72 часа зарабатываю на вайбкодинге — получится ли? — Transcript
Full transcript
- 0:01I gave myself 72 hours and 20 bucks to
- 0:04build an AI app using v0-coding. I
- 0:07didn't write any of the code by hand.
- 0:09An AI model will write all the code.
- 0:11Then I'll have another week to try and
- 0:14sell it to someone. At the end of the
- 0:16video, I'll show how much I earned and
- 0:18what the expenses were. I've already
- 0:20been through it all, so I'm telling the
- 0:22story in reverse. By the way, making
- 0:24the product turned out to be the
- 0:26easiest part. Well then, let's go. Here
- 0:29is my idea. I used to build bots for
- 0:32clients. The client says: "I built it,
- 0:35got paid a buck, and moved on." What
- 0:37can I say? It was just regular
- 0:39freelance work, but with v0-coding. I'm
- 0:41talking about these two videos; if you
- 0:43haven't seen them, you can check them
- 0:45out. But this time, I wanted to create
- 0:47my own product. I build it once, and it
- 0:50works for me even while I sleep. This
- 0:52thing lives right inside a Telegram bot
- 0:54. It’s a mini app. No App Store, no
- 0:56need to install anything. You just open
- 0:59the link and start using it. Let me
- 1:01demonstrate my idea. I stopped
- 1:04wondering where to write down my
- 1:06thoughts and deleted all my task
- 1:07managers. That's why I v0-coded my own
- 1:10app, GX Planner. Now my entire life is
- 1:12in one Telegram bot. You don't even
- 1:14have to open Telegram to use it. For
- 1:15example, I can hold down the button and
- 1:17say: "Tomorrow at 3:00 PM I have a
- 1:18meeting with Andrey. For lunch, I had
- 1:20borscht. I need to remember to pay for
- 1:22the internet. And I also spent 500
- 1:24rubles on a taxi." GX Planner
- 1:26automatically figures out what's a
- 1:28meeting, a meal, a task, or an expense,
- 1:30and sorts everything into the right
- 1:32sections.
- 1:33The system organizes everything itself.
- 1:36One message instead of ten apps.
- 1:44That’s the product I built, and I
- 1:46even sold a bunch of subscriptions for
- 1:48a very nice sum, but more on that later
- 1:50. So, what do you think? Pretty awesome
- 1:53, right? Or is it complete garbage?
- 1:56Whatever. Moving on. The rules are
- 1:58simple: 3 days to build, 20 bucks for
- 2:01Cursor Pro, and 5 bucks for the OpenAI
- 2:04API. There are no other investments
- 2:06here. Then 7 days of promotion. I'm not
- 2:09buying any ads. My TikTok is starting
- 2:11from zero. I have my own Telegram
- 2:13channel with a small audience. It
- 2:14helped, but not by much. I’ll
- 2:16definitely show you the stats for all
- 2:18social media. I think the idea is clear
- 2:20to you now. Now let's rewind to the
- 2:22very beginning. Just as the idea was
- 2:25starting to form, 10 days ago: 3 days
- 2:27to build, a week to sell, a blank
- 2:29canvas, and zero code. Hello everyone,
- 2:34my favorite v0-coders. Let’s start
- 2:36our experiment. This is day one. I’m
- 2:39not touching the code at all. To start,
- 2:41I’m putting the idea together in
- 2:42Figma. I already have an idea, and
- 2:44it’s not just some made-up notion,
- 2:47but a pain I experience myself every
- 2:49single day. Look, it’s simple and the
- 2:52same for everyone. I often have great
- 2:55thoughts, but nowhere to write them
- 2:57down, so they just get lost. Or another
- 3:00example: I spend money and forget how
- 3:02much. I don’t keep track of my
- 3:04expenses at all. I also don’t keep an
- 3:06eye on meetings, and sometimes I might
- 3:07miss something. That’s why this
- 3:09feature needs to be implemented. And
- 3:12since I watch my diet and work out—
- 3:14well, for athletes, we need a fitness
- 3:16tracker. If you try to do this manually
- 3:20, you end up with about seven apps, and
- 3:22you have to choose where to put each
- 3:24entry. But that’s just absurd.
- 3:27That’s the choice I’m getting rid
- 3:29of. This is how I came up with the SGX
- 3:32Planner. One place where you dump
- 3:34everything by voice, without even
- 3:36opening the app. And the built-in AI
- 3:39will automatically figure out what it
- 3:41is and where to put it. A couple of
- 3:43ideas went straight to the trash, like
- 3:45a feed inside the app and various
- 3:47streaks. Well, that’s just overload
- 3:50and has nothing to do with the problem.
- 3:52Then I thought: "How can I make this as
- 3:54lazy as possible?" Typing really takes
- 3:56too long, so voice it is. And it would
- 3:59be even better not to take the phone
- 4:00out at all. That’s where two features
- 4:02were born. I’m marking them with an
- 4:04asterisk. I’ll come back to them a
- 4:05bit later. It’s going to be really
- 4:07intense. I’ve also already defined
- 4:09the pricing plans, how we’ll monetize
- 4:11, and in what way. Essentially, my full
- 4:14MVP is ready, which is pretty cool. And
- 4:18once the picture came together, I
- 4:20started turning it all into a prompt.
- 4:22To build the prompt, I immediately
- 4:24create a mind map with all the app's
- 4:26features and work out the architecture
- 4:29so the AI understands the structure
- 4:31instead of it being a mess. I
- 4:33immediately decide where the app will
- 4:35live. I’m making it right in Telegram
- 4:36. I get the mini-app bot token from
- 4:39BotFather, give it to the AI, and then
- 4:41it connects everything itself. No App
- 4:44Store, no moderation, no waiting weeks
- 4:46for a three-day experiment. My tool is
- 4:49just the standard $ 20 Cloud Code.
- 4:52Although I’ve long been on a more
- 4:54expensive tier from Codeium. I got the
- 4:56cheaper one specifically for this video
- 4:57, the same one most people have. And
- 4:59yes, another very important point,
- 5:01without which I won't make it. I'm
- 5:03talking about skills. These are
- 5:04pre-made instructions. I have five of
- 5:06them for the front-end so I don’t
- 5:07have to explain every time that I need
- 5:09a decent design. On its own, AI does it
- 5:12just terribly, in the most generic way.
- 5:15And I will also use MCP—not exactly
- 5:18skills, but direct model access to
- 5:20services. It goes to the database
- 5:23itself and pulls UI components without
- 5:25me. I'll drop the name and links in
- 5:27Telegram. Next, I feed the prompt to
- 5:30Claude, and it gets to work. And while
- 5:32our miracle is being written, I realize
- 5:34what I’m actually doing. It’s an AI
- 5:37B2C Six7 SAS. Yeah, a lot of jargon
- 5:40that just makes you want to barf. If
- 5:42you translate it into normal terms,
- 5:45it's a project made for everyday users
- 5:48with a subscription, all powered by AI.
- 5:50But what’s up with the "Six7" part?
- 5:52Honestly, I have no idea, I'm just an
- 5:55idiot. Everyone and their dog is
- 5:56churning out apps like these. And
- 5:58almost all of them are losing money. Oh
- 6:00well, I’ll make my own money-losing
- 6:01project just for kicks. Why not, right?
- 6:04At first, I’ll probably be right
- 6:05about losing money. But there’s a
- 6:07reason people make fun of these things.
- 6:09They never validate pain points or
- 6:11demand, so the projects don't last a
- 6:13week. My approach is different. I’m
- 6:15building it based on my own specific
- 6:17struggle. But will people pay for it?
- 6:19Well, we’ll test it on real people
- 6:21via social media. At the same time,
- 6:23I’m mapping out on this canvas how it
- 6:25should make money. The video leads to a
- 6:27bot. Everything in the bot is free, but
- 6:30when a user hits the 2-action AI limit,
- 6:33they want more and buy the Pro sub for
- 6:36290 rubles. Per month. Looks great on
- 6:39paper, right? But will even one
- 6:41stranger actually buy it? We’ll find
- 6:43out in a week. Well, end of day one.
- 6:46Claude built me an app, the framework
- 6:49is set, features are polished, and a
- 6:51marketing plan is ready. Everything
- 6:53looks better in the morning. See you on
- 6:55day two. Day two. Today’s plan is to
- 6:58fix the mess the neural network made
- 7:00yesterday. I woke up, ran the app
- 7:03quickly, and saw enough errors that
- 7:06need fixing. I also need to add the
- 7:08design and implement those two "star"
- 7:11features. Let’s get started. I’m
- 7:13describing the features in two prompts
- 7:15to handle it all step-by-step. And I
- 7:17sent it all off to Claude. As soon as I
- 7:19have some results, I’ll be sure to
- 7:21show you everything. Feature one. New
- 7:23iPhones have a side Action Button. Most
- 7:26people don't even know it exists. I’m
- 7:28mapping it to thought capture. I don't
- 7:30even need to unlock my phone. A couple
- 7:33of seconds after I record something,
- 7:35it’s already sorted into sections,
- 7:38and the AI categorizes everything
- 7:40itself. My philosophy for this app is
- 7:42that the most convenient tracker is the
- 7:44one you never even have to open. And
- 7:47for those who think it’s just for
- 7:49iPhones, no—you can record the same
- 7:51thing with a regular voice message, you
- 7:53just have to open Telegram. The second
- 7:56feature is a bit more hardcore. iPhones
- 7:58have a "Back Tap" feature. Almost
- 8:00nobody knows about it. I’m setting it
- 8:02up for quick expense entry. Like, if I
- 8:04buy something, I tap the back twice,
- 8:06and the expense is already recorded.
- 8:09You don't even have to unlock the phone
- 8:11. These aren't just some extra buttons.
- 8:14Every app competes to make the
- 8:17prettiest screen. But I want screens
- 8:20you never even have to open. Honestly,
- 8:22the best interface is no interface.
- 8:25What else is inside? Let me show you
- 8:27quickly. We can count calories just by
- 8:29snapping a photo of our meal. It’s
- 8:31elementary. Or we can track money in
- 8:34any currency. It factors in monthly
- 8:37limits, plus tips from an analyst. You
- 8:41can also track your habits and check
- 8:43trends. We have eight sections in total
- 8:45here. You also get morning
- 8:48notifications with a daily overview,
- 8:50plus a bunch of useful dashboards.
- 8:53Today was an awesome day, I got a lot
- 8:57done, so I can safely go rest and
- 9:00finish our app tomorrow. The second day
- 9:04really turned out to be the best.
- 9:05Everything was going great. I was in
- 9:07the zone, but on the third day,
- 9:09everything almost fell apart. Okay guys
- 9:16, day three, and two red flags at once.
- 9:19First: my twenty-dollar subscription
- 9:21limits are almost gone. I'm working
- 9:23right down to the wire. I really
- 9:25thought I wouldn't finish in 72 hours.
- 9:27My hands are itching to dive into the
- 9:29code, but we’re holding steady,
- 9:31brothers. This is a real experiment.
- 9:33Second red flag. I gave my vibe-coder
- 9:35brother a feel of the—no, not what
- 9:38you’re thinking—the app. And I got
- 9:41a report almost immediately. And it was
- 9:43for the exact feature I was bragging
- 9:44about yesterday. He asks the bot, "How
- 9:47much did I spend at cafes this month?"
- 9:49Yes, we have a built-in neural network.
- 9:52And it writes back, "I couldn't find
- 9:54anything in your records." Even though
- 9:56he has a ton of entries, it’s a
- 9:58pretty interesting bug. I need to sit
- 10:00down, look into it, and fix it now.
- 10:02That’s vibe-coding for you. Just as
- 10:04it is. I’m not cutting this out,
- 10:06I’m sitting down to fix it. Phew. Man
- 10:15, I struggled with that bug. I
- 10:17basically wasted the whole day on it
- 10:19because I was fixing the wrong sections
- 10:20. The neural network couldn't
- 10:22understand the problem for a long time.
- 10:24I barely managed to fix it. It turned
- 10:26out I hadn't properly defined the
- 10:28finance section. So it was just
- 10:30answering blindly. Anyway, I fixed it.
- 10:32Now it calculates from the records
- 10:35first, then replies. Same question. And
- 10:38we finally have a proper response with
- 10:40a category breakdown, so it works
- 10:43perfectly. A word on security. In short
- 10:45, the keys are on a server with UPass.
- 10:47They never touch the app. I didn't want
- 10:49to build payment processing myself. I
- 10:50connected a payment system through a
- 10:53manager, and that's how I’ll handle
- 10:55all payments. You don't even need any
- 10:57complicated layouts there. Everything
- 10:59works out just fine. Well, the 72 hours
- 11:02are coming to an end. Our timer is
- 11:04already at zero. I made it just in time
- 11:06, but the product is alive, that's the
- 11:09main thing. It's far from perfect, but
- 11:11it's usable. I think I'll display it
- 11:13for you in the background. The first
- 11:15task is complete. Building it in 3 days
- 11:17is realistic. And that's where
- 11:19something I was completely unprepared
- 11:21for begins. The product exists, but
- 11:24there's no one to use it. Absolutely no
- 11:27one knows about it. My idea on how to
- 11:31sell this miracle. I'm cutting the
- 11:33product into short clips, one per
- 11:35feature, and posting them everywhere at
- 11:37once. And let's see what comes of it.
- 11:40If you've enjoyed watching this so far,
- 11:42by the way, smash that like button.
- 11:44Let's keep going. Our favorite money is
- 11:46waiting for us there. Fast-forwarding a
- 11:49week ahead. I want to say that the
- 11:51clips are already posted. And right
- 11:53away, before they start writing in the
- 11:54comments, I'll say that I posted from
- 11:56several accounts on each social network
- 11:58. I tested one clip in different ways,
- 12:00changing the post time, description,
- 12:02tags, and seeing where the hook would
- 12:04hit. So all the reach further is the
- 12:06sum across these accounts. The first 2
- 12:08days there was silence. I filmed how
- 12:10nicely the voice input works. And, well
- 12:13, very few people watched it. I already
- 12:16thought I’d failed the challenge, but
- 12:18then it hit me what the problem was. I
- 12:20was showing features. Viewers don't
- 12:22wake up in the morning thinking they
- 12:24need speech recognition in a tracker. I
- 12:26rewrote it specifically for that pain
- 12:28point, for example, that I tap my phone
- 12:31twice and the expense is recorded
- 12:32immediately. Showing it in action. The
- 12:35first video was about the function, the
- 12:37second about me. And that one took off.
- 12:39It got 15,000 views in a day, which is
- 12:42just wow. I really felt that my product
- 12:45wasn't made in vain. And then
- 12:47everything started growing. The results
- 12:50of the week turned out to be
- 12:52magnificent. I was simply in shock. It
- 12:54ended up with 88,000 views. 1,000
- 12:57people visited the bot itself. 11
- 13:00payments were made from Telegram. That
- 13:03is my actual audience. The rest are new
- 13:06accounts from social networks. The
- 13:08first thing I saw were messages like
- 13:10these. It felt very good. And then came
- 13:13what it was all started for: after the
- 13:16first two days, the first paying users.
- 13:18And finally, we've reached the money.
- 13:20Remember the blurred payment screen at
- 13:22the beginning? Well, here it is. 623
- 13:24dollars for the week, or 47,000 rubles.
- 13:29A fifth of that amount came from two
- 13:30people. They bought lifetime access.
- 13:34Our expenses were 20 for the
- 13:36subscription, 5 for requests, and about
- 13:39598 dollars remained. This is an
- 13:42incredible sum that I didn't even count
- 13:45on. But 623 dollars isn't even the most
- 13:48important figure here. There's a second
- 13:50one, smaller, and it will drip into my
- 13:52account every month on its own. About
- 13:5518,850 rubles. per month. 65
- 13:58subscriptions at 290 each. But some
- 14:02will say, "People won't pay you for a
- 14:04subscription every month." But wait,
- 14:06290 rubles is less than a cup of coffee
- 14:09. I also factored in churn, because I
- 14:12won't be actively uploading videos and
- 14:14pushing this theme. By my estimates, if
- 14:17there are no new sales at all, it comes
- 14:19out to about 135,000 a year. This is
- 14:22just a forecast for now. The product is
- 14:24only a week old. We’ll see how it
- 14:26goes. And I won't be rambling on about
- 14:29how anyone can do this. Deconstructing
- 14:31a product and managing an agent is a
- 14:33skill. I had it before I even started.
- 14:35But previously, such an attempt would
- 14:37have taken a whole team or six months
- 14:39of work, while for me, it took 3 days
- 14:41and 20 bucks. Code is no longer a
- 14:43barrier. The barrier now is coming up
- 14:45with what to build and explaining why.
- 14:47Thanks to wipe-coding, the cost of an
- 14:49attempt has plummeted. This is enough
- 14:51to stop procrastinating and build your
- 14:54own project, or go build them for
- 14:56clients. I’ll post all the prompts,
- 14:58skills, roadmap, and promotion
- 15:00breakdown in Telegram. And don't forget
- 15:03to smash that like button on this video
- 15:05if you enjoyed it, and subscribe to the
- 15:06channel. Also, write in the comments
- 15:09what product you would test in 3 days
- 15:11and why people should even pay for it.
- 15:14Truly, it will be interesting to read.
- 15:16That's all from me. Go wipe-code
- 15:18something already. Hugs to everyone.
About this transcript
This page contains the full transcript of 72 часа зарабатываю на вайбкодинге — получится ли? by SGX, generated from the public captions YouTube serves with the video. The transcript has 2,739 words across 395 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.
What you can do with it
Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.
Free YouTube transcript tool
YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.