Why our growth kept breaking, and how we changed

Good afternoon, my name is Ivan and I'm the president and chief executive of Markster. Today I'm going to talk with my co-founder on what did we fuck up in the past two years effectively. Because we did a lot, but finally we found something that we managed to grow our company zero to one million in less than six months. And not to mention we are unveiling our brand new product. So stay on with us.

Attila, the show is yours. So we are already live, like it's Perfect.

Amazing. So Ivan, what are we talking about today? You mentioned something about fuck-ups. Yeah.

Not my life right I meant the company no like I thought that it's high time to talk about really what the heck are we up to because we have been a little bit quiet I know I'm doing the live streams if you are if anybody of you any of you watch thank you um but like you know I thought that I saw so many bullshit on the internet and like you we keep sending to each other and I thought that hey maybe it would be good to do a PSA and maybe we could own some of our things that we fucked up some of the things that are working and surprising and you know maybe just help a couple of folks who are you know doing a startup or building a company or even just working on a big company what the fuck to do with AI right now Yep. Yep. Um, um, because, and I will start, um, I thought that things would be so much easier, so much simpler.

And I'm sleeping less and less each and every day just to be able to keep up with all the things that we are doing and It's my fault. I got you hooked on the Claude Code. Yes, I have the Claude Code subscriptions, Codex subscription, I have super Grok, I have Gemini, like I tried now the 3.5 flash. It's like you know my gamble so I'm neglecting all of my hobbies now and my hobbies work I go with the bed like I decided I know that feeling.

Yeah I said that you know we are I'm not going to have a girlfriend on this series b so it's like and now I've fully committed like it's a joke but like you know now I broke up like you know it's like you cannot and I'm single-mindedly focused and even like putting 20 hours on average a day into this. For me it's very surprising how many times, and I think I posted on Twitter or X, sorry, that my, the favorite catchphrase, I'm using Whisperflow now, and my favorite catchphrase according to Whisperflow, I don't give a flying fuck because I have 20 different terminals open and as you do, you do, hey, you have more, I can only handle up to, I don't know, 12 properly.

And I just you know I don't give a flying fuck you just make sure that you follow my intent and the policies and I just you know I broke a record and sometimes it works sometimes you write me very angry slack messages so really The AI revolution is how the tech bros have sold us is different, I would say. And now, you know, even OpenAI offered the latest batch, two million for each and every YC company to build stuff.

And you as someone who's working with OpenAI every day, what are your thoughts? I think it's an interesting evolution of where we started from. I remember the very early. Very early GPTs, you know like where I started copy pasting stuff into the chat box and then got something back and then it was half correct, half not. You know, if I look at the coding side, it couldn't even produce like a simple Python script, you know, without syntax errors.

And now like it blows my mind on what it's capable of doing, which is I think completely insane. But at the same time, you can't just like outsource everything to it because we also saw in our setup that it can go wrong and it can go very wrong if we do it the wrong way. So I think like what I've learned is like we need to be somehow still like structured and we need to somehow like build in those automated validations and mechanisms and you know schemas and so on that helps it drive the right results. And then it also gets you know spanked like a bad boy when it does wrong things uh and I think our job is to obviously create those guardrails uh in an automated way so we don't have to be like hitting enter to approve commands Mm-hmm.

All day all night, which is sometimes how it feels like, right? Um You know, that's my biggest issue with this because now, you know, GitHub was hacked, you know, Don't yeah.

Get me started on GitHub. That's why I'm pretty annoyed with it in the last couple of weeks. No, no, no, like we have been doing this for two years now and together and, you know, for 18 months you've been telling me and trying to pressure me that, oh, it's so nice to put everything in Git, it's so cleaner, it makes so much sense because I use Google Docs, right, and all of the different docs.

Documents and I had my little spreadsheets and everything I'm like okay let's try this Codex let's try this okay and now we have what 150 repositories into organizations and now I can barely access them so that's I would say fuck up number one maybe GitHub Enterprise is not the thing that it was anymore but no jokes aside like I think I never in my life would have thought that I would be I would be doing this what I'm doing now. So if you even asked me a year ago what will be, you know, as Microsoft's CEO, what the fuck am I going to do? Oh yeah, I am imposing guardrails and, you know, make sure that you don't stash anything, push everything onto remote, clean up the dirty work trays, and I'm like, You know all the keywords now.

Yeah, I'm like, I'm like, I'm You're a deputy. Probably safe and then you're get Yes, push origin.

And I'm a deputy city now, so you... So, you know, I'm the captain now. I'm Uh but uh you know that so proud.

Uh yeah uh well uh you it was a long journey. But like I think that there is an enormous gap and like um whether we are it doesn't matter when we are in San Francisco or Budapest or anywhere in between. Um I think that this is the most thing that folks don't get on the internet, that hey, I'm using this yet, many people use it for free, uh or p pay the twenty dollar bu uh thing and w motor we pay spend now forty K a month, forty K a month on the inference. On the free site. Alone. Yeah.

And like the gap of where we are and what it could be is scary because we are still a very small player. I would say we are going very fast, but we are a very small player and I can't possibly fathom. What those folks with billion dollar uh funding are doing, right? Yeah.

Because when we are helping accounting firms or marketing agencies or whatever to do sales and marketing faster, that's relatively straight forward. And I think that you know I was really proud of ourselves when we got the trademarks and I am very proud of ourselves when we got ScaleOS, that now we have our own methodology and I think since then we fuck up less in that regard, because like the bots now know and maybe that's also something to be mentioned that one is to I started to think as an engineer in many cases, that's why when I'm working with marketing counterpart sometimes I don't know what the fuck am I talking about there's a marketing There's a marketing agency CEO. He keeps shouting at me. What the fuck are you? I don't know. I don't know who you're talking about. I don't know.

Hey, hello, Tom. So it's like these kind of things that now I see it on myself when I'm reflecting that I learned so fucking much I cannot even explain it. I couldn't explain what I'm doing five years ago to my father. Now I'm not able to explain to my peers.

And I... And I wondered that, you know, when you know you were in hardware like at with your team, you know had uh flashy customers and you built a lot of thing like crowdfunding I thingy I told you uh early on uh How does it feel as an engineer because for me just you know playing code monkey or chaos monkey at your production systems he took away my production keys. I begged it back, Monster. But you know, you know jokes aside like how do you feel that you know, like where is this going like you have been doing this for what two decades now So what do you think?

I think um like reflecting back my job has also changed a lot. Uh you know at Airtame I grew the teams team from like being the only solo, you know, developer into having a team of like forty plus uh engineers working on the product in various teams. Uh and now I'm being back to pretty much solo developer except I have a huge A_I_ team that's doing the work for me. Um so it's it's been a very interesting journey on how I also have to kind of get used to and develop uh both my skill set on how I interact uh with various A_Is and how to kind of chord them into the right direction of building a company, right? And at the same time it's also changing the whole work flow, like how we as people interact and how we with machines interact. And I think that's one of the like uh the struggles sometimes we still have, that there is a misunderstanding between the machine and the human uh where the machine wants to be very um like uh structured, very technical, very like to the point um but misses the whole kind of background of why are we doing it and like how to actually tie things together and that goes back to your like uh 120 repo problem, right, that we have so many of the components of building a system and now we're like plumbing it together, but it's still hard to even comprehend as a human on what are the moving parts and how are they connected together. And the machine also sometimes struggles because we're so inexact with the like a prompting and uh how we like you know have parked downpals and Google Docs and so I think this is like a challenge and like also reflecting on what you said about people and where they are like I also talked with a bunch of my peers and some of them, you know, they treat A_I_ as a search engine, a glorified search engine.

Mm-hmm. They just put in like whatever like how do I find the best cloud hosting provider or whatever. Stack overflow. Yeah. Yeah, exactly. And um and I kind of like I think we're way ahead of that curve and I think the technology is still very early. So it's like I think our job also with is how to like bring that technology to people and explain it to them and so that they don't have to, you know, get into cloud code and invest spend twenty hours a day hacking away at it, but actually a way to give them uh like this power like make them superhuman, you could say, uh using A_I_ as a tool, as a technology to really bring them forward.

Well And that brings to my point because yes, we are plumbing now the new components because we are, I think we are iteration number seven now. At least. The big ones, the small ones I don't count, I can't count the small ones. But like you know, we were a tool, then we were a platform, we were multiple kinds of systems.

And like this is one of the conversations that we had recently that as an engineer, How do you know what's worth building? Because in San Francisco at least, I mea like I meant my mission for three months, I went to an event every single day in the city and I met hundreds and hundreds of engineers, some of them couldn't find jobs, some of them wanted to be a big start-upper. Um they simply couldn't find a they started to build stuff When the project after project and they were in five different projects because yes, they are superhuman they can build anything but like I think that we are kind of moving to the age of distribution and I think many of the conflicts that we are having is like coming from different philosophies because I don't care if it's hold hell together, duct tape, some pray, right, and faith, but it works and it works and it doesn't have to be necessarily scalable and even we have like big clients and they don't know what the fuck is happening, sorry about that, but like that's true, because not everything has to be so compliant, dealt, but like that's why it's just...

Just uh like even when companies like that got haggled to get up like the motor we're talking about like for me a and I don't want to talk about the big companies I really want to talk about the normal average people doing a business starting a start-up what the heck is worth building now? The thing you require right like um building has become a lot cheaper and a lot quicker. So think anyone can you know uh bring up Cloud Code and then like suddenly prompt uh their way to actually building something. Um the problem is like how do you get it out there. Who are your customers, what the pains you're solving for them. Like how do you really like uh have this kind of like user understanding, doing user research. Talking to people because like I have a billion ideas I've built like almost a billion of them but I don't know what you're talking about. But I also know that they like just by publishing it online you know it doesn't get people to even find it right and I think it's actually in that sense getting worse because like so many people can build so many things and So many of the same things, you know, it's easy to like, oh, I'm going to clone Airbnb and then build it overnight.

And then what, like, how do you actually get it out to people? How do you get people on board? How do you explain to them, like, what is it about? Why is it better? Or is it just the same? And then why?

Like. And I think that part is really hard, and especially as an engineer, you know, one thing is the building part, you kind of lose the interest also in some way if it's like too quick and too painless because as engineers we like to suffer, as I usually say like, you know, early pre-AI like era it was about like, you know, writing some lines of code by hand, then of course it didn't work.

And then fixing it and the fix didn't work and then fixing it again and that didn't work. By in two hours. Oh, and that's a semicolon not a normal now you know AI just like one shots it and it gets it right at least the build phase so that kind of struggle is no longer there now the struggle is more like distribution as you say and building the right thing finding the right market getting it out there and then like yeah if you grow it into a million user company You might have other struggles as well, that's more technology, but I think that's actually rare like it's it's usually the first part on how do I even get any users or Yes, customers.

And it's very funny that you say that to me because like it's like the marketer is talking about engineering Yeah, exactly, I'm just with cold. Yeah, yeah, yeah, so here's my title then there you go no but like because for me it's very interesting to see that all of these kind of things um and this is something that we fucked up as well that you know oh this is so nice it's so fat excuse me and we started like scoop creep now I understand it very intimately and I would like to issue publicly an apology to you that hey maybe it's not always we need to do that you know and When, you know, getting the basics, like it can get you 80% of the way very fast, but in order to make it sellable.

It requires the 100%. And we don't necessarily know what the 100% is. And it's like I quote myself in the loops, oh, I have this problem. I have my advisor bot, advisor bot's device, then I have the brief writer, okay.

And so the Grok writes a brief. In Grok, I put it in a perplexity max. I have all of the different kind of sources. I generate 10 different briefs. And then I put it in a Grok Gemini so that we have the conclusion. And then we have the... And Yep.

I'm doing these loops and I'm like I have so I have so much market data that I absolutely have no practical use for like you know yes the actual marketing agency market is now 164 billion who gives a fuck like you know like I think and this is what you said the other day that this kind of AI delusion or like the kind of fugue state is what really is the issue now and you know getting the fundamentals right like cash equals offer plus traffic boom Exactly. That's it. We can sell these glasses.

If you're listening, I'm holding a glass. We can go to the street and then we can pitch people. Hey, do you want to buy for this dollar? And you might may or may not sell it. And that's technically cash and that's technically an offer sold in business.

Like many businesses have offer, many businesses have activities and services, but they don't have offers and they don't have a way of thinking of how is this. This is going to be valuable and I think And we kind of rubbed off on each other in two years that hey now we intimately understand that now even when we have that kind of inkling of an idea to do something who the fuck will buy it we don't have any domain like and now we are building what fifth six domains a day like we don't have anything there we don't have the distribution there let's not do that and we are building now in clusters for customers I think this is something that more folks need to understand because maybe I'm looking at here the US It doesn't necessarily make sense that many of the businesses now being handed over in generation and there is no one to hand it over to.

So there are a lot of lot of businesses that are like lifestyle businesses in the more traditional sense that hey you can make a decent living but it's not going to be like the Lamborghinis and all of that sort so you can have that because to be really to do like a lovable type of thing that you know you're a bunch of 17 years. Or 17 year olds building something that you are technically able to build in 45 minutes that's not going to work unless you have these kind of fundamentals right you can garner people and rally people and for the boring parts like if you're selling accounting like lawyering and all of this sorts like yeah then there is a system and you just need to have the meetings and you need to grind it out so I wonder that what is it that you felt that like it was a fuck up that is worth sharing because for me that was it that hey we built a lot of things that we didn't even think who would buy it and then we try to kind of put it and then we got one response that was like very out of blue like I don't even work now I'm retired and I shut off the whole campaign like I don't want to have deal with this shit you know so that's also I saw like you I'm like a gen z like you know edwin and hunter. But like what is the one fuck up that you could share from engineering side that is like very more tangible that you know, maybe you should not have done that.

I think um I think that uh a concrete fuck up is like And this is also like augmented by AI in some way that it's very easy to over engineer things, you know, it's like very quick to turn a very simple project that you know, pre AI, I would just like, you know, make it very concrete, very small, very like something I can read from like in 15 minutes of just looking at the code to becoming this like huge, you know, project with a lot of databases and components and like you're like really like the AI is designing for scale for like Amazon type you know company when you don't even have a single user to use it and I've I think I fell into that as well with some of our projects where you know like I'm already thinking about this big scale instead of like really like scoping it very narrow and very sharp I think that's one of the other things the other one is of course like AI makes you feel like you're superhuman and all your ideas are amazing and you're never wrong I you're never wrong never yeah also like made my AI like be a lot more like yeah exactly more stern no stern and like throw back the stuff that the ideas that I'm throwing at it because I really need that kind of counterpart you know I'm saying like no this is a dumb idea like don't do it but we have like the one of the investors Axel like he has his own remarketer you know I made him in the skill and like you can see the pattern with him because based on the transcripts now every single time I ask him something the skill I mean not the person he says Ivan stop you are now a step free now let's go to stuff like literally he's German guy I love him and like you know he's doing that so we build this kind of guardrails to also keep us maybe it's not just the machines but keep us humans in line and check I think that's that's also like more important than ever that Yeah.

Hey know what the fuck you're doing right that you know because it's easy you know you open up but it was also like in our defense just to be on the both sides of the court even before ai you know guys opened up a shopify trial and then they have a t-shirt that no one bought and the instagram handle entrepreneur like so I think if you if somebody wants to delude themselves they are able to do so without yeah but it's much easier now right yeah you can kind of cheer later yeah amazing yes oh and like the practical usage is that you know what I was really pissed off Why are we not talking about now OpenClaw at all? Like you know, It's is it disappeared, right?

Came and went. Uh the okay, the guy uh you know, he's a very loving guy. But like I'm pretty sure that something was engineered about the type like uh you know uh that's my favourite conteo. Um and you know I was very surprised that hey in the podcast studio like we were talking to the guys and then uh I don't know how we got there but like yeah I didn't talk to you yet. But then when we had the emails, I'm like, no, it was actually my AI writing all of those emails.

And he mentioned that I wished him a great day. And then how nice this guy is. And I'm like, no, it was like it was trained on my email. So it knows that I'm polite, especially in first contacts, but like, otherwise, no. And, and like, I feel that all of these kind of courses and communities and all of this kind of bullshit, like Can you not buy anything for a month and just observe what are you doing?

What are email, what is something that you do more than once in a month that could be very easily automated. Yeah. Yeah, so and you don't even need to have like Mac and NA10 because that's what the gurus like to sell because they can sell templates and that's how the money is made. So folks don't be scammed. But like, you know, what is something that you started?

Started to automate for ourselves that you really like. I mean I think I like automate most of the things that I do and like try to focus on only the you know human aspects of the core like how to design this in a scalable architecture way but like for example infrastructure is like fully automated for me so anything from like buying domains to setting it up to having it on the server to making websites to like everything is pretty much just like one prompt and it's done fully end to end which is I think it has like made my time so much more productive I think if I have to put a number it's like at least 20x on what I'm capable of doing today in the same amount of time compared to before and all it took was tying up the systems to the right AI and making sure that I have my own guardrails so like it cannot just like oops delete production because it wanted to free up some free space free space for this brand new website or something like that. I don't know who we are talking about, but my favorite invention of yours is definitely the review bot that now it checks me, that doesn't let me merge anything until it checks all the diffs and the CI and then, hey, actually you have a core conflict and then, oopsie And I think daisy.

Actually that's one of the things we're doing well, like we're like scooping out various roles to the AI and then like specifying really like how they should operate with us and also together in order to actually execute properly with our business processes. For example like now I have my own developer team AI and then whenever it actually generates stuff my review AI actually reviews it and I'm getting more and more confident that the review AI is actually doing a pretty good job. I still sometimes look at it of course like validate that everything is in place. But most of the time I would say like 90% of the time I can just like give it the green check and I'm happy with it Which I think is already a big progress especially in the age of AI being able to generate tens of thousands of lines of code per hour Don't talk, don't get it out to me, but yes, yes, something like that.

And, you know, like especially like my contributions also that I think that because that's because I'm now able. Able to meaningfully translate my vision and you know I back in the day I was at pre-code camp I was going to boot camp I learned how to code with you know HTML CSS Python all of these things I never I never worked as a developer right so it was not really something that I could use and now that at least I understand the concepts and now you're giving me that those kind of things I think the original setup of like hustler and hacker I think that's right, super uh impactful that now we have that kind of habits that you know, I know where you are sometimes going over shot, shoot and scre Yeah. And you know me when I'm I'm in a different uh galaxy uh than this planet and uh I But just to jump in on that, I think it actually helped a lot on our collaboration as well because now you can go and prototype things and experiment with it and you start to also understand my struggle sometimes, right?

Right along with it. But it also gives you a way to communicate your ideas and your thoughts to me more than you know just writing it down in a Google Doc or something because you can actually ship a product yourself and then I can make it proper right and it's the same thing the other way that I can communicate back to you in a more technical way because you also like learned about it and Yeah. I think that's a great way to also you know enhance the collaboration between as you say the hacker or the growth hacker and the engineer or developer or startup whatever uh But that's something I could suggest for people, also for non-technical founders to start playing with it and experiment and that you don't have to push production code from day one.

But like hashtag yolo like dangerous is capable of it you can but yes but if it's local and you do it for yourself okay Then it's it not, doesn't then it probably matter maybe won't harm that much.

It might be yeah like maybe my cyber my yeah security background still like creeps out sometimes yeah like it's okay like I don't know we never leaked any kind of public no key no we have we have pretty good uh I yeah think guard you called rails you called him multiple times before I wanted that so that's good and um I don't want to create like you know multi-hour episode I'm not joe rogan but I want just uh I really want to tie up with one thing um is that you know I think it's time that will be a big thing that we have been working on and I've been tripping it on search in the live streams and you know I generously let you work instead of doing those with me but now I had I had to I had to force you now because the people wanted to see you that he that you're not an avatar you exist but maybe like who I am.

Knows who knows but no like I think that it's time that it we at least took a just a couple sentences about engine because like I think uh it's time that we talk about it um because that's the big thing that we are building uh towards and we have multiple surfaces for it um but I really would like you to showcase that you know how is now being built because we have been spending two years on development another year on ideation on it and maybe it's uh now you can do it better because marketer will tell you everything. Um so Yep.

What is Markster engine today? So the engine is um Yeah, maybe I can start with why are we even doing RAD. The idea is that we want to bring the technology that we have today to run businesses for people that are not like, you know, coding 24-7 in cloud code. Have So, lives. Exactly.

So the idea of the engine is to have Markster operate your business. Uh end to end. And that includes all the business processes that you might have in the company that requires marketing and sales, uh anything from generating leads to running cold make campaigns to doing uh social media and content and so on. And the engine itself is designed in a way to learn about you as a company and uh people within the company build up this kind of company um context in order to be able to act uh for you, right in some human in the loop way obviously. So but sometimes not if you don't if you don't want that. And to be able to actually help you through the whole flow of running your business uh by generating uh content, artefacts, uh actions, tasks to basically the whole flow um. And it helps basically by making sure that you don't have to think about, you know, how that uh engineer is running and you can just kind of outsource it through the technology and you can focus on what you're really good at uh and that might be accounting if you're an accounting agency or you don't like a developer if you're a developer, you don't have to learn about the intricacies of actually running something like this. And the uh Mom would might ask that hey I have Grok, I have ChatGP T, I can connect it to my Gmail, I can connect it to Google Drive. How is this different? What would you say?

It's true and you can and you should probably because I think it's a great way to play around but it's it's not something you can implement or you well technically you can because we actually open sourced our playbook on doing that. But as you also know, it's it's a lot harder uh than it seems. It seems very simple to just prompt your way into it and you can get by to some extent. But if you really wanna roll it out to your whole company, make sure that it's running all the time, make sure that it doesn't crash at uh two a.m. And that you don't have to spend six hours cleaning it up afterwards. Uh Never done that. Never. Then that's why we're here um to actually make it proper. Yes, and that's kind of the point that, you know, even as simple as uh creating social media content that was one of the first use cases even three years ago, we spent last week a full week on re-hauling that system to making sure that it's right every single time every single way and uh, you know, it generates and it's really you.

Exactly, I think that's the main point that it needs to be you and it doesn't need to it shouldn't be J_G_P_T_s. These general AI slop, right, that's now like spreading everywhere on the internet. Yes, and that's why you know you can see that you can get viral almost immediately, you can get leads almost immediately, because everybody's using this kind of generic and lazy solutions, like really, and that's why being different, being standing out in a good way is one thing, and secondly, that you don't have to hire big consultants, hire courses and etc.

Etc we have ScaleOS which is our trademark methodology I have worked my ass off to codify how you can actually grow a business through because I have you know burden I you know very well but some of them don't know that you know I burned my seven-year-old relationship and we broke up on our anniversary because I was in the business and I it consumed me because I was not able to step away for a day and even when I tried I just kept thinking of it and I was physically there mentally not there and like I think that's the super important thing that you know many people are struggling with life balance or actually the kind of pressure of your AI is taking away and it's going to be all we are going to Get the fuck out of here. Hmm. That's first of all. Uh second of all uh it can just make your life easier if you are actually using it to a template. So we worked with many different kind of companies and now I can think many different ways like from services to products to small and big, from Europe to North America and then beyond. Uh so I just wanted to thank you Attila to being and helping I think there was thousands of people who use Markster today because that's your magic hands. You are Jesus now.

And, you know, that we are now revealing Engine. So, and obviously details will follow. But I'm very, very proud of it. And I hope that this kind of fuck-ups that, hey guys, we also fuck up, like, that's Everyone does. That's the part of the learning journey. That's to learn from.

Learn from it, yeah. Yes so it's now not like you know five years in university like you still do university but it's not five years and then like it's much easier and faster you can get to a very good level if you are just trying things so Thank you for the g for pulling away from your all of your terminals and Thank you. My agents are already like yeah like a crisis. And um if uh any of your listeners and watchers you want a little more you need to tell us in the comments so that I can I can invite him onto the stream with me so you don't have to see me all the time so uh thank you very much and you have a great day.

Same to you. Bye guys.