan market tests. Let's launch this process. It will take you 6 to 10 minutes. Therefore, during this time, I will comment on some points using the example of research that has already been done, and the link to it is here. This is already with deep research. We are launching this thing with deep research. You should have enough credits even on the free version to, accordingly, launch it. If, by chance, it has completed, send a plus sign. Read, send a plus sign, and start reading and noting if there were any interesting things. I will guide you through what was interesting when I read it. So, let's go through this. Firstly, what's great is that we get confirming evidence. Usually, when a founder pitches, they have some section where they write that, well, the market really needs this. And what I like is that, essentially, deep market research, which we are launching, prepares information for a pitch deck if we decide to make them and show them. And the most interesting thing, when these links are found, is to go and read deeper, if you are interested, if it is your...
So that what you want, what you want to do in the coming weeks, well, a couple of months. Therefore, follow the links, read in detail. But what I like here is that we have confirmation. And notice how the result is built, that it extracted the first premise, which is hidden inside my product idea for tracking one's mood for founders, that founders generally have stress and they need help, well, mental help. We see that this is indeed the case. But it's interesting that there is contradicting evidence, yes, that in principle, that many, that this stress normalizes and even founders are the same as you. In fact, despite the fact that it was difficult, still, they get on this needle and can't get off it, respectively. Well, and remember, we asked him to do a confidence level so that we understand how confident he is in this. Well, everything is fine here. We, remember, we had an idea that we could go to accelerators, yes, and that accelerators could be a separate B2B this. It's interesting that he found evidence related to this. which, yes, that there are initiatives from accelerators, for example, from Y Combinator, where they help founders work with, yes, with mental difficulties, and, accordingly, they even formed resources there that help with this. An interesting point is that, well, remember, I wrote that this, well, this application helps specifically to track mood and energy. This is how we saw it for ourselves, the product that we were making. And therefore, the researcher LLM extracted the premise that, yes, I assume that tracking these two parameters, with some context, will reduce the feeling of stress and improve my health. And it's interesting that there is such supporting evidence, but it's not enough. and that, yes, just tracking is not enough. This already gives us an important thought for when we think about the product or refine ideas about the product. If tracking is not enough, then what does that mean for us? We need some kind of feedback system and some behavioral advice to reduce stress. Well, what is called, for example, cognitive-behavioral therapy. It's cool here that we are starting to see some competitors. So here we see that, for example, there is some company Dailyo. And below we will see that there will be a separate section about, remember, we asked in the prompt about competitors. There will be, accordingly, we already see something about competitors, a little more about those, about the desire to pay. It's actually interesting that, yes, two very useful products are given as an example. You may have heard of them. Calm and Headspace. And that, for example, Calm has 4 million paying users and so on. But the question is, how many of them are founders? Yes, how much, yes, and the risk is that early-stage founders are very cost-conscious and therefore, maybe they won't have money to pay, and therefore you need to be careful. Well, here is ready information about competitors again, it's the same as what is usually in a pitch deck, about their traction, about if there was information about the paying audience and so on. And we can, accordingly, get all this here and, if necessary, research deeper. And now the third section. Remember, we wanted him to think about what differentiation options could be, that is, how we could differentiate ourselves, what free spaces are there. And it's interesting that the first space is what we broke down into when making the flow. In general, what is the problem? The problem is that discipline to regularly track your mood, energy, enter this data, is very difficult. And therefore, it would be cool, and this is the first idea, it would be cool to somehow automatically understand these parameters. Or you can ask the question differently: "Can, yes, what parameters are important for understanding the level of stress and mood, could be captured?" And because if we did it automatically, then it would help us increase the product's value, because a person does almost nothing, and they just receive advice on how to live better, as they say. And indeed, we encountered this in Inflow, that people are not ready to wait until enough data is collected to give insights. They want to get it as soon as possible. And most users simply did not reach the stage when there was enough data for us to provide any relevant insights and recommendations. Therefore, we narrowed the audience to people who, according to the diagnosis, need to constantly track their mood. These are people with bipolar disorder. But this is very important, I think. Retention is exactly that. That is, the proportion of people who use it after 15-30 days is very small. And I would say, at least from the experience of INFOW, that one of the key problems was precisely tracking, that is, people simply don't enter data, but they got the market and often it turns out that these research results can influence our ideas, that we will want to redo them and so on. We will not do this part now, because, again, our task today is to go through the entire process from zero to prototype, but in the next meetings, we will refine the ideas based on the market research results. We will just skip it now for simplification. Okay. Well, let's move on. Now we need to formulate. We have two artifacts. We have a detailed description of the ideas. We have market research. Now we need to understand what our minimum viable product (MVP) should look like. For those who don't know the concept, I will explain it soon. But the main idea is that we have some value idea that we want to bring to the client. For example, that it will allow them to overcome distance with less effort and faster. And working on an MVP, iterative work on an MVP, is when we formulate not some piece of a product, for example, a car wheel, which we plan to make in the end, but we formulate a minimal set of features that are sufficient to deliver the key value. And therefore, in the bottom picture, the skateboard, yes, it's not a car, but it's already something that delivers key value. I overcome distance faster and with less effort. And then, based on this initial version, which delivers value, we start improving it. Improving, as we are doing throughout our meeting now. Our task is to create a full prototype, and then in subsequent meetings, we will improve and improve, because our task now is to get a working prototype in minimal mode. Therefore, our key task in the third stage is to understand what is the minimum version of the application that is sufficient to still deliver the key value. And we understand that tracking as such does not provide value. We even saw this in the report. That tracking along with insights and recommendations on how to influence it, gives us the value of understanding and managing our emotions. Secondly, the less you choose for the first version of the product, the easier it will be for you to understand why a particular feature is not working. It's not working, in the sense, from the market perspective. And AI will find it easier to generate an application for this. That is, again, complex systems that work grow out of simple systems that work. Therefore, we also need an MVP to formulate this minimal version of the product. And the question we should ask ourselves constantly when thinking about an MVP. And we will talk about this in more detail in the third meeting. It is what should I remove from my concept, but still deliver key value. It's like Occam's razor, yes, when we, by removing the unnecessary, achieve a minimal simple solution that still solves the problem. So, how do we generate this MVP? At this stage today, in a simplified mode, the MVP will be formulated for us by the LLM chat GPT. But when we talk about this in depth in the third meeting, I will show that this process cannot be fully delegated. These are your decisions that you will have to make, and AI can help you, but in the end, you make these decisions. So, how do we do it? We have a prompt. So, here we are already saying that you are a great product, yes, and also a technical architect. And your task is to write a so-called Product Requirements Document or Technical Specification. And we immediately say that we plan to give this Technical Specification to the tool service Lavable, which will generate the prototype. Why do we add this to the context? I think it's clear to you, because we are trying to orient it immediately, that for example, an architecture needs to be chosen. You will see in the result that it will choose an architecture that supports Lavable, that it will make the option simpler, because, well, a prototype so that it is actually generated and so on and so forth. So, in User Input, we actually insert the two results that we have. This is the product from the first step and the pre-search findings from the second. The depresearch mode is not needed here, just insert. And what do we want? We want a description of the product, we want a description of the problem. We want a description of who it is for. This is very important. We want user stories. We want it to describe, only the most important user stories that need to be supported, and features. Let me open an example, you launch it while I show you what we have here. Well, what do we want? Well, it chose, because we didn't explicitly specify, but it chose to make it a companion, the first direction, out of the three directions that, remember, we had in the product concept, we could leave only one. Accordingly, the problem statement is precisely about retention, how many people track their mood per day. And one of the values that, well, we are trying to create, is that we don't want them to just track their mood. We want it to reduce the feeling of stress, and not lead to burnout and so on. This, yes, what are the user stories? Well, first of all, we must be able to enter, well, this information. We must see a card like what I have day by day in terms of mood. If the mood trend goes down, the application should draw our attention that we are possibly sliding into this, yes, mood cycle. We want to see the trend, by the way, in Inflow, this was our paid feature, the mood calendar, and people paid very actively for it and asked to add it. Now we have the technical specification, we have the ready technical specification, which we can now feed into the prototype generation tool and get a prototype. We go to lavable.dev. We don't add anything yet. We start with what you should have there within the free version, I think five generations, five messages per day. The first thing we do is in the prompt, in the prompt document, we take this part, and we essentially initialize an empty project. You will understand why I am doing this now. That is, we are not asking it to generate an application yet. We are just asking it to create an empty starting template for me. It will handle this quite quickly. We generated an empty project without a UI. It just created an shell for us. And now the most important thing, these documents that we generated, we generated them not just for nothing. We must put them into the so-called Knowledge Base of the Lavable project, so that later we can simply say: "Hey, you have all the necessary information in the knowledge base. Generate the product according to what is described in the technical specification, in this PRD, in your knowledge base." How is this done? Here in the upper left corner, most likely, here it is called slightly differently, the project, but you will see that there is a Settings section. We go into Settings, and there is a Knowledge section. This is precisely the section where all the contextual information that your Lavable project should know to generate the application in accordance with how you want it is added. And it is here that we copy our PRD. I copy it and paste it right here. And we see that we have put the entire PRD into it and save it. That's it. Lavable now has an understanding of what this project is about in its context. And it is enough for us to tell it: "Now generate the application in accordance with the PRD that you have in your knowledge base." It will work for some time now, but notice that it immediately, I didn't tell it anywhere, except that I added it to the context, that it will be about, well, our PRD, I didn't say that it's about Mood and Energy Companion, but it already says "I will create the founder Mood and Energy." That is, it has read from its project knowledge the PRD, it understands what it is doing and starts doing it. It's interesting that sometimes when it initializes a project, it might show you this button right away, and you can paste it here. But I showed the whole process because sometimes it doesn't show this button. Here I can choose mood, energy level, only energy level doesn't work, but I can fix the mood. Here it is, you see? It added it, and I see the current mood, energy for today, well, and some trend. It's clear that these are all placeholders, that is, not real data. And we will show you in one of the next meetings how to connect a database here. And, immediately, so that all, well, loggers, well, everything that you enter, it is saved. We, in principle, see, look, trends, and, well, calendar integration coming soon. But, in general, you see what else I want to show. Again, you have a limit for those who are on the free plan, so please be very careful with this. But, let's say, I don't like the UI. Two things I would like to show. First, the first thing you want is to show it to someone right away. Here in the upper right corner, you will see the Publish button. And if you click on it, then in the free version, it will offer you what the address of your application will be. You click on Publish and send the links in the chat. I already see screenshots. Super cool. So, you see, this application, this prototype, is already available on the internet. Anyone can open it. I can send it to my clients, my potential clients, my beta alpha users, and so on. Therefore, this is a very useful feature. To extract and deploy immediately, notice, you don't know anything about the technical aspects of how to do it. I want to show the second feature. Let's say you, like me, let's say you are a fan of, say, Apple's design or some other design, and you want the UI to look more like the design. What I like about Lavable, they have a feature that allows you to give it a screenshot. We have, for example, an Apple screenshot. And I tell it: "Please change the look and feel." Attach, that is, I ask it to redo the UI, to redo the look, so that it looks more like the screenshot in the attachment. And it will try to redo it now, and we will see how successful it is. Well, we see that it, well, this is a typical button of this Apple style. You can see the tabs, well, yes, it took the Apple look a bit. I see Sasha shared "Apple Journal App Inspired." Okay. Well. Cool, cool. What I wanted to note is that, you see, you have a chat, someone wrote, Alex, I think, that the data is not filled in or something else. You, well, can ask it to do it. That is, this is an agent, imagine that this is a developer, a front-end developer, to whom you gave the technical specification, he assembled this, but you don't like something, and you, well, let's say in this case, I have "logs today." And let me tell it, "Draw today's pulse with some dummy data." I'll open a couple more. While it's generating. Well, I liked what Sasha got. This looks cool, great. And the fact that it changes the entire background. Here's another one. And this is, look, it even made some onboarding and some privacy statement appeared. Oh, it's cool that you see, it's visible that it's for founders, first of all, Founder Mood Tracker, and secondly, what is on my mind now, what is bothering me, yes, I can indicate. Cool, cool. And I can even export to CSV. Well, okay. Let's look here. Log energy save entry. And it reflected it, but it saves to local storage. I will teach you how to connect the database later. Also, ML logger crashnet class. Yes, mode. We see, yes, how much we did the same work, but the results are different. Here is some rugby application for players, I understand, rugby. To search for clips, yes, clips, the necessary fragments from games, I understand, matches, yes? Okay. Well, let's assume I filled it in, let's assume I logged in. Well, it's clear, it doesn't work here anymore, but the idea is clear. And track. And this is precisely the dashboard of these assistants for or a broker. A broker, yes, probably Mikhail or someone did it. A broker for these for track loaders. Cool. Well, I think you are convinced that this is real. And yes, these are prototypes, yes, they may not be fully functional yet, but we see that in general, well, a little over an hour and a half has passed. Yes, of course, I talked a lot and distracted you, but in general, within 2 hours, we got a prototype, and then we can work with it. Or we can show it to the team or the customer and work on it together with them. Or we can share it and gather beta customers and show it to them. Or we can go on a call with a potential client and show it to him, showing it as an illustration, to gather more requirements and feedback on what functions, because when people see, they better accept it. Lavable supports import from Figma, that is, you can connect Figma designs and convert them directly into an application, so indeed, from the UI perspective, it will be much cooler. But let's try to summarize. And we went from idea to a working prototype in 2 hours, yes, we didn't do each stage very deeply, because our task was to go through it from zero to the end in breadth first, and in the next meetings, we will start going in depth. And you will see how, for example, for some ideas, we can use other tools, we will use other tools, including, one of the tools that I will recommend for internal app cases, because maybe you are making an internal app for automating your work, your team's work. There will be a slightly different product, but globally my main thought is that there will be different tools, and some of them require less technical knowledge, some more, somewhere deeper, somewhere more superficial, but in general, this is our task. What do we do for homework? Please, now do and fully spin this whole process for your own idea. Moreover, if you want to approach idea generation more thoroughly, then I would recommend a very cool prompt that will allow you to generate 100 ideas for your digital product. And not necessarily digital, I just made it software. You can make a physical product, a Telegram bot. In general, it's expandable, but 100 product ideas that could earn us $10,000 a month. And here's a very important point, that each idea should, and here you should specify the Target Audience and country or region. And I strongly recommend that you choose a Target Audience that you know or are well acquainted with, or that you represent. I will show you right now with my own example, what I would generate for myself. And you will get 100 ideas, some will still be repeated, although we tell it not to make identical ideas and so on. So don't be upset, it's okay. Our task is to get many ideas in breadth and choose. And each will be described, you will come up with a name and describe, 40-80 words about what the key idea is, how it can earn $10,000 a month. And, accordingly, what is its coolness. And let me show you what I generated, as an illustration, if I were generating for myself. And look, I say: "Give me 100 software ideas, specifically for immigrant tech founders, that is, I define myself in the USA." So this is essentially me, and let's read the idea. There are just some really cool ones. And it's interesting that the tone, that I added specifically "immigrant tech founder," that there are spins on the theme of immigrants, on the theme of the USA, on the theme of immigrant founders. Let's look. Well, this is a understandable idea, actually. You make, like, a set of prompts, for generating certain documentation that needs to be submitted in the USA if you, well, work with a foreigner, yes? Well, in principle, it's the same in Russia. And this idea is about that. And this one is cool, like, you know, if you have multiple passports, a very common situation, how, you know, not to pay extra money for, well, extra taxes, you know, this is like your accent. In general, this is a very expensive feature. Well, how to properly format your LinkedIn, how to understand the culture. This has happened to me very often in the first years. In general, when you are from another culture, you may not understand some jokes, some idioms, and some pop-cultural aspects. And therefore, it's cool, a cool assistant for this. Well, and you can read the whole list later. Why did I give you this prompt? Homework. If, by chance, you don't have your own idea or you just want to see what other ideas there are, use the prompt, describe it for yourself, launch it, and generate a prototype. A question about how to pre-test an idea. Yes, in the next meetings, we will evaluate and pre-test ideas using virtual users that we have prompted as an LLM. Yes, this will happen soon. Regarding prompts, there are two aspects. First, I always start. I ask ChatGPT to generate a prompt. Then I read it and correct some things, adjust them. How do I adjust them? According to my own understanding, experience of what should be. And you will see when we are evaluating ideas, that I have a certain framework for how I evaluate ideas, and I will offer you all these prompts, but nothing prevents you from rewriting them as needed, taking into account your aspects. For example, you may have some additional information or structure for market research or for what the technical specification looks like, or you may have a ready Figma with coding. I will show you how to connect Figma, and this aspect too. But globally, I start with GPT and then adjust. In the next meetings, I will show you how to connect some things, for example, to fully make a product preview plus registration for a waiting list for this product, which then can be used to drive traffic to this site. And if people convert, we will talk about all this, what metrics to track, and so on. In general, that's it. I hope it was useful for you. I hope a prototype was created. I hope there is this joy of learning, the joy of anticipation, that, damn, I don't even know, technically I can assemble something without developers. I think this is always inspiring. Yes, I warn you right away, we will still encounter difficulties. Not everything is as simple as it seemed, but I tried to guide you as simply as possible so that there is this feeling that it is possible.