I Make $80k a Month With Claude Code... Here's How To Build Apps Like an Engineer, Not AI Slop

0:00So, I worked with a a company that was building software for restaurants, and they sold for Oracle for over a hundred million dollars. Most software is going to have most traffic from AI agents, not from humans. So, when you build software, you need to build it in a way where AI agents can do things on your software, not only humans. Everybody out

0:19there is saying how great cloud code is. When I look at it, it's it's actually quite quite bad. It's not because of the models or anything. It's about how it's being used. The planning is more important actual building. When we think about building a new product, we have five stages that we usually go through. First, of course, is what we call it What do you do? What is your job? What

0:41is your business? Yeah. So, um we use the term product studio when we describe our company, like not a dev shop, even though like probably everybody thought we are developers. We are a product studio, and we work with non-technical founders to go from idea, concept, or prototype that they have to a real market product in 2 weeks, 3 weeks, usually this around this timeline. Normally, the founders that we work with, like they are domain experts in some area, and they want to build a product, or they could be running, let's say, a service business, and they either notice like an adjacent problem next to their service, and they want to build a

1:17product to put it into the market, or they want to build an internal delivery system, like how we did, and to use it to deliver outcome instead of selling access to the software, making them some something that's called like an AI native agency. You've been doing this for exactly how many years have you have you been building now for other

1:33Oh, 10 years. 10 years. And how many hours would you estimate that you helped build? Only in the past 2 years, we've built more than 50. Wow. But, throughout my career, probably significantly more. Although, back in the day, uh I used to stick with a project a lot longer before moving on. But, we can say, just looking in the

1:53last 2 years specifically, so over 50. What was the biggest month you have for your business or your agency? Well, actually the our biggest month is actually going to be this month. It's because of all of the work that we did last month. So, this month is like it's time to collect everything. It's going

2:06to be $80,000. It's not like huge yet. Uh but if we compare with like say just last year, I mean our whole last year was 250. I've been the blocker for my agency because like over the years I didn't know if I wanted to build a product of my own, if I wanted to build an agency. Like I've been conflicted with all this stuff. And closer to the end of last year I realized that I can actually marry these two things and become an AI native agency where we build our plat-

2:32platform and then we sell the outcome. So, that's kind of how we got here to this point. What is the most successful project you've been a part of? Yeah, so I worked with a a company that was building software for restaurants. So, specifically like in the back kitchen, back office, stuff like that. And they sold for Oracle for over $100

2:51million. Wow. Yeah. And what was your involvement in that project? Were you helping build out the engineering infrastructure or what were you doing? Yeah, so I was the lead engineer. So, like I was essentially leading the development team. Uh trying to essentially turn business requirements into real software, real features. What else have you been working on? Like who do you help and and what have you

3:09seen grow? been quite a few. Uh so, for example, there's a founder who I've helped go from essentially 25,000 MRR to over 70,000 right now. There's been another founder duo where like they've been trying to build their product. It they completely failed with a different team. I rebuilt it in 2 weeks. They went to market, raised a $300,000 seed round, we continued working for about half a year, they raised another round of $10

3:33million. Wow. So, you've seen I guess in the back end of a lot of different software, a lot of different apps. A lot of people out there want to build apps, want to build software. And cloud code is hot right now. Yeah. What's your take on cloud code as a you know someone with an engineering

3:47background? Everybody out there is saying how great cloud code is. Uh in reality like when I look at it, it's it's actually quite quite bad. And it's not because of the models or anything. It's about how it's being used. Because the marketing that is out there right now, like it talks about how you can release a product into the market, but it's like the only thing this does is it helps you with the code part of it. But when you build an actual product, code is just a small portion. We do a lot more work before we even write a single line of code even back in the day. So

4:21now with Cloud Code, you're able to shorten the amount of time that it takes to actually write the code, but it doesn't really help you completely cut all of the other parts that it takes to actually build a good software product. I wanted to say first of all that we, like all of our code for the past year, has been fully 100% AI written. Like we haven't gone into the code. So like we're not against AI written code. We're just against what everybody is saying the way you are supposed to use Cloud Code right now. The planning is more

4:46important than the actual building. However, the planning is not just these steps. When we think about building a new product, we have five stages that we usually go through. First of course is what we call like the discovery stage. Like figuring out like what exactly are we looking to build. And when we're looking to build, we're trying to figure out what's the core workflow. Like the main thing that this product is supposed to do. So once we discover that, well, we also need to think about like who are we solving this for? Because it could be that based on who you are, uh who you're building for, you're going to prioritize

5:16one set of features versus another. Then when you go into the next stage, that's the product requirements documents. That's essentially where you put figure out like all the different parts that actually define the software. And we use four different categories, or sometimes we call them tiers, of like how we organize different features. So first of all, you look at all the features that implement the core workflow that we talked about in the discovery. Then after that, you think about supporting features. So that's what the core workflow uh essentially rests on. So these are features that essentially make the the whole core workflow even better. Then we

5:50look at table stakes, and that's something where like every product in this category, people are expecting those features to be there. That's like, you know, authenticating or like billing. Or if we're talking about like a marketing type of product, you would need to have some sort of dashboards. Like that's something that you would absolutely expect from that kind of product. Then after that, we we look

6:08into the category that we call delight. So, it's like the cherry on the cake. Like, what makes this product really different in the way that it does everything else that it does here. Because there could be a lot of other products that solve the same core workflow, but it's your delight that is going to make customers to come back to you specifically. A lot of people might feel like, "Okay, I'm just going to do these features." And maybe they're going to focus on one of the categories. Like, we believe and we've seen that it takes features in all four categories to

6:33actually have a market-ready product. But that still is just planning what like the requirements of the product itself, like what features are need need to be there. The next stage is the where we actually start thinking about engineering. That's what we call the system architecture, where you think of all the different pieces that go together and how they communicate with each other. You plan that before you start building. Because now, with a product requirements document and with a system architecture, now you're able to take all those plans and then give it to Claude code or Codex or any of the other tools that you prefer, and they're going

7:11to have a proper roadmap for what needs to be built and how things need to communicate. Because if you don't do that, most of the code that these models have been trained on is what we call spaghetti. What's going to happen is a few months later, you're going to start building, you're going to start working, then things are going to randomly break without knowing. And that's from the spaghetti code. But if you give it the exact rules of how things need to communicate and how you're organizing them, well, you don't have that problem anymore. So, these are the things that are like pretty much everybody that is

7:41non-technical that hasn't actually built and doesn't know all the different steps of building a software product, that's what they're missing. And Cloud Code itself doesn't know to tell you about that. Yeah. Could you tell me who you were as a kid? Where did you grow up? Where did you did you get born? Where did you get raised? Like, your entire journey up to the point that you're running a business

7:59out of Dubai. So, I was born in Moldova. That's a small country sandwiched between Romania and Ukraine. It held the title of like the poorest country in Europe for many, many years. I That's where I grew up, that's where I went to school, that's where I lived. I At some point, I got interested in computers very early. Um

8:16then I started to learn how to program. I specifically wanted to build a specific website. Instead of like going and playing with the other kids after school, I wanted to go home as soon as possible so I can continue to learn how to program with PHP back then. Then I was just like I I built a set up like a specific game. Like, it was a popular MMORPG that I had set up. I made a

8:37little bit of money off of that. Then I moved to the States, specifically to San Francisco. Uh that's where I went to school. I left to go back home, became a freelancer as a software developer, of course. That's where I had the opportunity to work with like a bunch of different founders. Tried to build a startup uh

8:54myself. And what age is this now? Uh I was 21 when I started freelancing. My first freelance contract paid me $3 an hour. So, it's like but I I I was excited because what this meant to me was like I was in Moldova and I was able to get a client from like a completely

9:08different place. And so, that that meant I was no longer stuck at just the rates and the money that you could make in Moldova. Back then, the average salary was $250 a month. Wow. So, it's like I was able to close somebody who's not in Moldova, which meant it opened up big doors of opportunity. Within a few months, I was making 10 times more than the average salary in Moldova, right? But during that time, I did try to like make an agency multiple times. I tried to quickly hire a few people. I got busy like from morning to until I would go to sleep. Tons of problems, issues. Then I

9:42downscaled and I repeated that several times. Every time, of course, I more experience I started to ask for more money and so on. Last time I grew the team to about five people and we were working like on a weekly basis also with a founder. Then like the last week that founder disappeared with $11,000 in unpaid bills. Like I couldn't recover that and I was like, "I'm going to downsize again just to make sure that like I I don't know if I want to do this." Well, during this time I also

10:12tried to build my own startup as well. The thing is like the first time I wanted to build my own product was like a year after freelancing. Like I was still doing freelancing but I wanted to build my own product. And the reason why? Because I wanted to build a better freelance marketplace because I was

10:24freelancing on Upwork. I didn't like it. There was a lot of issues that I noticed even back then. So I wanted to build a better version. But I treated that as a technical as a software problem. A marketplace is not a software problem because the software is just a facilitation. There were a lot bigger

10:41issues. And I tried that. I raised a little bit of money like not a lot. Plus I used my own. And then three years later I shut my doors down. What was happening like after all that chaos I scaled down. But what I did is I just I figured like, "You know what? Let me actually just find one person who wants to learn how to program, how to build software with

11:00me that I can just teach from scratch." A few years before moving to Dubai like I've been doing this and he started helping me out. That was like the first time when like I kind of started setting the roots of like maybe I'm going to build a product studio but I still wasn't sure yet. And I started

11:15delegating a few things to this person. So I've been working on the things that only I know. And then when that that person started helping me out with things. At some point we became pretty stable in terms of revenue and I was like, "You know what? It's time for me to move somewhere to make a change." Like I want cuz like in Moldova it's not that fun. It's a poor country. There's a lot of political instability and all this stuff. So, I wanted to move somewhere. At first, I was thinking, let me figure out a way to move back to the States, and then I found out about

11:40Dubai. And I watched a lot of videos about Dubai, and I was like, that's like a better version of US. Yeah. From traditional countries, Balkan countries, there's a lot of pushback when it comes to entrepreneurship and going out on your own. You should go to school and you should have a certain career. Yeah. What kind of pushback did you have from

11:56people around you? Was it supportive? Was it more conservative? You know what I mean? Like, how about this? Like, my grandmother, she she has been uh a teacher for more than 40 years. Now, she's retired, but like that's has been for a long time. So, like, the push were like, you should go to university, go get a degree, all this stuff. Like, she's been trying to push me in this direction, but like at that point, I knew that I wasn't going to do that. I mean, I did go, like that's why I kind of started going to university. I was like, and I I missed half of the

12:24semester, and then like I still got top grades. So, I was like, what is there for me to learn here? So, that's why I dropped out. But like, she tried to push a few times, and and that's basically it. Like, because I I became a freelancer, I just like shut down from everything else, from the whole environment, right? Like, I rented an apartment, I went in there, like I spent most of my time in there trying to figure this out. Right? So, like, I didn't have a lot of connection with the outside world, if you will. That's why I

12:51didn't really get that much pushback. Because in my case, like my parents were very supportive of everything that I did. A lot of people don't have this Yeah. you know, this privilege of being supported by their parents. In my case, I'm lucky that I did have the support. But even if you did have people that tried to hold you back or, you know, had hate towards you having, you know, going and doing your own thing, you didn't

13:09hear it because you were locked in. Yeah, I did not hear it. Exactly. Because it's like, I didn't go to hang out with people and so on, cuz like I was I needed to make this work. I knew there was like, there's no way I would go back to be an employee, to have a

13:24job. Would you want to touch on like, what would be a three-step foundation to for someone who wants to go and build something on their own? Uh I would say don't start with the build. We usually help some founders like a smaller portion through a system we call the launchpad. And it's actually just has three phases. It's research, pitch, and validate. So, with the research like you start from the idea and then you go and do a deeper research to understand like what's the competitive landscape, like what are some of the gaps that you can find in that market. Maybe you like if you already are in that industry, you can

13:53essentially feed into the system to be able to like to have a good starting point. But like you should still not build purely from like what you think you know. Like you still need to do some research to see overall what's the landscape. And in today's world, this deep research is pretty simple. You get cloud code, you get a good subscription, or let's say you have OpenAI or Perplexity, right? And you tell like what kind of deep research you want to do. So, you want to figure out like what are the market trends, uh what is the competitive landscape, what are some potential key insights that you can with

14:21you know pull from there. Uh what is like actually define your hypothesis of what you want to build. Then you go into the pitch and that's where you define okay like what is your positioning. What are what are you going to say? And like try to even build something like a pitch deck even though

14:32you're not going to pitch to investors. It's just a good exercise in general to know what your messaging should be for this product. And then the validate is where you actually design the experiment. So, we we like to use a smoke test. That's a landing page that essentially makes you feel like that's the real product like the product is live. You don't want to use a waitlist or a coming soon kind of strategy. You make the product feel like it's already live. And the way to do that you don't normally want to show screenshots of the actual product. So, to do that part you also build a what we call a happy path

15:00prototype. That's essentially like it showcases the experience of the product with that with pure dummy data. Like you don't build a backend, you don't figure out how to store data, all this stuff. You just focus on the actual experience that the product represents. You use that to pull screenshots to build different areas like for the for the landing page to make it feel like it exists. And when somebody tries to opt in you're like, "Oh, oh, you caught us a

15:20bit early." Yeah. And then with every person that opts in, you go and you set up a call, a meeting with them. During the call, you want to interview them to figure out like if this is really a problem for them. If this is your ICP that you're thinking that you're going to have. Also, on the call, you can show them the prototype just to get their reaction. The goal here is either to get a letter of intent or even to get them to prepay if the problem is painful enough. And then, you go and build. You don't build, you start from here. Cuz now this This is the part

15:50that became super simple. In the past, that was still the the the easier route, but it was a lot harder to get them. You may needed to spend like weeks and months to do this. Now, you can do it like you can do all the setup in 1 week, and then you run the actual experiment in like 1 or 2 weeks, and in less than a

16:08month, you get a go or no-go answer. I want to ask about the next coming 12 months. People are saying a lot about software with AI coding. It's dead or we're entering a fast-fashion era of software. What is your prediction? Well, fast fashion is correct. That means you the bar for standing out goes up. Software is not dead, not even close to dead. It's just shifting in like what it means to build software. All of the software is going to have a lot of AI agents to do things for you rather than like you having to click buttons. But, the clicking of the buttons still won't

16:43go away because there's sometimes it's easier to click a button to get something done, to run a workflow, than to write a prompt. So, like everybody that's saying that like the the graphical user interface is the opposite disappearing, there will be only agents, it's not entirely true. Like, you just need to find the right balance. And the third, which I'm seeing as like the most important trend that is happening, is a lot of software or most software is going to have most traffic from AI agents, not from humans. So, when you build software, you need to build it in a way where AI agents can do things on

17:16your software, not only humans. Because there's been a study where if your software is purely used through a graphical user interface, like clicking buttons and so on for humans, that costs about 45% more in tokens compared to software that is built for agents. So, now if we take two different pieces of software, one is because it's a beautiful UI and an amazing experience made for humans, but doesn't have an MCP server that AI agents like to use. And the other is not as good maybe visually, like but it does the same things except like manually like you you're not going to find the experience that enjoyable. And

17:52so in the past you would choose the other one. Now, people are going to choose the worst experience that is built with MCP servers and with the experience for AI agents, which we call agentic experience. So like people are going to choose this. So that's the most important trend within the next 12 months. Software is not dying. Thank you so much for coming here and

18:09sharing the book. Happy to be here. I learned a ton of things today. I hope the viewers did as well. Where can people find you? Where do they look you up? Yeah, so I'm on LinkedIn, just Alexander Dodon. I can easily be found or you can go on my studio's website, risework.ai. And from there, if you're interested to build something, you can book a call and

18:29we can chat about it. All right, Alexander. I great talking with you. Great talking to you guys. Hopefully we'll do another one soon, but yeah. All right guys, thanks for watching. We'll see you in the next one. Thank you.