How I Launched an AI Bot in Telegram and Made Money

[00:01] I see a bunch of bots in Telegram with tens of thousands, millions of users. And it even seemed to me that I was missing something, that something was passing me by. And although Telegram is still the most popular messenger, and I think it will remain so.

[00:14] how Telegram is growing, how they've reached a billion active users, and having listened to Pasha Durov talking about how great everything is, I decided to launch something of name is Sasha. I am a solopreneur. I launch projects alone using

[00:29] wipe coding. I do everything myself, from development to marketing and sales, without a team or large expenses, mainly with the help of Ii. Well, on the channel I share my experience, I show everything as it is, honestly and without stupid hype. In this video, I'll tell you

[00:41] how I launched my bot, how I chose the idea, did a rearch, analyzed , uploaded it to the server, and finally launched the project itself. Well, to proper wipe coding, you'll need a subscription to the basic neural network. If you

[00:56] guys from Payter can help you with this. This is the sponsor of this video. The cool thing is that with Payter you communicate with a live person who helps you at every step. And in general, all this happens very quickly. On average, it takes 8 minutes from submitting an application to paying for the service

[01:09] account login and password. You will simply be given card details with the bank details that you enter there yourself, and the payment will go through. Or they will simply pay the invoice using your link. You can pay with any Russian card via the Fast Payment System (SBP), even with a

[01:24] credit card. Well, if you have a studio, or you work on commission, or you are an agency, then Payra works with legal entities and provides all the necessary documents. I'll leave a link to Piter in the description. Switch, pay for your subscriptions,

[01:37] use foreign non-Russian networks, and do it with the lowest commission on the market . The first thing I usually do is look for competitors to get inspired by their solutions. In general, since

[01:49] we are now looking for an idea, we need to understand what we are going to do in general. And for this, I put together a Minecraft map where I wrote down everything that can be created in TG . And then there are wrapper bots, yes, the standard ones, these are

[02:04] GPT chat bots, where there is a simple model, you just provide a certain amount of generation or communication in the chat for free and then take money simply for giving people access to the GPT chat. They are the largest in terms of users. They have

[02:19] 500,000 here, there are bots with millions of users, and it seems that this is already such an overheated niche, but I still see new projects being launched in it every month . The second topic is more

[02:33] specialized bots. That is, we don’t just take the bare GPT chat, but we somehow improve it with the help of pronts, additional functionality, and somehow specialize it. That is, we say in advance that we will not be for everyone, we will be for a

[02:48] certain niche. We're narrowing down a bit. Well, for example, and the photo sessions, yes, I liked it. Then there's a cool bot for reviving old photos. There's a really cool theme there. You upload some old photo and it turns

[03:02] it into a high-quality color photo. And, uh, in short, maybe even make a video out of it . They advertise themselves there in a very cool way. You can see there how old relatives and grandmothers are crying over these photos. This is a cool topic. Then there are all

[03:17] photos. This is a cool topic. Then there are all sorts of bots for creating songs. If you want to drop your rap album there, then you can now create a song further. And all sorts of specialized bots. Well, there are psychologists, maybe a bot

[03:32] there, a gardener's assistant, where you send it a photo of your apple tree, it tells you what's wrong with it, whether it's sick or not . In general, we are narrowing down a little into a more specialized niche; we are not doing everything for everyone. The next topic

[03:46] doing everything for everyone. The next topic is, uh, mini-ups. The funny thing here is that Telegram claims to have 400 million active users in these mini-ups, but I have a hard time believing that. And you can create different services using mini-ups.

[04:00] All sorts of task trackers, habit trackers, you can create services for business, CRM systems, all sorts of project software. But somehow I didn’t want to go here. And looking at what's here in general , 99% is some kind of crap, all sorts of

[04:16] slippers. Well, in general, I don’t really want to go here. The next topic I considered was B2B. Well, that is, uh, to make the same CRM system, to make some software for sales departments. For example, there was an idea

[04:30] bot that monitors whether a manager responds to clients and provides some feedback or helps them follow the script and sell. That is, a bot for analyzing correspondence for business. Here.

[04:47] clients here are either experts, bloggers, or companies. And I didn’t want to come here either, because the sales issue here is very complicated. And I don’t really like B2B sales in general. There you often have to sell for a long time. At the same time, transferring someone from the

[05:04] X CRM system to ours will be super difficult. In general, this is definitely not a project for a ten-hour challenge, so I missed all that too. Well, and write out the topics. This is a custom development. That is,

[05:18] you can simply say there: "I make bots in Telegram, I make projects in Telegram," as if taking orders and making money from it. But I don’t really like it here either, because you have to sell there and sit there. Oh, and I'm really annoyed by all these

[05:34] acts, closing documents, amendments. In short, prepayments, then you have to collect the accounts receivable. In general, this is definitely not a challenge topic, but there is money there . Here. Well, by the way, I already

[05:53] refuse such work. And then the creation of TG channels. Well, this is also a complex topic. Now, it’s probably not particularly relevant. The market is oversaturated, and you need to invest a lot of money in traffic and generally be careful

[06:07] not to get bots. Well, and also all sorts of automation agencies, wipe-coding agencies, I also dismissed all of that. And in the end, I decided to settle on, uh, some kind of specialized bot. And at first I just wanted to make a bot, uh,

[06:22] that would stupidly generate pictures of a banana. In general, I went with this idea . I think, well, okay, since there are a lot of such bots, I don’t see anything wrong with copying. I'll make the same bot, try to advertise it, get

[06:35] new users, and launch the whole thing , try to sell it. Here. But then, here's a spoiler: the idea changed, and the next step is, uh, I

[06:47] wrote out a plan in my mind map, like how I launch projects in general. The next I launch projects in general. The next step is resarch. Whenever I get an idea, I usually first go to the GPT chat and start discussing it with him

[07:00] . So, I went to the GPT chat and pitched him my idea for an image generator. He was skeptical about it, saying it was the Red Ocean and easy to copy. In short, he doesn't advise me to do this topic and

[07:15] advised me to switch to some more specialized software or specialized bots. Well, it’s like what I showed you about specialized bots. One of the ideas he suggested to me was a

[07:29] carousel builder for Instagram. And I liked this idea. She's cool. I took note of it, but decided not to do it through the bot, because it’s inconvenient, because there are, well, a bunch of small nuances in the interface. And here's some advice for you right away:

[07:43] you should immediately think about what kind of flaws might occur in the product during development. Well, for example, let's take the same carousel generator. How does this work? Like, a person posts on Telegram, and the bot

[07:57] can then share on Instagram. But if we do this in Telegram, there will be a lot of nuances. Well, for example, how are we going to show him these carousels within Telegram? That is, we will send him their photos. OK? And if he

[08:12] wants to fix something, what will he write in the chat? Correct this text for me on the in short, it's not very convenient from an interface point of view. And there are still so many

[08:24] visual bugs. Well, that is, if Nerona starts adding too much text to a slide, or the layout is somehow crooked, or there are or there are word wrappings, crooked letters, in general, eh,

[08:38] it’s easier to do this in visual editors. In short, there will be a lot of edits. Therefore, such thinking in general, that it can go That is, the more projects you create, the more you will see where

[08:54] you will stumble during development . So just try to do something. And then you will immediately understand right away, like, well , most likely there will be some problems here. This is where I run into a bunch of edits. Here I

[09:06] will spend 3 days to get this sorted out. Eh, and right away, in short, you will know. Another idea was the same: these are the photo shoot professionals. So what kind of pain is there? People need photos for social networks , and, and, and, they don’t have beautiful photos

[09:20] , and they need to go to photo sessions there, uh, take pictures of themselves somehow, and you can make a bot that does exactly these kinds of business photo sessions. Ah, in short, this is also a cool idea. Again, there are already many such bots. I also took

[09:36] note of it and then I kept pushing it, pushing it, asking questions, and eventually a cool idea was born to make a bot meme. bot meme. Here. What's the topic here? I thought

[09:49] the idea might be something like this. This is an entertaining bot where you upload your worst selfie, and the bot generates a photo for you in the style of successful success. Well, like, there are 101 roses in my hand, I'm next to a lambo, I'm having breakfast like a

[10:04] billionaire. All these things. It seems to me that this could be a hype, given the agenda that everyone is discussing now. Basic minimum, luxury maximum. In short, everyone is trying to appear richer than they are. And this

[10:17] topic can get out of hand. The idea is, of course, completely cringe-worthy, but it can be played out in a funny way. There's a chance it will just go viral. And immediately, when the idea came to mind, I was like: “Damn, I wish I could make a series of Shorts

[10:31] .” You've probably seen these Shors with the voice of a news anchor. And make three 30-50 shorts, where with such a presentation, like infacigans launched a new bot that allows you not to fly to Dubai, not to rent cars there. Now

[10:45] you can do it in a couple of clicks. And these are the kind of meme-y things you can throw in. There is a chance that they will get a lot of views. Well, there is always a possibility that nothing will work out at all and the whole thing will be complete [ __ ]. Nobody

[10:58] will be interested in this. Here. Well, in short, I chose this idea for the challenge. I because it could be shown from a Vibecoding perspective. It has everything you need: balance, development, some kind of admin panel, and image generation.

[11:13] Overall, it's cool. Next, the next step. Even before development, I decided to check whether it was even possible to do this. I just asked the GPT chat test prompt that I can just generate somewhere and, well, generally

[11:28] see if it will work or not. Ah, I went to freepic, I have some generation left there. I chose my worst possible selfie, made a couple of variations, and overall I was happy with the result. Well,

[11:41] I would have appreciated it, for example. Of course, you can tweak it further, train this model somehow, write cool prompts, but, well, the basics turned out to be easy. And I decided everything, here, here I have already decided for sure, I take this idea and do it. Next we need to

[11:54] start development. And then those who launched startups might say something like: “Well, we need to check the demand there, analyze the audience.” They also often conduct Kozdevs and interviews with clients. And in my opinion this is

[12:10] no longer relevant now, but now I will explain why. Back in the day, when development was very expensive, and before we went somewhere to develop, we had to be 100% sure that the idea worked. So, we would first go, schedule

[12:24] calls, find out what users were struggling with, sell to them, and see what they were willing to pay for. Now, almost any MVP can be built in, well, a few hours or a couple of days. And that's why this thing with the goats is

[12:39] simply irrelevant. Well, it seems to me that you will simply spend more time searching for , and then analyzing all of this, than, well, wiping code it and than, well, wiping code it and developing it. That's why I think we

[12:52] need to immediately start some minimal project, buy happens there. In short, don’t get too carried away with these long interviews, these long interviews, market research and all that. Therefore,

[13:05] we immediately move into development. And in this same chat, I simply ask GPT how best to knows their history, the context that we discussed. He starts recommending languages, frameworks, and libraries to me. and some connections on which all this

[13:21] can be launched. But I will still say here that the architecture of neural networks is a very weak point. They often offer hackneyed or non-scalable solutions. For the first start this is okay, but later

[13:35] problems may arise. And here, well, I don’t really have anything to advise you. You just need to be knowledgeable about it, have a sincere desire to understand these technologies, and truly love it, and then you will already know what to knit with what and how to do it. But

[13:50] again, there are test questions that I always ask the neural networks. That is, you can drive him, ask him a little questions. For example, what if ? This

[14:03] architecture that you proposed, will it even withstand such a load? Or, for example, if you have a different idea, you can ask: "What if I later want to can ask: "What if I later want to

[14:16] easy to do or will I have to rewrite everything? In short, try, I don't , or something, test it with some kind of test questions, and then it will start fixing itself on the fly and offer

[14:29] cooler solutions. In the end, I ended up with this stack. And the interesting thing here is the Grammy library, which is specially made for working with TG bots. And the same library for working with AI. Hey, this one is also free,

[14:47] it allows you to make requests to neural networks and do a bunch of other things in literally five lines. Very convenient . You don't need to write kilometers of code there, force the neural network, and then fix all these bugs. There you go. There is already a

[15:01] library where everything is tested, everything works. For those who don't know what libraries are, if I explain it completely Basically, it's like an app solve certain problems. So, someone's already written the functionality for you. It's the same thing here

[15:17] . And it's cool that in the programming world, most of these libraries are free. The only thing that bothered me about this stack is that, well, it can all crash very quickly under heavy loads. Yes,

[15:31] even if 50-100 people start generating photos at the same time , it will all hit limits, queue up, and hang somewhere in the background on the server. All this will heavily load it, and the bot may simply not respond or respond for a

[15:45] may simply not respond or respond for a very long time. But at the start, I usually, well, skip them, because if the project ends up being live, the whole thing will added. Okay, let's move on to development. Next, I go to GitHub and

[16:01] create a new repository there. This will be where our project code will be stored. Then we'll transfer it to the production server from here. Deploy. But first, I need to wipe-code it all, write all the code. I opened AntiGravity, cloned

[16:16] this repository to my computer, and before starting development, I installed the necessary skills. Skills expand the capabilities of the standard one. And here's the list of skills I came up with. I'll leave it on my Telegram channel with links. I also have a video about the skills

[16:32] on YouTube. Check it out . Lately, I always start with another brainstorming session inside the code editor. And there's a special brainstorming skill for this, which will allow me

[16:47] to answer some additional questions, and, overall, the result will definitely be better. I describe the prompt in as much detail as possible: what I want, what stack I need, what I want, what stack I need, the idea of ​​the bot itself and useflow, how

[17:01] it should work for the user step by step. I describe the logic. I say that I need an admin panel, and I list what I need in the admin panel. I need to see what settings I need to look at, how many

[17:13] users are registered, and I'll list all the libraries. He specifically said that I absolutely need to be able to configure these prompts and able to configure these prompts and presets for photo generation in the admin panel. And then,

[17:26] long story short, I ran into some glitches in Antigravity itself. I had to wrestle with it a bit Antigravity itself. I had to wrestle with it a bit , creating new chats, simply returning an error message for every message I sent . So, if we have

[17:40] this brainstorming skill installed, it starts asking us additional questions, and there are sometimes very high-quality questions in such narrow areas, asking what needs to be done and how. I recommend you use this. When we simply answer the questions step by step,

[17:55] it launches another skill, which we also installed. The skill is called writing plan, and it records all our requirements in a separate document and then creates a task list for itself . And it works much better than the

[18:08] standard planning in Antigravity. Well, everything is basically clear here. Your neural network is simply She leads you step by step. You answer her questions, and then she gets to work. And here, by the way, it's important: if you, for example,

[18:22] understand that the design of your bot is critical, and the bot doesn't ask about it at all, it's better to point this out. Well, just write it outright: "For me, design is also super important; it's key for me."

[18:36] Let's dismantle it too, think about it." That's it, Antigravity starts development and at some point just crashes. I think they were having some kind of problems. Not a single message in the chat worked at all. Anyway, it managed to

[18:50] single message in the chat worked at all. Anyway, it managed to do a couple of tasks here and in the middle of creating the database schema, it just crashed and never came back. Oh, and I crashed and never came back. Oh, and I thought that it would get back up in 5 minutes

[19:03] and decided to switch to another task for now, to come up with a name for the bot. I went back to the GPT chat, generated several options. In the end, I chose this name on the rich one. Antigravity never got up again. And I had to

[19:17] switch to the Open AI codex, which I also have. And, honestly, then I, well, was super worried. I thought that now, uh, the codex wouldn't be able to pick up this project and finish everything. I would then be exhausted fixing bugs, because

[19:33] Antigravity stopped there right in the middle of the task, and I didn't even know if the codex would have enough context understanding to To pick it all up and finish it. I asked Codex to read the project, read the documents we created using

[19:47] our brainstorming and skill, and understand where we left off and continue development. And here I really respect the guys from Codex. He read everything,

[19:59] understood everything. He immediately told me, like, we stopped right here. It seems that this is ready, but this is not yet. And I was really surprised at how cool he picked it all up. It's funny, though, that at some point he decided to cut corners a bit

[20:12] install the library for a beautiful interface in the admin panel and would generate some of his own components. And then I stopped him and told him: "No, let's still move according to plan, download, install, everything as it should be." In short, it

[20:27] took him about 15-20 minutes to write it all. And it was really fast, really fast, once again my respect to Codex, right here, their new model. Well, it's a pleasure to use. And

[20:42] then, and he gave Give me the instructions when you're finished, explaining what to run and how. And if you're not sure what to do, just ask for more detail, like, "I don't understand this, give me a step-by-step explanation."

[20:56] give me a step-by-step explanation." There. Then my next task was to go to Superabase and create a database there. Subase is a create databases, create authorizations, and store your files.

[21:09] They have a great plan where you get two projects for free. Next, we'll need some software to generate the images themselves. I'm using fil AI. It's the most generating images specifically for developers. We needed to

[21:23] developers. We needed to get an API key there and connect it to our software. So I went to the interface and got the key. Basically, it's all easy. The neural network will guide you step by step and show you everything in detail. I don't really

[21:36] want to show you the screen here, because my project is actually working. I'm afraid I'll just forget to blur out some API keys somewhere. And that's it. Someone bad could take advantage of this. When everything is ready, we launch our project. And immediately I got an

[21:52] error about some endless redirects. Anyway, I immediately got nervous again, Anyway, I immediately got nervous again, thinking that now, because of this change from Antigravity to Codex, I'll drown in edits, and I got a little upset. But I threw

[22:06] this error into Codex a couple of times, and everything launched successfully. It's interesting that Codex decided to make its own design here. An unusual one, actually. It's a bit wonky, of course, but it certainly doesn't look like the standard purple

[22:21] design from Claude. The only thing that was wrong was the white text on the active sections in the dashboard. I quickly fixed all of this in a couple of prompts so as not to irritate me, but because I'll be spending a lot of time in the admin panel . And I just filled in the

[22:36] missing settings. This is the choice of models for generation and the choice of models for determination, and the gender in the photo, for gender analysis, because for us, gender analysis, because for us, depending on which photo is on which

[22:48] gender in the photo, The image will offer different presets. Next, to work with Telegram, we need to connect Grok. I'll explain why this is necessary. Telegram works by requiring an external address to send all the

[23:01] messages users write to your bot. But since our project is currently running on a local network, it doesn't have internet access—or rather, it's inaccessible from the internet— we need some kind of intermediary. We need a

[23:15] redirect everything to our local network. Here. And for that, there's the OroK service. It's free up to a certain traffic limit, but the free version is quite sufficient for testing . And again, a piece of advice: if you're working with Telegram bots,

[23:29] test everything and connect Grok. Well, having learned from experience, I asked them to create a special field in the admin panel where I can simply to create a special field in the admin panel where I can simply paste the link it provides and click "

[23:42] register Webhook" so that Telegram sends everything to this address. Messages from the bot. This is also a life hack for you, so you don't have to edit them in the configuration files or in the code in the project itself. You can

[23:56] the admin panel. It's very convenient. That's it, when everything launched, we entered all the settings . We're testing the bot. I launched the whole thing and tried it out. All the commands are working. That is, the response came at the start. It accepted the photo and started analyzing it.

[24:12] It accepted the photo and started analyzing it. Then an error occurred. I went to the codec and told it, "Error, fix this whole thing." And these are the first photos I got. I see that it didn't include my face. Apparently, the neural network

[24:24] just generated the image gene for us there, not taking our reference into account. We fixed that too and got the first acceptable result after a couple of iterations. By the way, yes, I also moved the settings for these presets to the admin panel, so it would be easy to

[24:38] create them later. When I got the first result, I saw that it was already 2 a.m. because of all these stories about changes and changing the location, I should have gone to I decided to continue in the morning, and while I

[24:52] was lying in bed, falling asleep, I also calculated the business model, what prices we should make some money from this. Well, basically, I came up with the rates in advance, because it could be done easily from the phone, not at the computer. In the morning, I woke up,

[25:07] tidied up the bot a bit, added some messages, changed the text of the starting message, set up, uh, this meme-like rating of the photo itself. Well, that's how it works. The bot analyzes the photo and determines the gender.

[25:23] There are different plots for generating for males and females, those same presets. Well, for example, for, uh, girls, it's "gave me 101 roses," for men, it's "closed a deal for a yard." Basically, different things. And if there's no person in the photo at all , then I have to

[25:39] inform the user about it, so we don't get bullied with all sorts of cat photos and people don't just throw in some test images. In short, it determines that it's a person, uh, and displays it depending on Different settings for each gender,

[25:52] filters, presets, whatever you call it. Then I went back to the GPT chat and started generating prompts for each scene. I created five scenes for each gender and

[26:06] added them directly through the bot in the admin panel. I already had all the functionality ready, I just copy-pasted it. After running a few tests, I realized it was running a few tests, I realized it was working at an acceptable level. I usually

[26:18] create the concept myself, and I always add monetization. And it's cool that here I literally set up payment in stars for one prompt. I simply specified the required tariffs, specified that I needed a Telegram Star, and how many generations to give for

[26:31] Telegram Star, and how many generations to give for free. And again, my respects, the code worked right away in one prompt . And I tried making a test payment on my own , and it went through without a single error. While I was doing all this,

[26:46] yes, I realized even more, damn, what a cringeworthy idea this is. And in general, when I test projects with this particular bot, I always have a ton of ideas about

[26:58] add it there, and maybe we should make it so that people can enter prompts themselves. And here it's important to really slap yourself on the wrists, to forbid yourself from adding the plan, because we need to launch as soon as possible. Wipe coding is a

[27:13] topic that's very exciting. You see that you're successful at something, and thousands of ideas immediately come to you about what to do next, how to improve it. It's such a creative process. And here it's important not to get bogged down, because you have

[27:27] some reserve energy that's devoted to the project, some kind of motivation. And to launch this project and generally get results, you need to spend this energy very wisely . Because if you get stuck in development for 2-3 weeks,

[27:43] you'll be tired by the end, you'll sit there thinking that something isn't working correctly, your motivation will disappear, and ultimately you won't launch the project. So my advice: slap yourself on the wrists and say, "I'm not going to do this now."

[27:56] My goal is to launch as As soon as possible. Anyway, that's it. The next step is deploying everything to the server. This is called " PLAY," when we upload the project from our well, right now, if I, for example, just turn off my computer, the bot won't work,

[28:13] but on the server it will run 24/7, and even if some error occurs, it will automatically come back up, uh, and come back to life. Basically, well, this is a production version. I rented a cloud server on Beget, and the

[28:29] project itself will be running on Quiify. It's a free utility that allows you to quickly deploy any project from GitHub. Also, if you're deploying something, I recommend you install Quiify and figure it out. It's

[28:44] Then I just go back to Codex and say, "Prepare everything for deployment and give me instructions on what to do." I will deploy on a VPS, I will deploy through quify. And when he finishes all this, uh, I just followed the instructions

[29:01] , and everything deployed on the server the first time. And this is actually a big victory, because usually I spend a couple of hours fiddling with my projects here. And again, the code, my respects, did everything without errors the first time

[29:16] . How cool is neural programming, this wipe coding is developing. This is just crazy. I was very surprised. That's it, the bot has been launched on the server. You can safely turn off your computer, it will still work.

[29:31] I checked everything to make sure everything was generated. And with this, we have completed the development of the first version. At this point I consider myself to have completed the challenge. The total work time was approximately 10 hours. The project really fulfills its function. You can

[29:46] laugh about the idea for a long time, but in fact, we did what we wanted. Well, then it was a matter of marketing. Eh, I didn't have much time before releasing the video. I just decided to post a bot on my TG and write a couple of posts in

[30:02] my Threads. And literally in about three hours I already had my first 100 users, and I even somehow received one payment. I really don’t know who payment. I really don’t know who could pay money for this. This is very

[30:16] interesting. Then came the first feedback. For example, someone wrote that the service is cool and useful, but they need to work on it so that it doesn't change the face and add generation based on your own request so that people can do business photo shoots

[30:32] so that people can do business photo shoots . Here. The next steps are as follows. You can tweak all of this, improve the prompts, improve the quality of the generations themselves, generate new presets, and roll out new features. We'll try to

[30:45] get to some kind of systemic payments. Because at some point, something will happen when you see that more and more users are entering your product and more and more people are starting to pay. There is no need to be immediately

[30:59] disappointed if you do not have the first payments. This is completely normal. Still, we missed the stage of this interview, calculating the pain points of the clients there. Eh, so here we just need to gradually tweak everything and bring it to perfection.

[31:13] If you have really bad metrics here and no one even uses it, that is, people don’t do more than one action, then, damn, the whole idea needs to be questioned. Maybe no one needs this. But if people

[31:26] spend time there and even receive some initial payments, even if at a loss, in the minus, that, I think, is already good. What marketing advice can I give you? Well, I have Tres working cool in my projects. My

[31:39] account there is not really a media account, it's not really promoted, but Threads provides very good organic reach. And if you just talk about how you're running a project, your idea, and sharing your successes, the most interesting thing is that

[31:53] the posts that get the most reach are the ones where you say, "I've reached 100 users, my startup has launched, congratulate me." And people like posts like these and ask, " What's this project? Send me the link so I can read it." Works great. Well, from paid traffic,

[32:07] it’s YandexDirect, the easiest one for advertising in Telegram. It sets up very quickly. They have this thing working automatically with bots, everything is synchronized. You can even ask there, I don’t know, go to the shop and ask someone for

[32:21] 3,000 rubles. They'll set up at least a crooked test campaign for you, but it's important to get some traffic going here, even if it's at a loss, to generate your first money. Because if you generate your first money, even at a loss, then you

[32:35] remove keywords that don't work, add those that do, and gradually it will turn into a plus. Here. But, well, my economy of this bot won’t handle paid traffic. I'll try to make this all go viral

[32:50] here, because Telegram takes a 35% commission for whistle stars. Uh, and there, well, in short, the economic situation isn't all that great. It will be difficult to count on paid traffic here That's all. In short, I considered this challenge complete. The project has been developed and

[33:05] launched. Someone even paid for the whole thing for fun. Of course, these were my subscribers. They might just give me 500 rubles for fun. throw. Here we definitely need to work on getting people who don’t know me to pay for this. Well,

[33:20] basically, just go into marketing and that's it. All. I will gradually finish this video on my VTG channel. I'll continue writing about how this bot is doing. Were you able to find any viral mechanics ? I will show how

[33:36] users grow and how the product changes. Well, or I'll close the whole thing if nothing works out. This is completely normal. I am generally in favor of closing down ideas that don’t work. So go to TG and subscribe. Thank you for

[33:49] watching to the end. I hope you liked this format. Subscribe to my YouTube. In the next video we will be making a service for creating carousels. And I want to try to set up marketing specifically through agents. Well, that is,

[34:03] so that I have a marketer who creates companies himself, uh, monitors their effectiveness himself. These are the kinds of things I want to do. And of course, we will create a service for creating carousels. This idea really resonated with me. I have a

[34:17] couple of thoughts on how to do this. Very cool. That's it, thank you. Very cool. That's it, thank you. Bye bye. M.