Fireside Chat with Jozef Gherman (StealthGPT)

This transcript captures a fireside chat between Gary Sheng and Jozef Gherman, founder of StealthGPT, in front of Gauntlet AI graduates at Alpha Schools HQ.

Gary Sheng: Alright guys, this is an experiment. I think like, hey, good to see you. So look, everyone's building in the ed tech space, education space.

Jozef: I think it's really important that we get people that have earned spiky POVs. So, this is a friend, Joseph. You're here for how long?

At least three months, but with the way that's going, maybe longer. Okay, cool. And what inspired you to be in Austin?

Well, the tech scene over in Tampa was lackluster. So I just wanted to go somewhere that was more friendly to tech. I didn't want to make the hopscotch move to California just yet. And I heard that Austin was like a phenomenal place to go.

Gary Sheng: Yeah, Austin's interesting because it's free and it's Hi, everyone.

Jozef: So welcome to Austin. Thank you.

Gary Sheng: So, Joseph, you are the founder of StealthGPT. Do you want to talk a bit about what that is?

Jozef: Yeah, sure. So, StealthGPT is basically the world's first undetectable AI platform. So, what we did is we figured out how to take generative AI content from ChatGPT, Gemini, Grok, wherever you may find it, and convert that into kind of like an output text that these AI detectors couldn't pick up on.

So that was like our offering to people, and then we kind of just wanted to see where it would go. Like when I first built the technology, I didn't know who was going to need it, but I knew that people were going to need it, and so that was like the founding thesis. Now we have a bit better of an idea of who's using it right now, obviously. It started in February 23. So, it's been a couple of years since then, and we do have a customer profile right now.

Gary Sheng: Awesome. We're going to touch on a lot of things. I bet you've started on a lot of things that didn't work out. Do you want to talk about failure?

Jozef: Sure, I mean, well, if you Google me, my first company was kind of a failure. I launched it when I was 23. I was getting big into the crypto scene and all that kind of stuff. Got into crypto back around like 2017, rode that bull market, 2018 hit, you know, it was like 80%, 90% haircuts all the way around. Bitcoin fell from like $20k to like $4k, crushed my business, was a horrendous failure.

And then since then, I was kind of like meandering around, just trying different things, doing some advisory work, doing like other, you know, kind of jump-starting ideas, but nothing seemed to really connect until, you know, StealthGPT. So that was a good four or five years.

Gary Sheng: How'd that feel?

Jozef: Never give up, man. I mean, like, that's just like, at the end of the day, I've always kind of, you know, luckily, my father ingrained this into me, like, when I was very young, like, just believing in yourself and believing in your own capacity to make it.

Even when I was going through all this stuff, and I've gone through a lot of things during that time, very tough times, I never lost faith that I would actually make it. It never got so dark to where I was like, oh man, it's actually over. That never occurred to me. It was always like, I'm going to keep swinging, and eventually I'm going to hit a home run. It's just going to happen.

Gary Sheng: So you mentioned your father. Are there other people maybe around your age that were encouraging you during that time?

Jozef: Not really encouraging me, but what I did do again, like lessons from the father, which is really important, which is like kind of, you know, being cognizant of who you're friends with, you know, like making sure that the people that you associate with are like high effort people, people who are doing well in life, people who aren't making like very dumb decisions. So that was like impressive on me.

And so when I had recognized like a friend of mine, he's gonna be a doctor, this guy's gonna be a lawyer, this guy's gonna work in the government, like whatever, I just stuck to them like glue. I would just be like, I'm gonna, I'm gonna hang on to that connection, those friendships, because these people know what they're doing.

Having us investors in our company would not have been possible had I met these people prior to StealthGPT. So the network I was building up over time, I could later leverage into becoming investors into the company or introducing me to another person who became an investor in the company. And then as far as directing StealthGPT and figuring out things over there, having these experiences and having these failures definitely ultimately led to success.

You know, unless you're insanely lucky, you're not just gonna just beeline to success. You talk to anyone, they always talk about how many failures they had. That's just like, that's just part of the path. Like you gotta accept failure is gonna happen. You've got to accept the risk, you've got to take the gamble, but through each failure, it's not a failure at the end of the day if you learn something again.

There's always cliche sayings that do sound very fake and cheesy when you first hear it, like yeah, yeah, yeah, bedtime story type stuff, fairy tales, but there's a reason why they're cliches, and there's a reason why it's echoed over time, because they kind of are true, and you kind of just got to keep following along the path and hope that things kind of connect for you, because there is a certain role that luck does play into the whole process of becoming an entrepreneur.

Gary Sheng: What have you learned about the education system through doing self-DVD?

Jozef: Um, well, I, I mean, I, I will say I haven't learned anything too new because, you know, so not sure if you noticed, but, um, back when I was in school, I was like, I was a student body president over at USF. So, you know, I was very familiar with the school system, super familiar with tuition fees, the academic experience, what a student goes through and struggles with loans, you name it. I basically advocated for it and did work on it.

And one thing I knew about schools - the first thing I want to talk about is that they are resistant to change. They don't want to change anything. They don't adjust their curriculum, they don't adjust with the times. They're saying the way that school is being taught right now is very, very similar to the way school is being taught in the industrial age. The only difference is you've got computers and stuff like that.

You can't use a calculator, you can't use software, they have not innovated. You know, when I built StealthGPT, like I said to you guys in the beginning, I had no idea how it was going to be utilized. Now, later down the path, Turnitin, which is the biggest academic software in the world, they implemented an AI detector into their system.

So now every single day, when a student submits their homework, it's automatically checked for AI-generated content. That was done months after I created StealthGPT. Well, guess what? Now that these AI detectors are in schools, guess who needs stealth GPT to beat these AI detectors, right? So students need it, right? And so we started to see a lot of students start to use our software, and we kind of started gearing towards that demographic, because, hey, that's where we're getting the sales, that's where we're going to go, we're going to chase the buck.

But I kind of figured, and I knew, even at the time, it's like, we're beating these AI detectors, and this is back in February 2023. We knew that they weren't going to change, and to date, they still haven't changed, you know? They say it's a big problem, but they are resistant, and they're just anti-agile, you know what I mean?

Just because of the systems that are built up over there, they're almost... They're disincentivized from changing for the better to adjust to modern day technology because changing requires learning something new, doing something different. And, you know, personal growth, personal growth.

But for what it's worth, these teachers are lazy. You know, I had a finance professor that looked like any day he could pass. We could have gotten an email like, hey, sorry, professor's not coming in today. The funeral's next week. And the guy was handing out flyers in the class, like handing out this handkerchief. In this lecture, I'm going to talk to you about how to... It's insane that you're still giving out this paper from 20 years ago.

You expect students to change, you expect students to do everything in good faith, but you as a teacher, you're giving out the same stuff that you've done for 20, 30 years. Why don't you do something different? But the school system is anti-change. It's resistant to change, that's the incentive structure, that's the teacher unions and everything like that.

So even as people talk about the threat that ChadGBT has and things like CellGBT have to the... I mean, yeah. Again, when the school system adjusts to a modern day calculator and software, I'll actually start to be concerned, but that was a big point I brought up when I advocate on behalf of the students and stuff like that.

I went to a Florida state legislature, and I'd be like, why don't they give students, why don't they teach them modern day softwares? If you're a computer science major, again, I'm not too familiar with what the school system is doing right now, but why aren't they teaching you how to use Cursor or Windsurf? If you're a finance guy, why aren't they teaching you finance software?

Why aren't they teaching you accounting software? Because when you go out into the real world, you're not doing a, you're not doing a standard deviation by hand. You're using a software to do that, right? Software is not going to make the same mistake that a human error could do.

So, and when you're going out to computer science right now, if you don't know how to use cursor or windsurf or these IDE's that are connected with these AI LLMs and stuff like that, you're at a huge disadvantage. And it's like, why won't the schools do that? Well, that would mitigate and basically eliminate a lot of the grading. And now they've got to rethink everything. Now they've got to make something new. That's the problem. That's what it all comes down to. And they don't want to do that.

Gary Sheng: What is your vision for the ideal school system? Let's talk K-12.

Jozef: Well, luckily with AI and stuff like that, we have the ability to create a very atomized, unique, individual experience tailored to each student. Luckily, AI is not lazy. If you ask it to write a paper, it writes the paper pretty much.

Now, you might ask it to fix your code, and maybe it doesn't know how to do that right now, but it's getting there. But AI is very good when it comes to generative content. If I ask AI to draft up a curriculum, it can do that. If I tell it I have ADHD, it can also factor that in and develop a curriculum with that in mind.

And so, one of my ideas, I'm very passionate about education and stuff like that, stemming from my experience of being a student body president. It's a huge passion project. I think that we can use AI to create these unique learning experiences where you can make a tailored curriculum based on a goal, you know what I mean?

Like, say for example, you have a new kid, they're born, and you want them to be a doctor, you want them to be computer science, you want them to be whatever they are. Imagine you put your kid on a track, like you can basically tell the AI, hey, this is my kid, they're this age, this gender, this list of disabilities, and they have it or none. And then I want them to be on this track and then imagine creates this custom like K through 12 curriculum for that kid that can also be adjusted in real time because as we know when we're four or five years old, you know, some of us want to be an astronaut or fireman, but then we change our minds.

You know, we get to 12 and we're like, wow, I'm really good at drawing, or I'm really good at math, or I'm interested in the stock market, whatever. And now we want to change our path. Well, with AI, we can augment that path in real time, right? And so we can change the way that we're doing stuff.

But also, the powerful thing about AI is that we can create all sorts of learning experiences. How do you learn best? You learn from reading. Some people are great readers. Me, personally, I'm a visual learner. So I'd like to have more visual displays. More charts, more graphs, more like PowerPoint presentations, stuff like that. Do you like to learn through games? Do you like to learn through audio? Do you like to learn through video?

Whatever you may have it, we now have the capability to almost do everything I'm talking about. You've seen people create games with AI, vibe coding, like I'm sure a lot of guys... Maybe not. But if you're familiar with Levels on Twitter, you know, the guy created a game within a few weeks.

Gary Sheng: I think everyone's familiar with Levels IO.

Jozef: Yeah, exactly. I'm in good company over here, so you guys are familiar with him. He creates a game in a few weeks just with AI, not really much to it, and it's making like $150k a month. You know what I mean? That's the type of stuff that we can do with AI right now.

Imagine what we're going to be able to do with it in two to three years. It's only going to get much more powerful and stuff like that. As it stands right now, what I think the education system should go and can go is having the ultimate tutor, which is AI, which doesn't get frustrated with you, doesn't need a vacation, and knows far more than your teacher could ever know, have that teach you to be a better person.

There are some components in there because there's much more to school. Like for me, the college experience was not what I learned in the classroom. It's what I experienced outside the classroom for me as a person. And so that component needs to be solved. And that can't be solved by, you know, looking at your phone or tablet or computer, which is the brick and mortar, being in a physical space, being around people, like minded people or whatever.

And that's that, in my opinion, is the toughest component to solve. The educational component. The educational component can be solved like that overnight with AI. But then it's like, how do we make sure that these people aren't socially awkward humans for the rest of their lives, right? We don't want that to happen.

Gary Sheng: I mean, especially if you, you know, Joe talks all about learning science. If you basically feed it the learning science that you think is relevant, you say, hey, I create the curriculum, like, theoretically, over time, you can just instantly do the exact kind of curriculum that you want that fits that science. Right. Yeah. So you talk about creating good humans. What is a good human?

Jozef: That is like a very much like beauty is in the eye of the beholder type thing. It is very subjective. I do think in those very great terms, like George R.R. Martin kind of colored that for me. It's like, what is a villain? A villain is a hero of the other side. It's like, what is good, what is bad?

Well, maybe what he's doing is good, and it's benefiting someone in a good way, but it's bad for me, right? So maybe I interpret that as bad, and that person interprets it as good. You know, I don't know. I mean, at the end of the day, you kind of got to use, you know, what I would say is situational ethics, you know?

Sometimes, like, you know, we're taught all these platitudes, and we're taught all these, like, You know, kind of things like, you know, through hard work, I'll always make it. Well, not necessarily, you know, maybe the guy who's a little bit more cozy with Ross. Or like, be humble, which is the exact opposite advice that you have that you want if you're doing marketing.

Gary Sheng: Yeah, exactly. Yeah, definitely. You can't like if you go into a VC meeting and you're like very humble and you're very conservative. You're not getting funding. You're only getting funding if you're like, my platform is going to change the world. And in three years, no one's even going to use chat GPT. They're going to throw that in the trash because they're all going to be using stealth. And if you don't say that, you're not getting the investment.

Jozef: And then you say that, and then three years, it's like, ah, hey, man, yeah, we got close up shop. Yeah, we're gonna win. And now you're in the A-hole because you lied, you know? So it's almost like setting you up for failure.

But to that point, like on the humbleness, it's like selfishness. I think that's a really good example where people always say, don't be selfish, don't be selfish, be selfless. And I think that there's a place for both. There's a time and a place where you have to be selfish, and it's a good thing, and it's fine to be that way.

But there's also times where, for the most part, you should aim to be selfless and stuff like that. But don't be afraid to do things out of your own self-interest. Because trust me, other people are out there, they're working on their own self-interest. They're looking out for themselves. And if you, at the end of the day, are not looking out for yourselves, and all you're doing is focusing on looking out for the other, You will get screwed.

I've learned that lesson time and time again by being too generous, by being too nice, by not being pushy and letting myself be a doormat. And then just people step on you at that point. Know your value, know your worth. Sometimes you got to look out for yourself and do some hard shit. So I want to point that one out because you point out humble, but selfishness, I think is a very controversial one as well.

Gary Sheng: That's a good, what we call a spiky POV. Let's get—we're going to have fun. We're going to zoom in and out on different topics. So, one thing I want to—I want you to get tactical. What should everyone building at TechSass know?

Jozef: Well, you should know your audience, right? You should know, like, 18 to 24 college, but if you're building something really cool that any student could use, as far as 15 to 16 could be your target. That, for us, is in our target demo.

And keep in mind, the attention span of a person is really, really short. I mean, it's like, you guys have heard it over and over again, but again, you've got to really pay close attention to that. If you're not talking brain rot, you're losing people already. If you're not talking Skippity Toilet, if you're not talking in their language, and you're totally talking in these very high, almost like SAT-type words or whatever, as developers, sometimes we get a little bit too technical for people, where it's like, what was that?

If you're talking in that language, you've already lost the plot. So when it comes to doing EdTech SaaS, if that is your target, if you're going after students, You know, know the target, you know, talk to your siblings and stuff like that. If you have cousins, talk to your cousins, talk to people who are in school, you know, and get a feel for the landscape, get a feel for what the problems are.

If you're trying to solve a problem, know what that problem is, know what you're trying to solve, and then talk to those students, because that's like your focus group, and that's one that you don't have to pay for. Now, obviously, if you're targeting admin or you're targeting teachers, if you're building like tools for them, You know, obviously you don't need to be talking skippity toilet with those type of people.

I'd probably freak them out. They'd probably be like, what the hell are you talking about, man? So you want to, you want to make sure that the speech that you're using is very, very tailored to the type of audience that you got in front of you. That, that, you know, code switching, I think is what they call it.

But like that ability to just switch up your language and your marketing speak is going to be super important because I'll tell you this too, you can build the best product in the world. You've heard it a number of times. You guys are all engineers, coders, you guys are doing it right now. You can build the best product in the world.

If you don't have marketing and distribution solved, you are dead on arrival. It doesn't matter how good your website is, it doesn't matter how fast a page loads, it doesn't matter how fast a user gets an output. If you cannot get distribution, you can't get eyeballs on your website, you're dead. That's it. Point one.

That was our biggest thing that we had to discover, because we are a very product-focused lab. And our exponential growth, like when we went from 40k a month to, 40k what? 40k MRR a month. And then our biggest spike from that point was like 150k. We're at 330k right now.

But when we made that big spike from 140 to 150 was that realization that product led growth, feature led growth was not going to give us, it just wasn't going to grow us. What was going to grow us was working with the influencer, finding out a good way to market the product and get it in front of the target demo.

At that point, it's just a numbers game now. It's just a lot of big numbers. I have a 3.5% conversion rate that holds steady. All I have to do now is 10x the eyeballs. I 10x the eyeballs, I 10x my revenue. It's that simple.

One of my friends right now, he's actually my COO, good friend of mine, we met in college, he was more on the marketing side. His whole thing was he worked for high-ticket sales, or high-ticket SaaS sales. And when I first built StealthGBT, we had all these free uses built into it. And we met one night, I showed him the website, and he's like, so you have 10 free uses on StealthGBT. And you're not getting that much customers. We were probably at like $500 MR. It was like really beginning stages.

And he was like, how many of these people are used, like of those 10 users, how many people, how much are people using on average? And I was like, well, I don't know. I'm not tracking that. He was like, well, you should really track that. I was like, okay, so I track it.

And then we see, okay, on average, two people, like people use two uses, and then they just don't use anymore. Okay, so change it from 10 uses to 2 uses, that's it. All they get is 2 uses, and after that they have to pay for it. Just by doing that one change alone, we went from $500 MRR to about $4,000, $5,000, right off the gate. Just that one little change, one little tweak, allowed us to grow 10x from that point.

(transcript continues with discussions on finding good people, creating a strong team culture, and philosophical thoughts on success and faith)

Gary Sheng: You're an amazing communicator. What's the importance of that in building a successful business?

Jozef: I mean, to me, like, I'm super authentic. I did have to catch myself cursing a few times because I'm so used to not cursing when I'm talking in front of an audience and stuff like that. But for the most part, I speak from the heart and emotions on the sleeve type thing where it's like, I am an open book.

I've had genuine conversations with my own competitors. Where I see people who are running websites that are literally doing what I'm doing, or copycats, I reach out to them. I'm like, hey, I'm Joe, I'm super competitive, I'm going to try to beat your ass, but at the end of the day, I do think that the pie is big enough for all of us to win, so if you need to call me or have any questions for me, go ahead and do it.

I'm not going to give you my source code, I'm not going to do that, but if you need some general advice and stuff like that, I have an open door policy with just about anyone. You know, that's just kind of how I am. I'm very free flow, and I don't care who you are or what you are. If you're a celebrity, if you are a billionaire, or you're just a dude who started a startup and it flunked, it failed, I still keep in contact with people like that, you know? I don't have a gatekeeping mentality.

(transcript concludes with additional perspectives on leadership, faith, and maintaining authenticity)

Gary Sheng: Hopefully I did some motivation.

Jozef: Hope this experiment was good. Hope to bring more people like him around. Also, I'm really chill, so like, I don't take myself too seriously, so. I gotta go, but like, I'd love to come back and chill, and if you guys see me around the office and you ever want to just pull me aside, feel free. I'll bring you around. Yeah. Alright, thank you guys.

Gary Sheng: Have a good one. See ya. Bye.