Vibe Coding Doesn’t Scale. Here’s How I Build Now!

Where do you feel like you spend most of your time still, in terms of like manual cars, things that could potentially be automated? I think I burned over 200 billions in tokens in the last couple of months because it's not that easy to build stuff for an agency with 300+ stores.

Alright, it's the middle of October already and the pressure is definitely increasing. We are supposed to ship a lot of things for the agency, for the software Scaley AI. Yeah, the time pressure is definitely there, like Friday is around the corner, so the features need to go live for the agency. We already have a lot of people on the waiting list for our software, right? Which is great, but at the same time we also want to fulfill the demand. Kind of a roller coaster of emotions because at the same time it's super exciting, like we are almost there. 99% is done, but then this last couple of steps that still needs to be done to get things, just know they are the hardest, the steeper.

So we push hard, I will show you a little bit on how we build things.

So at the same time you can see we have multiple sessions running with AI, they are pushing hard 24/7 basically. At the same time we have two developers also fine tuning and getting everything ready to make sure we have the perfect mix of speed at one hand, but also quality, right? We cannot launch some random AI stuff. And there are levels to this, but at the same time speed is super important nowadays more than ever and especially in the software game, right? So in this video I also want to show you a little bit on how we build with speed and quality, how we create this dynamic mix of AI, vibe coding, software development with top tier senior software developers, right?

So we'll show you the pros, the cons, the strengths, the weaknesses, everything you should look out for so you don't make the same mistakes that I did myself in the beginning, especially when it comes to the vibe coding more complex features, right? So stay tuned and let's dive straight into it.

I'm not a developer and I never will be, I never was. So why do we structure our team like that? Why do we push so much with AI multiple sessions a day running all the time 24/7? I think I burned over 200 billions in tokens in the last couple of months alone. I posted a screenshot of that in one of my, my latest stories. So we are heavy on that, right? Despite of me having two developers in my team. So the question is, why do we structure it like that? What are the benefits and the obvious strengths that you need to double down on if you build your own software, your own product?

So we establish the strengths first and how we do it, why we do it. And then later I will show you with the obvious or not so obvious or next and potential weaknesses you have to look out for.

So first of all, like I said, I'm not a developer, but I do understand exactly what I need, what I want and how I want the product to look like. I can create a version, a sloppy version, not perfect at all, not running perfectly yet, but I can vibe code a potential first draft of the feature, how I want it to look like, especially now, right? Because we have a lot of underlying infrastructure already, different skills, different tools, different features already in place where you have everything mapped out properly. We have tested it. We know it works. We have the API paths, right? So I can create a sloppy first draft and I say sloppy on purpose.

It's obviously not perfect, right? It's very fast. Sometimes it works, but like it's not built for perfection. It's not built for scale yet, right? But I can do that and give that to my developers and they have a much better understanding of what they actually need. Because 80, 90% is already there with the right API paths. They know, okay, this is how it works. This is how it should look like without me brainstorming memo that's 10 minutes long and explaining them.

"Yo, you know what? This is what I want. I want the BMEX to do this and I want the search campaign to do that." It's very difficult for them to understand the vision. They are not media buyers. They don't have to only comprehend. They don't understand the pain points that you have to go through. That I understand with over six, seven, eight years of experience in this niche already. So there is this obvious gap and that's where I is really effective. At one side, you have me. I have the brain. I understand exactly what I need, but I can code. I can build. Not to the same degree as my developers who are extremely proficient in what they're doing, right?

But they don't understand Google ads. They don't understand media buying. They don't understand what we need to have in the product. And that's where I strong before you had this like massive gap. You couldn't really close it because they're building your thinking, but there is this massive gap, right? But now with PI, like I said, I have the opportunity to build it myself to a certain degree. And again, it will never be as good as what they are doing. The point of what I'm trying to say is like you can close the gap now, right? You will still need the developer if you want to build something more advanced.

You want to build something that doesn't break in the middle of the night. The moment you onboard more than 12 users, that's also why it's not that easy to get into the software game. But you can close the gap now between vision and building more effectively than ever before. At first, we had massive issues beginning of the year where AI wasn't that advanced. We were working with this software development agency and we made zero progress. Now we have our own developers. We vibe code 80, 90%, at least the first draft, at least the vision. And then we close this gap right between vision, strategy, what do we want to build?

Why do we want to build it and actually building it to perfection? Because it's way easier to work with something when it's already there, when you can already see it, when you can already see it working to a certain degree.

So massive gap before, now with AI you can shrink that gap and therefore massively increase the speed of development. But the speed has a price and you will potentially pay for it if you are not careful. Let me tell you a story almost over a month ago.

I was in Tokyo at that time in Japan. There's one more split that we have basically, right? Our software developers, I make them focus mostly on the software businesses we built. We built Scaley. That's what we focus on here primarily. But we also have another company coming. It's called Maporio. It's a marketing software for local businesses. It doesn't matter too much. We focus on Scaley and the agency now first. That's our main priority that's almost ready to launch. But the thing is, I make them focus primarily on that, right? It's a software. It's a bit more complicated. It's its own product. It needs to be able to scale to thousands of users and beyond.

And me with the agency, especially the simple things, and because I understand it so well, I build a lot of features mostly purely vibe-coded. And then the more complex ones, again, also the developers come in, help me support it, make sure it's 100% stable when rolling it out, right? So the story was the following. Like I said, mostly purely vibe-coding. What I do with the agency worked super well. But what we are trying to roll out was basically a client dashboard as well or in its simplest form, purely a data connection to Shopify, right? And make sure that it runs smoothly based on the data.

We get a lot of insights for our data analytics work, right? So we can push winners faster. We can exclude waste as fast. So we see which categories are trending up, which product categories are trending down. Basically like quite advanced data analytics, right? Not just like, "Oh, I set up a campaign. I scale it up. I scale it down." We want to take it to the next level with the agency. Dude, I'm still embarrassed of it, you know? It was a couple of weeks ago. So I basically told every client, "Yo, this is coming. Look at this. Absolutely game-changing. This is sick. Let's onboard you as well." I got everyone on it.

I had the Shopify connector sent out, right? And everything started to work smoothly. Client dashboard, boom. Obviously we had some like alpha tests first, right? With a couple of bigger clients, everything worked smoothly. Great. Highly advanced data analytics access. No other agency has it like that, right? Yeah, we could start to improve our work with it as well. So I'm like, that's it, right? I need it for every other client. So we are starting to onboard more and more and more of their stores as well. So we onboard all these stores. Everything starts to work properly until one day.

And I wake up in the morning and half of the dashboards are not syncing correctly. Data is completely outdated, not even accurate anymore in some cases. I'm like, cute. So that was my first kind of epiphany moment where I was like, damn, okay. That's why you hit the limit with vibe coding, right? Then the wipe suddenly turned dark real quick and it's not that fun anymore. I guess the most important lesson here is what actually happened now in the first place with my product, right? Because if it's just a minor bug, all right, cool, just fix it. So I basically had a couple of days doing that, right?

Just fixing and fixing and fixing. And it took so long because there are so many stores, so much data, right? So it's not even like a one-to-hour session, right? It starts to become more and more complex, especially if you want to build it properly. You want to make sure you don't fuck things up. You have a couple of micro-tests, try runs, make sure everything works smoothly. And it worked. It actually worked. Until I hit the next bug, the next morning. And it just started to become more and more and more.

And I got slightly pissed as well, right? At some point, we basically came to the conclusion, okay, the main bottleneck is we built something that works perfectly fine for a couple of stores, up to 20 stores potentially. The problem was I have over 300 stores in my portfolio, and we connected over 100 already. So everything I built, everything that worked so beautifully with five stores was basically doomed from the start. And why? Because I didn't build it for scale in the first place, right?

So the whole system architecture, the whole thinking ahead, the whole orchestration, it just wasn't there, right? You tell Claude something, it will do it. It will get it done. It will work. But it will not think ahead. It will not think about system architecture, right? It will not think about security. It will not think about all the ins and outs of payment processing, right? You can do it, but again, at some point, I would run into issues, right? So I told clients, hey, this will launch. And for the past four or five weeks now, we are basically fixing and rebuilding the whole thing because I wasn't considering the right system architecture in the first place to build it for potentially free 400 clients.

So that's where, again, like the experience of the software developer comes in, which you just cannot vibe code that. So yeah, painful lessons like that make you realize, okay, there are genuinely levels to this game. Yeah, you can vibe code some stuff. Yeah, it might work, but it will definitely break as well if you don't consider those things. So insanely grateful for the valuable lesson because, yeah, now we are building things very differently and things finally started to click because what we did after that is, of course, rebuilding the whole foundation properly with the developers

because we had to rebuild the whole thing basically, right? You cannot launch something that's made for 20 stores if you needed to be ready for 300 plus. And guess what? Now, even the vibe coded part runs smoother, right? Because the whole architecture below it is solid, that's safe, that's proven, that's tested. And on top of it, we can test new features easily, right? Without breaking the whole important essential thing underneath. For example, this week we launched our dynamic labelizer

fully vibe coded so far, right? But different because the foundation it fuse on is already there. The architecture is already there. And the feature itself is quite simple, right? It's basically a rule based label system that labels product based on performance. But the underlying data where we get it from, again, that's already proof now, that's fixed, that's working. So we could launch and test it really quickly, right? Because it's not a complicated feature. The whole data flow is working properly. It's accurate and suddenly out of nowhere, we build one of the most advanced labelizers in the game, right? Because it's not just a rigid rule based labelizer that gives you like, after a certain level spent, you get this label.

We build it in a way more dynamic fashion, right? Where it actually considers way more variables than just like, oh, I spent hits a certain level and if rawers hits a certain level, you get this or that label, right? So we look at multiple factors, like, of course, spent, but spent on variant level, spent on product level, rawers, sales in total, timeframe, of course. When the product got launched, what's the price band of this product? And all of this fuse into this labelizer, right? And that's also where our data advantage comes into play, right? And that's something I want to show you later

when we talk about how to build a proper software. How do we actually build great software in this new age of AI, right? Because like we outlined before, what's totally a valid fear, and you probably saw it on social media as well, right? Like we all scroll and if you're in the same kind of niche, you always see like, oh, software is that, software is that, software is that. So that was very encouraging for me as well after investing a shitload of money into building my software businesses. And I'm like, fuck, is my product really like that easy to copy, right? Can someone just come in, vibe code this stuff, and basically have a one-on-one copy that works perfectly, right?

But as we just outlined in the earlier stages of this video, it's actually not that easy to build a great solid product purely with vibe coding. To be honest, like if you face a very aggressive competitor, a very capable competitor with a lot of resources, a lot of money, their own development team, it will be 100% able to actually copy your product and have a version that functions completely the same, right? It's not rocket science anymore. And honestly, the IP of software, it doesn't exist anymore. Like that's true. All these videos on social media, they are right to a certain degree. It's not that complicated to rip stuff off one-on-one or just build something similar.

For example, for my own business, and I think I mentioned it earlier in the video,

we completely killed Asana. I wasn't happy with it anymore. We paid a shitload of money for such a simple tool. And then it wasn't even properly integrated in our own data. So I ripped it off. Took me 30 minutes to build it for my own company. Another one, two hours to like fine tune it and make it good. Another tool was Linktree. Absolutely shit product. I paid nine euros a month just to have this link on social media. First of all, I have like very little customization options. And then I see the little Linktree ad in the corner, right? Like I'm paying for it. If it's like a free version, I get it.

Makes no sense, right? Took me literally five minutes to build it and another 10 minutes to like fine tune and deploy it. 50 minutes in total to replace Linktree. So software products like that, they will be cooked, right? Like you don't have any chance. And don't get me wrong. These are huge companies, Asana, right? Like they have a lot of additional features and they are obviously not sitting still. They will deploy a lot of AI in the product as well. And they are huge, right? They have obviously a huge advantage of being that big already, being first in the market, being number one in the market.

They will be able to adjust, right? But if you come in as some random nobody with your vibe-coded tool and you build a tool like that, that's that easy to copy, then of course you will have issues, right? You will not be able to build a multi-million dollar company based on something I can copy and build myself in 10, 15 minutes. That's certainly the wrong way and that will not work. So there are a couple of things that make a software stand out and things that you cannot copy-paste and that's also why I'm not worried at all. Let's look at Scaley for example here.

So the Google Ads AI media buyer that we are building, right? And again, you actually could rip off the product in a very simple fashion, right? Although it's a quite complex software with a lot of rules, a lot of things to consider. Like you could do something similar. Dude, honestly, I would give you the source code and I'm confident in the fact that you cannot rip it off like that. Why? Because there's way more to it, right? That actually makes this a great product. And the first and most important thing that we have is the huge agency behind it. So I run a seven-figure agency.

Like I said, we manage over 300 store and there is all this data coming back and forth between the agency and Scaley, right? That makes it extremely powerful. So we don't just randomly create the software product out of nowhere. No, we have a sick amount of data. It's actually sickening. And if you go to our website, you see like, okay, over 200,000 Slack message, over eight hours of audio process that fuel into the chatbot of Scaley, right? So that already means it's not a standard random AI tool anymore, right? Regardless of how it's built or what it is exactly. Like because of the data, you already have a huge advantage.

So that's the first thing, you can copy that. You cannot copy the data, right? You cannot just copy the agency and build it like that.

And just to demonstrate how valuable data is, I will tell you a little story. There was basically this airline, I think it was in US airline. I don't remember the name that went bust, right? Like broke, done. And anyway, different things went for auction, right? Obviously all the planes, all that stuff. And what they also sold was their data. But not just any data, not even customer data. They sold their internal data, means older internal emails, older internal chats like Microsoft Teams, like whatever they're using, right? And that went for auction as well. And you would not believe how much these AI companies bid for that, right?

So Google bid for it, multi-millions, just for years of history of internal communication data. So you might think that it's not even that interesting. We're not even talking about customer data, right? Which obviously would be super valuable as well. I'm not even sure if you can sell customer data like that. But anyway, it doesn't matter. But that made me thinking, like actually I was sitting here, I was like, dude, I have an agency that makes seven figures a year, right? I was sitting here, I was like, I couldn't sell that for that price they were bidding on, right? There's no way I would get that price.

The agency is not broke, we didn't went bust. And you were talking about an airline that's basically done, right? And they sell their internal data. And just this aspect, right? Is worth 10 times, 100 times more than my agency that actually generates healthy profit with over 20 employees, right? So that made me think like, okay, that's sick. But it again emphasizes the value of data, right? So data is huge.

Forget the agency, it's about data, right? So data it is, and if you build your own software, you need to have the data. You need to find your variable that cannot just be copied like that, right? Another one, in the example of Scaley, I will show you. Another thing that we have is a huge community. So that obviously includes Instagram followers, YouTube audience, so just like the audience in general. But I'm not just talking about that, right? That falls more on the marketing and distribution. What I mean, for example, is our Skool community with around 2,000 members right now, right? And again, what do we get from the community?

First of all, a lot of data, right? Beautiful, isn't it? A lot of data that's important. But the second and even more valuable aspect of the community is that it creates this bubble around everything, right? Like a closed ecosystem, like a strong community. And again, this is not just about my software. It's about what can you learn from that and implement it from your own software. So it's not just a product that can be copy-based like that, right? Because if you can build it, someone else can build it as well. Data, the community around it, the personal brand, your brand overall, these are like some main things.

And then obviously, like other aspect distribution, of course, right? So, okay, once you have the product, everybody can build the product. You also obviously need to be able to market it, to sell it, right? So that's always and will always be one of the main things. So distribution is super important, and that has to be also one of the main focus points for Scaley. It looks very tiny here, but it is of massive importance. And then besides that, obviously, like other factors like product design, understanding system design, understanding the needs and pain points of your customer, understanding the whole flow of how the software develops, and really the vision behind it, right?

Little things that you might not even see when you used to do. So it's not just about the code. It's like, okay, how does everything tie in together, right? Is everything smooth? Really like understanding the vision and the design behind it. And everything ties into each other in the end. So then it's even more important.

Really looking at it from the perspective of like, okay, do I understand what my user needs in the first place? I will give you one example with my product. We are basically almost ready to launch, right? So I was testing it last week. Developer made everything perfect, front and beautiful. And I log in with a test account, go through the whole onboarding, and I sit here and I haven't tested my own product for quite some time, right? I let my CEO and my developer do its thing. And I sit here and I'm like, okay. And I've tried to see it from the user perspective, right?

So like these things are super important. You always like, you tend to get a little bit like hyped about your own product, right? But you need to be very critical, like, okay. The most critical, you have to be the most critical in the world. Like what can fail here, you know? What can go wrong here? Like you have to look at every single variable. So little things like that matter a lot as well, right? That goes way beyond just software and oh, someone can copy your product. It's about understanding the whole sign of it. How to create a great product, right? How to improve retention.

How to decrease churn. And in the end, how to market it properly.

All right, that's about it for today's episode. See you in the next one. Next one will be quite interesting as well, where I have a very in-depth strategy search with my CEO for Scaley, where we prepare for launch and really dial in the last couple of things so everything runs smoothly once we're ready to go live. So stay tuned.