Alex Hormozi’s New Playbook: Entrepreneurship in the Age of AI

I'm gonna give you five different things to do today, but the one that could triple your business is going to take five seconds. >> I wonder if we can put Chromosi bot in a funnel. See, that would be a very valuable thing if you had a bot that could crawl a funnel >> and then make suggestions. It takes a picture of each page. >> Yeah, exactly. >> And then just doodles. Yeah. And then just says like, I think you have an offer issue here. And it could crawl your ads library page and the hooks are off here or the hook isn't congruent with the headline. There's all these different check marks that you have to go through. I wonder if we collaborate with you to have like a funnel scan. >> Yeah, that'd be cool. It' be an incredible lead magnet. >> Yeah, >> there's so much content out there about entrepreneurship and that's often good. But often people think that they need to fit a certain pattern and they miss the things that are around them. Family members that can help them. Uh co-founders that might not fit the bill of a >> of of a co-founder you might expect. Yeah. >> Um and um >> well, it's usually the weird ones that win. >> It is it is it is a it is a tails game, right? Like it is not an average. This is not the average game. >> Um and so what I like to tell founders and entrepreneurs is that there are no rules in this game. >> There are no rules. >> You know, my uh hold on, I'll show you something. This is the first book. >> There are no rules. Amazing. Amazing. >> It's my only guiding principle for the book. There are no rules. >> By the way, I loved about this this book when you start with here are some things that people said about Alex. That cracked me up. >> Okay. >> I I like how even when you started I I I like I started watching your videos, you know, when you first came out. >> Oh, no way. >> Um >> Well, thank you. >> Yeah, of course. And No, thank you. I mean, I learned a lot and um >> you know what attracted me to it is like, yeah, I don't have anything to I'm not trying to sell you anything. you know, I'm just talking and it's clear you have >> this demeanor about you that is kind of relaxed and I don't know where where that comes from because a lot of fanders are just sort of so highrung and honestly I think it's just I have a worldview that most people don't share and I think it bothers a lot of people and so I don't talk about it as much because I just um but when I do talk about it like I always like I get the most meaningful messages from talking about it which is just like I believe in cosmic irrelevance >> um and so you know like the idea that like someone is is famous or not famous or has status or doesn't have status. I I just so hard whole wholeheartedly believe that in in four generations like I will be completely forgotten and no one will care. >> And I think that just like >> it just takes a lot of the pizzazz out of out of like >> special snowflake, you know, isism. >> Yeah. >> Um and it takes a lot of the anxiety out of the day-to-day operations of a business. The upside and the downside. And I would say that I think a lot of people, this is gonna be a little bit of a side quest, but I'll wrap it around. So like >> I think many people desire emotional or mental stability and I think it's absolutely a prerequisite for being a good entrepreneur because there's just so many fires that come every day, right? And so I think people have difficulty with that because they don't define what it really means. It seems like a very amorphous thing like emotional stability or mental stability like what does that mean from a behavior perspective? And so I see that there's three components to it. So you have um you have uh mental toughness, you've got mental resilience, and then you've got mental fortitude. And I'll define each of them. So toughness I see as your fuse. How many bad things can occur in a row without you changing your behavior? That would be toughness. >> The next is um I'll go for two seconds probably better. Uh fortitude is um once once you surpass your fuse of bad things that can happen before your behavior changes, how steep is that curve of your behavior change? Is it a little drop or is it a you go ballistic and need to go you know do heroin for five months you know what I mean like I don't know just like something completely crazy right >> um so that's second element of it and then the third element would be resilience which is like okay how quickly do you bounce back >> um so what is the time between uh baseline behavior change in behavior and then recurrent to the original baseline >> and so if you think about those three vectors it's to me it's it's helpful so that I can think okay do I need to exhibit bit more mental toughness right now as in like cuz the ideal scene is that just >> you have maxed mental toughness and so that nothing changes your behavior but we are human we are fallible and things do happen right and so if I am going to you know have my behavior change then I want to think okay I want it to be as as little of a steepness as possible um in terms of my change as little intensity and then once I do notice that there's a change then I need to execute resilience and say okay well how quickly can I return to baseline and so what's interesting about that is that all of those are behaviors rather and moods or emotions around it because you can say, "All right, well, I can still be upset about this thing, but if my behavior returns to baseline, then I can resume original function, which is going to disrupt minimally the system that I'm operating within or that I'm a key key role in in making, you know, work." And so, um, that's my little side quest. Coming back to that, I think what allows any founder is having some worldview. It doesn't have to be mine by any means, but like I mean I've seen super religious people who like they're like, "Well, all of this is happening because God wants this to happen and this is happening for me, not to me." Great. That's just as strong maybe stronger of a worldview. Um but I think you should have some overarching um >> to couch the experience. >> Yeah. schema. Yeah. Exactly. That that everything operates within. And I think that makes it significantly easier because I think if you lack that >> um you become the center of your universe. you this is almost like a victim aspect of it. >> Yeah. I can't believe this is happening to me. Like all of that kind of stuff. It's not fair. This other company's doing better. They got this better funding or whatever it is. Um and it's just like it just doesn't matter. >> And in in either of those worldviews, you know, like if you have the cosmic relevance worldview, it's like, well, none of us matter. Uh and if you have the other one, it's like well the endgame was never the achievement, but I am the I am the goal. And if I am the goal, then all of these things happen through me, not to me. And so the idea is how can I use these things to my advantage to get if you have a a god worldview what the creator wants me to learn from them. But both of these things can either make it a null point or make you better. >> And I see that as at least I think the worldview of more successful founders. >> Fascinating. >> Yeah. Um that's that's a really interesting model of uh how how to think about I mean just hearing you talk I just see it in myself like there there are breaking points I get to where my behavior turns into like sort of more addictive type behavior. Maybe I'll spend like too much time on Twitter or >> or and I like catch myself like what am I doing? Clearly I'm there's an escapism aspect of there and and then how can you and over time the resilience aspect started >> um started to get better but you need to to focus on it. Uh and and this this idea of cosmic relevance is is really interesting as a as a philosophy because I think people see it as a some people see it as a sort of like a form of depression or something like that. But it but I think it is when you read the classics >> there's always you know there's the sort of the religious worldview there's the um cynical worldview >> and not in a negative sense and there's also the niic worldview and not in a negative sense I think uh when you >> our our age today like people don't read philosophy because it just feels like it's useless in some sense it feels like >> you you know, Cisophian task of like just like carrying the rock. You know, I spent like 2 years in San Francisco. There's this like really crappy coffee shop that would go to every Saturday and we would read a philosophy paper and we would just like discuss it endlessly. >> And it it just h just understanding how people thought in the past have >> have been hugely valuable because because I I can also pick these mental models or ideas from different >> different you know schools of thought. And the other thing and and maybe this is where we get into the 100 million uh money models. Um it's so you make it sound so simple. Uh in some sense it is simple. It's sort of like like Warren Buffett talks about investing is about buying a business and holding it, you know, buying successful business. >> Pick right. >> Pick right and pick things that are going to be here in 100 years. You >> and pick the right price and and and and that's it. um in in your case uh acquire customers that could pay for themselves in 30 days, right? The idea of an unckillable business that actually is is no one in Silicon Valley thinks that way. >> Yeah. In Silicon Valley, there's this aspect of like burn as much money as you can, capture as m much market share, then at some point ruck pull every uh and and and and actually try to focus on fundamentals because because the public market, you know, we've seen this with Uber and and other cases like that. >> Um >> I love the fundamentals are just making money, but Yeah. >> Yeah. Exactly. Yeah. >> So So may maybe uh give me your your thoughts, reflections on that. What what made you go the your route the the the sort of the more tried and true versus the Silicon Valley cuz I'm sure you know Alex you know 25 year old Alex could could have easily gone to San Francisco and like went that route. >> Well, so a couple things. So one is I I never once in my whole younger life considered tech as a as a as a career path. >> No one I knew was in it. >> No one that I went to school with did it. Mhm. >> So like I only discovered it far later in my career that it was even like on the menu of options, right? And so you know for me coming out of college it was just I I I didn't have this big drive to like change the world or make some product. It was just like I would like to not be poor. >> That would be great. And so that was kind of my my judge was like get a job that pays decently well. Um and so I did that and then obviously you know fast forward here I am now in more traditional business because that's what I that's what I kind of got into. Um, but I do think that there's kind of two equal opposite perspectives on taking market share and I think Silicon Valley focuses on one of them and I think there's nothing wrong with that. I think it's probably the better way, but there's there's there's an equal opposite way of doing it. So, one is that you drive CAC as low as as as down low as you possibly can. >> Um, which I would imagine is more the Silicon Valley viral product way, right? >> The other way is that you drive LTV so high that you can outspend everyone else in the marketplace. And so, but it's still cact like it's still the same dynamic and you one of them you just want to go to either infinitely low as it approaches zero or infinitely high that you can outspend everyone. Both of them fundamentally accomplish the same thing. It's just that most people you know you can do this so much so much more from a product I guess you can do both from a product perspective. Um but there's typically more marketing and sales that goes into the side and um >> I think it also depends on the scale that you're looking for. like at a certain point paid acquisition can get increasingly costly and inefficient >> and you get rock pulled by the Facebooks of the world often and all that >> right and they change the algorithm when they know that you're making money and they all of a sudden they just charge you twice as much for you know impressions etc etc but like um >> it still worked very well because again I do 80% you know of businesses in the United States are or 78% are service based businesses so it's interesting because like the a huge percentage of market share is Silicon Valley but from a logo perspect perspective, it's astonishingly few comparatively to the number of businesses that exist on basic fundamentals of charge more than it costs you and do it over and over again, right? Um, and so I've lived more in that world of just always fundamentals. And so to the same degree I um bootstrap I've bootstrapped every business that I have. Um, and I think that my my end goal has always been freedom um and maximum flexibility of options or maximum optionality. Um, and I think as a result that's shifted some of my decisions because I think to a degree like um, do you guys have do you guys have VCs who back your stuff? >> Yeah. Yeah, we have a lot of VCs. Yeah. >> Yeah. Um, and so if you you have a bunch of VCs like there's there's covenants, there's controls, there's voting, there's all this stuff. And um, not to say that maybe someday in the future I won't have that, but like that all felt heavy to me. And if if I believe what I say I believe, which is that in five generations it won't matter, then I'd prefer to ride this ride the way I want to ride it. Right. >> And and Okay. So, uh c can you introduce the the 100 million money model? >> Uh and um and you know it it it looks very simple. Why aren't more people doing it? >> I think it takes skill. I mean, I think it just takes almost as much skill as probably less skill than building a product that goes viral, but it takes some skill. >> Um, and so fundamentally, I call it client finance acquisition, but basically is that one customer comes embedded within it enough gross profit to pay for that customer the cost of delivering to that customer. So, CAC plus COGS and CAC plus COGS on the next customer. If you can accomplish that, then and within a 30-day cycle, then almost all businesses have interest free cash available to them on a 30-day timeline, which is why I picked 30 days. Now, if you can get interest free cash on a 90-day timeline or a 60-day timeline, um then you could expand that same concept. Um >> but the idea is that at that point, cash is no longer constraint to growth. You'll still have other constraints for sure. >> Um but at least cash won't be the constraint to growth. So, how can you grow a bootstrap business like it's venture-backed with that kind of aggressive growth? >> You do client finance acquisition, which is accomplished through a money model, which is what I wrote the book about. >> Um, and so like, you know, we had a software company that we went from zero to like almost $2 million a month in six months. And that wasn't because we had a viral product, but because we knew we could outspend everybody, and that's what we did, right? Um, and each of these businesses and like my gyms, I was able to fill up >> all my gyms before I opened the gym. And that's not something that's is common, you know, common place to do. And I would do those all in 30-day launches. And so that idea of well, if we can just get the customers to pay for the next customer, then I just need to make that first sale. And then after that, I can spin the wheel as fast as fast as my operational capacity can can handle. >> It's a flywheel. >> Yeah. And then at that point, the only thing that's going to happen is, you know, CAC's going to potentially go up. Um, but if you have, you know, tremendous LTV, then you're pretty much good to go still even in that in that shorter timeline. And so all the stuff that I talk in here is basically 15 mechanisms that I use to try and pull cash forward uh with the customer so that we can accelerate the cash conversion cycle so that we can ultimately remove cash as a limitation for growth which I think from a Silicon Valley you know uh VC backed perspective even on the low side if you already have VCs if you no longer if you eliminate your burn rate then you gain tremendous leverage of course right and then also you can time your rounds so that like maybe maybe this is a bit of a building year and you don't you don't want to you know you could grow faster but you know it's not in the best interest of the business right now. And so it's like okay we're going to have a slightly lower growth year because we want to fix some of the fundamentals of the product. But if you are burning cash then you're you know you could potentially get forced to raise a down round and if you get raise a down round you're screwed and all the momentum's gone and it sucks right and so having these types of skills or tools in your back pocket I think can be tremendously valuable for a founder who's obviously bootstrapped because that's how you can grow really aggressively but even if you are somebody who has you know venture backing it can get you out of sticky situations. >> Yeah. So I um I maintained full control of the company. It's not like I don't trust the VCs but >> uh every round I was able to dictate the terms partly because every round I raised it with you know boatload of money still in the bank. >> Yeah. >> Uh we're still like a money burning business. Uh but we've been fairly capital efficient. Uh, and I think I think that's like the main that's the main leverage like you you be again going back to how you want to spend your day, how you want to build your business, that kind of freedom. >> Um, it's very easy to build a jail for yourself. >> Yeah. >> Uh, and astonishingly easy. >> It's very easy to to be way unhappier as an entrepreneur than you were as an employee. Um there's a sense of freedom as an employee where you know you clock out at 5 and you don't have to worry about anything but you can build a pretty miserable jail for yourself uh as a as an entrepreneur. >> Um what kind of businesses have you seen built with this with this model? Like you give me a story of someone that >> applied the recipe and was able to to to to to grow a business. >> Oh, I mean there's like tons I mean any category you can think of that's like a traditional service business can can use this. Yeah. >> Plumbers, HVAC, gyms, B2B consulting agencies, like I mean it it works. >> Yeah. Yeah. But do you have a story of of you know someone um like maybe give a like a personal story like how did they get started or >> two kids wanted to start an agency for float tanks which is just like a random niche to be in >> and they read my gym launch book where I talk about this concept not as elegantly as I do in this book but I was like hey we run these you know six week challenges by doing that and they were entering a flow tank space where people were trying to get people on um like $49 or $99 a month memberships but the cost of acquisition was closer to $200. plus. >> And so they would have to wait >> two or three months to break even with gross profit in order to just recoup payment, which was really difficult for flowch owners that already had expended a ton of cash to open these open up these facilities. >> And so they came in just with a different offer, which was like a six week stress release thing. And so they, you know, bundled in um some mindset app and some like journaling and like a once a week touch base with somebody who just text them. And by doing that, they were able to charge $600 on the first transaction. >> So that's the attraction part, >> right? That was an attraction offer. And so they were able to do uh they did a win your money back offer but not off of like you know losing weight or something. They just did it based on behavior. So if you do these you know you attend twice a week for at least this period of time and you rate your stress levels but between the beginning end if your stress levels don't go down by at least two out of 10 >> um by the end we would give you your money back right and that was the offer. A lot of people like oh that's pretty compelling. Um and as a result they were able to you know offset acquisition costs and and the model fundamentally worked. And so that's so that's like an like there's hundreds like there's a zillion examples of this. If you were a dentist, it' be the same thing. You do Invisalines um and it cost you a X to get an Invisalign. Person comes in, you make three times your money and then after that you get them onto the subscription, which for them is going to be cleanings and whitening, you know, on ongoing basis. And so almost every business can structure it this way. It's just it's interesting because most businesses want to have a low barrier to offer thing, which I think is really good when you're talking about what the very first thing that someone touches in your business from an advertising perspective, but not necessarily the first meaningful transaction. >> Yep. Um, and so I I come from the perspective that someone is taking action because they have some level of deprivation or some level of problem that they're trying to solve in that moment. And so I think people have very large amounts of motivation for very short periods of time. And so when they have these large, you know, moments of motivation, we would like to capitalize on that with the largest transaction possible rather than the smallest transaction possible. Um, and in so doing, you know, larger commitment from the customer, you know, and if you structure some of the mechanisms that I talk about, it's like you can actually get them to be more likely to activate. Um, even if you have a B2B software that's, you know, higher touch and if they integrate their, you know, their their tech stack or whatever it is, um, you can put rebates in there so that they pay $5,000, but then they get $1,000 back if they do step one, two, three, and then all of a sudden, you know, that 5x is LTV on the back end, but you still liquidate the acquisition cost. And so, >> um, that's those are those are the variables, right, that we like to play with. And different businesses lend themselves to different things in terms of which structures like buy X, get Y free, uh, works in some businesses and less so in others. you know, a decoy offer or pay less now pay more later offer or when your money back offer or giveaway offer. Like there's all these different offer structures that work phenomenally well um to pull cash forward and also generate a lot of leads which is kind of the best of everything. You know, uh the the sort of internet business mantra has always been low friction. And yes, there's an aspect of it that's very true. Low friction wins in many ways, but there's an aspect of it like we've had many experiences at Replet where we we find that higher friction leads to to more conversion because you're asking more of the customer upfront that leads to uh more investment and maybe a song cost feeling, but also it might set them up for for success by putting in the work. This is there's a great hypothetical extreme that you can take when you're like trying to explain this to someone, which is like, okay, well, if less friction were truly the answer, then we should just have ads that just are just like buy here. >> Mhm. >> And we know that that typically doesn't work. >> Yes. >> Right. And so if that's if that's not true or if it's just like call this number right now and buy our thing for this amount of money like that, we we know that doesn't work because we don't see that very happen very often unless it's super lowc cost consumer products, right? Um, and so then it means that there's some level of friction that literally makes it more efficient to spend the money by adding friction. And so then the question, at least for me when I think about marketing in general, is what is the sweet spot? And so that's what I think all marketers are trying to find is like you want to find just enough friction that you can have as many people exposed to the ideal path to getting them to, you know, to ascend or to activate. >> Yeah, I love that. Um, uh, reductio at absurdum, right? like like find the like the most absurd case and that helps you prove the point. >> Um uh and and then uh the other parts of the funnel, the upsell, the downell and the continuity are all very important. One thing that we haven't done very well is the downell >> and like after after reading about that I'm I'm just gonna go back to office and just think about that. um like you know there's so many dark patterns on on churn >> where like you hide the you know I remember I subscribed to the uh economist uh newspaper uh and I had to call a number in England to cancel my subscription. >> Yeah. During their work hours. >> Yeah. During their work hours I was like I'm not I'm never going to recommend your magazine to anyone. Um but but I think instead this idea of like hey well okay you don't get the best value out of the service. Well how about we we you know give you um you know we we give you a plan that perhaps suits you better. Stick around longer. See if it works for you. >> Um >> something you might find really interesting. So we found this out when we did a big data analysis at gym launch which obviously B2B and higher ticket but I still think with psychology works the same way. Um, so we looked at all the LTV of all different customer segments across different service categories that we had. And the second highest LTV, you know, the first highest was the most expensive thing of the best avatar. I'm sure that was number one. But number two was the one that was super surprising to me. It was very close to that one and it was a significantly lower price point, but the churn was almost nothing. >> And what it was was customers who had been proactively downsold. So, not somebody who's going to cancel and then we say at that point, hey, let's, you know, let's downell you because you're about to leave, but instead someone that you saw that they uh were not using the complete feature set that there was a different service category that matched their exact usage case that cost less. And so, our team reaching out and saying, "Hey, you're not using this element of the service. We have another service category or tier that literally is 100% of what you're currently doing. If you want, I can go ahead and switch you to that." people were so like, "Oh my god, this is amazing." >> Yeah. It generates such goodwill. >> Yeah. And then they just stuck. They just stuck. And I found that um I I always kind of like remember that like we always have this fear as business owners, you know, to like I don't want to just don't bother them. They're paying us, right? But like you can >> pure margin. But sometimes it's like but you but if you can just crush churn and and triple LTV with that move, then like go for it. And also I think there's some element of ethics in there that how much word of mouth comes from that, how much good sentiment, reviews, all that kind of stuff. Less measurable. Um, but at least we could see LTV on those customers was the cohort that had the second highest LTV and that was like and it was at like less than half the price. Yeah. >> So to me I was like that's really compelling and interesting. >> Yeah, that's fascinating. Uh, I could think of an example for for Replet where you know some people build something, they deploy it, they're getting value out of the deployment, but they're paying the $25 and maybe that's a month and maybe they're not building every month. >> And so like downselling to say, hey, keep your deployment, maintain your tool that you built, but like you don't have to continue paying for the credits for for the agent or the AI that you're not using. And I I wonder because even the psychology because like I mean I love the psychology behind this stuff but like >> even making the offer that they could get some let's say it's $15 or $10 a month whatever >> and then if someone says no I'd rather still have the optionality >> it's almost like a reaffirmation that they're not going to like I would bet you even having to make the decision that they don't want the downell would make their current membership stick longer. >> Fascinating. I'm going to do that and report back to you. See what happens. That's very cool. Um, so let's let's get into AI a little bit. Um, >> asking you the questions. I don't know. >> So, actually, um, not many people remember, but you were one of the first people to key in on Chad GPT as a as a tool for business. >> Yeah. >> And I remember some of your early videos. >> It was like three three almost two three years ago. Yeah. It was a lot. >> Yeah. It was like a couple weeks after Chad was launched, right? Uh, in like um >> was it November 22 or something like that? December 22. >> Yeah. Um, so uh and I was very interested because I I always watch your world and I watch Silicon Valley world and I see like what is hopping over and when I saw that I was like okay this is going to be huge. Uh um so so what what what like what sparked the interest for for chat? What what what vision did it create for you? >> I mean I I think that there's there's fun questions of like what is it? So, I'm such a such a like what do I geek out on like what do I really enjoy a lot? Um is studying learning and behavior and you can hear it through all my my definitions and my business books. Like that's the through line that that marries everything. Um and so then then I mean AI obviously begs the larger question of like what does it mean to be human? And I think and I have this philosophical slant to my own interest and I think that's probably why I was drawn to it disproportionately. I also am a writer and so language is a natural thing for me. And so the kind of the the confluence of multiple these things together u made chat GPT really interesting. Um but just in like we we as humans tend to be very romantic about our ourselves and thinking that like no one can do what we can do and like we've been proven time and again that we're we're not as special as we think we are. Um >> that fits in your also irrelevant. >> Yeah. So totally in my worldview. Um a lot of people very upset by that. But with with each thing that it proves it can do because if if artificial intelligence learns almost exactly the same way that humans do through reinforcement training. And so you have you do a thing and you get an outcome. Yay or nay. And I believe at the most foundational level when you ask the question like why why are you so driven or why are founders so driven? We have gone through some reinforcement training which we may have been aware or not aware of that reinforce this set of behaviors that when you know stacked together so you see uh a skill as a behavior chain of multiple adaptive skills that are put in sequence uh to create an outcome that's maybe you know ideal or >> that adaptive chain becomes more complex set of >> uh behaviors then becomes a skill sorry um >> and so to the same degree AI basically learns the same way humans do >> and so if it learns the same way humans do then it will be able to do what humans do Okay, that's the very first principles kind of thinking especially watching Tachik. It kind of sucked back then but like being able to key on that. it just keeps tweaking and so um >> it it's it obviously starts with with just you know language but then it just continues to move move out and I'm curious what you think about like Tesla as because you know it trying to us trying to clean data sets to train on larger and larger you know L you know big data that it's that it's working off of um obviously that will have limits and then creating high quality data will become the constraint of the learning and so then it it will have to learn from first principles itself which is going to the experimentation in the world which means that it has to have some sort of link to the real physical world which is exactly how we learn right and so we're able like we have a so everyone who feels like you know AI is not as smart as humans it's like well because we've had a learning advantage yes >> because we take 10,000 inputs a day or whatever the number is% >> um if we take all of our senses times you know hours and whatever we can take in um as soon as we have robo babies and I say that not a like not a weird way but like um where it can experiment and touch and taste and do all that stuff in the world I think the reinforcement loops will happen so quickly and the other part of it that will happen is that it doesn't forget. And I think that's the that's the crazy part. So we've you know we've obviously trained AI for different functions within our business um and business use cases. And one of the things that was astonishing to me is that we were trying to train a um an SDR um so a sales development rep who would you know do outreach or do follow-up. And what was really interesting to me was that um obviously the hallucinations are still a problem right now uh with it and I think will always be a problem. I'm sure you I don't know the French guy, but the guy who Google listens to and he's some you know >> French special might be um but he basically was saying that like LLMs will never truly get into AGI because if you have 3% error it gets expounded you know uh at infin item >> and so uh >> he did moderate his view but I'll tell you >> oh yeah please yeah tell me um but one of the things that uh I found so cool for me though is that like you know it would it would make a mistake and I would you know we'd be like hey do this next time and immediate and every time it would do And I was like, man, I've trained a lot of sales people like scary in a cool way of just how quick you can learn and it just doesn't make a mistake after obviously there's hallucinations, but in terms of following the directions it would follow them, right? >> And so >> if if you It's funny because there's there's a lot of like really cute isms and in Silicon Valley like you know being a master is something that like some skills cannot be taught and only learned. Like there are these very cute like rhetorical devices but they're just not they'd make no sense. >> Like if you learned it you learned it which means someone could teach it. it's just not structured >> and we didn't know how we taught it because there's multiple factors but like if we if you control the conditions you can control the outcome at least that's my viewpoint of the world and so >> it just means that there's more complex environments in order to learn and so um I think I've just taken a lot of interest in that because you know the obvious of like chat GPT enters Optimus or you know Grock enters Optimus in the real world it's like >> oh okay well it can interact with all the tools that we've already made around the world to to interact with as humans and never make a mistake after that and then that just gets into lots of questions around existence and all that kind of stuff which I >> I I I think you're you're spot on perhaps more than you know because um even a lot of AI engineers would not understand what you just said and even some AI researchers. So what what happened is LLM's exactly what you're talking about train on the internet. Yeah. >> Right. Next word prediction just like keep at keep quizzing AI about like you know uh give it the title to a New York Times article quiz it on the first word. >> Yeah. uh if it's wrong you change the weights if it is right you reinforce the weights and so on >> um and so that that was the paradigm and that's how we got GBT2 three up to four that was the paradigm skip you know train on everything >> but that created uh all sorts of problems like hallucinations and then we ran ran out of data like that's it like a limited set of data >> which is crazy to consider like what that even means >> yeah um but then the next paradigm which is reinforcement learning uh in in in in in the like the paradigm after that. So reinforcement learning from human feedback was the thing that created Chad GBT because you know they it would get an answer and then there's a human that says okay that's not a good answer generated again that's not a good answer generator. So this is like voting basically based >> and so that that made it so that chat is a good chatbot. >> Um now then the next thing is learning from experience. >> This is not a human annotating the feedback. This is actually experience with the with the real world and this is what's creating really profound good coding agents right now >> because the experience from the real world and agent is actually still a virtual world. It's a virtual machine where it's like writing the code and executing it and getting feedback from it. >> And now this is being applied to robotics as well. >> So you know baby robot is exactly the right right way to think about it. We're going to go from um you know this idea of like humans teaching it to more of it like touching and experiencing the world and and growing from it. >> And by the way you know this happened before this happened with Alph Go. The way they were trying to train Go is the game is the game that no one was able to beat initially with AI >> but but uh Alph Go in 2015 was able to beat it. And it was because they moved away from training on human data >> to having the the machine play itself like a trillion times >> um and and that it created an alien intelligence >> and started inventing moves that we'd never seen before. Yeah. Really interesting. like really artistic creative moves >> that that till this day they're studying and they're referencing and it changed the game of go because people start thinking about it differently because we had this like very alien >> that is first principles perspective because it has no biases >> that's right so I think that's really the next >> step of AI where we're going to go from information systems which why I would call chat GPT perplexity all the systems that we had until today that that have achieved wide user base. Maybe there's a billion AI users today. >> Those are information systems. Uh we're going to go to action systems. So agents, uh the SDR you're talking about, that's what you're training there. Like an action system. Uh now Frano was this big skeptic of LMS like he's like look that's not how we're going to get to AGI because LLMs have the hallucination is >> what uh he invented this benchmark called ARC AGI and this benchmark a lot of the big model labs so his skepticism uh created new set of innovations uh so a lot of the model labs started going from training just on human data to training on uh on experience and that's how we got the 01 03 models, the thinking and reasoning models, >> and then and then they started scoring really high on his benchmark, the RKGI benchmark, which nobody scored really well on. Actually, there was a breakthrough two days ago with Grock, and it was like 79% or something like that, and we were like at 2% like a year or two ago. >> Uh, and so he's like, okay, I could see how LLM's could could actually scale all the way to AGI. So, now we're entering the action regime. Um, and this is what gets me really excited because we're building replet, but also this is where I think it starts to impact the real world. >> Um, so, uh, but but even even in the information systems, there's a huge amount of value that still hasn't been tapped. Uh, ACQ AI, acquisition AI, >> you essentially the way I understand it is you embodied your your knowledge into a system that people can go and and and and talk to. You're scaling yourself in a way. talk to me about that. >> So, um, so there's a few things. So, one is when we were building it, I wanted to be really clear to our dev team. I said, I don't want this to look like a chaty like I don't want this to just be like a chat interface because that's not going to be like I don't want to be compared to open. It's like we're not we're not going to win that game. That's not the goal. >> And so, there's a couple things that we did that was unique. One is we added in a context layer that would be permanently in stuck. Um, I don't know why all GPS don't do this, but like, um, so everybody puts all their business information in, which would be all the questions that I would probably ask at like at the at the top level before answering a question. And so if you were to ask because you know what our our a normal user would come in and say, okay, I'm going to talk to Alex AI for example, >> um, it's not it's ACQA >> and say, what should I do with my business? Well, if you were to ask me in real life, I would say, I don't know anything about your business, so I can't answer the question. And so I wanted to make sure that it would always ask questions prior to answering. And so it's a very questiondriven >> um AI in terms of how we built it. We did a ton of QA on it. Um but the first thing we're going to get is all the context around the business first. >> After that, then we're going to get context relative to whatever the the original question the person's asking is. And then in terms of where it gets the answer set from, um one of the things that we spent a really long time on was me trying to be very clean with the data that I was putting as inputs and outputs within the system. um for well obviously it has all my books and all that stuff in there. Um and so when it answers questions it literally searches um in in sequence it goes through all books and then it goes through all playbooks and then it goes through the notes. Now the notes are proprietary um like no one has access to those and it's basically all the consultations that I did over the last two years purposely to train it. >> Okay. And I have, you know, taken businesses from, I mean, commercial refrigeration to >> to a YouTube creator to a, you know, school community owner to, uh, you know, an AI company, right? Like everything under the sun uh, in terms of what those consultations were about. And patterns do start to emerge obviously the same way that you get good at doing anything. Like you see it enough, oh, this is another one of these and well, this probably has two of these three problems. Okay, let's fix it this way. And so, as a result, it started getting really, really good. And so, um, the feedback, honestly, I was very pleasantly surprised. Like, we put a lot of work into it, but you never really know. >> Um, but our usage has remained the same since launch. >> Yeah. I couldn't sign up because it was still overloaded, I think. >> Yeah. >> Uh, but it's it's it's it's um it has stayed the it has stayed the same. And so, I was, you know, if you have any retention in in a in a first product, you'd be like excited about it. But we have >> the same at launch every day getting used right now. Um, and so >> what's limiting you from opening it up to more people? >> It's more about because I did the launch and I did it closed, you know, with relation to the book. I'll probably reskin how I, you know, structure the offer, right? >> Um, in terms of like what I allow people to have access to, what data it is trained on that it's going to answer from. Um, and so I'm we're actively working on that right now, but we are still uh heavily investing in ACQ AI because if we think about what we want to do with ACQ, I see I think W you might have seen this WCOM came out and said like we're looking for full stack AI companies. So rather than saying, hey, build a tool for law firms, just be a law firm that does, you know, that's AI first. And so when we think about AI first, and correct me if I'm wrong because I'm super curious your perspective, but I thought of this as well, you can't be AI first, you need to be data first. >> And then once you're data first and have a data uh capture and enrichment layer first, then you can be AI enabled, right? And so that was kind of the the operating thesis that we've had for the last two years is like we want to capture data in a very specific way so that we can train AI on it. Um, but then with the ultimate goal of being able to be kind of replace what I see as a pretty big gap in the market and obviously where I make a lot of my content, which is kind of low midmarket. So there's tons of people who help people get their first five customers and I do that for free just for anybody. Um, and then there's obviously Fortune 100, you know, Mackenzie Gartner, whatever. I was like, but there's this gigantic gap between >> call it 1 million and 100 million a year >> where businesses in my opinion where the a huge percentage of enterprise value is created on private markets. uh you know going from a $10 million business that has you know $3 million of profit to a $25 million business with $10 million of profit is just a gigantic step up in valuation and wealth for the founder and that's where like I have so >> there's no advice at all >> right there's nothing there there's nothing it's just like either you got the whole Silicon Valley world which is just billy your bust right >> and then you've got kind of small business world which is like how to run a local shop but there's nothing for like how do you go from 10 to 100 and there's just not a lot and so we've done it a lot of times And so we feel pretty pretty confident to be able to help there. And so that's what the advisory practice has been based on. And we have heavy human you know intervention doing that right now. But the goal has been to capture the data such that every day the AI is um further and further integrated into our processes. So right now it's that ACQI is al is an internal tool first. Yeah. So our associates use that to come up with preliminary recommendations and the advisers kind of say okay we need to add this we need to tweak this and then making sure that cycle is as fast you know fast as we can so that overall eventually becomes you know in the beginning it's it's it gets one third right and then it's half right and it's 70% then eventually it's like this is 90% right >> is there now a version of this that we can offer that's cheaper than what someone else can do and better and faster and so anyways that's what we're working towards right now >> so you know uh to to your sort of question about how to think about full stack AI businesses. the the main question any founder who's building an AI business need to answer how is this different than chatbot 100% >> uh or why couldn't openai do it yeah right if it is lucrative they're going so much like they just started a job board like it's a it's clearly a company that's so ambitious you have to worry about what they're going to do what they're not going to do >> um and you know to me the question is what kind of domain specific information you have that is not on the open web, >> right? >> And that is your IP. That is your IP that no one else has. Like if you're a lawyer >> that specializes in certain case law that is very niche, no one else knows about, you can put that information into an AI and have that be the best AI about um I don't know like >> Yeah. IP defense for Instagram res. >> Yeah. Yeah. Exactly. like the more niche the better and there's given how big the internet is and how big the world is there's enough market for that for you to scale it up >> um and so you know if you have that and I tell tell a lot of people a lot of people ask me like okay what's going to happen when AI take more jobs or whatever it's like you're going to have domain knowledge that's not going to be out on the open web try put this into an AI >> uh you might be able actually to generate so much money without working and it's it's like the ultimate It's the ultimate passive income. Yeah. >> I you shouldn't go into it thinking that way, but I think the you know, >> so you know, my my view of where AI and agents are headed, >> we're going to get to a world where uh AIs will be able to go contract out to other AIs. So, I'm going to have an AI or, you know, society of AIS that's running my business. But anytime, let's let's say we got sued by Instagram res. >> Uh and it's like, okay, holy we don't know how to do that. Well, our legal agent can go out and find an agent that is trained on that >> uh and actually just facilitate the transaction. And you can sort of imagine how fast the world becomes >> when when it's like AI to AI protocol. Um, and I think uh I think you know there's going to be a lot of businesses where you don't have to hire a lot of people in order to make a lot of money. >> Um, and so I think it is such an exciting time for entrepreneurs, solarreneurs, small businesses. >> Uh, those are the people that are most worried today. But I think they're, you know, those who understand it should be the most excited. But it doesn't mean SAS is dying. It means SAS becomes a small business yeah >> endeavor which is I think that's great. Um and and so uh and so I I think that if you're uh if you're someone who wants to start a business um and you have an experience working at a company the first thing to think about is what kind of domain knowledge you know that chat doesn't know. >> Yeah. >> And that's a great formula for starting a business. >> Yeah. No, I mean I obviously agree. You know what it's interesting because um you know when we come from the investor perspective because I have you know that hat as well at acquisition.com. >> Um I kind of see like two operating patterns for investing. So one is I want to be one of the winners in AI. >> Uh the alternative is asking the question which is Basos's question which I love which is what won't change? >> Mhm. >> And then doubling down on that. And I think both of those are are great you know operating perspectives. I think the returns will be significantly higher in the um winning the AI, but there will also be tons of losses. >> It's funny, there's a book that it's called the Innovator's Dilemma. Have you heard of it? >> I've heard of it. Yeah. >> Yeah. It's it's it's a really interesting idea, which is um once you're successful as a business, uh you're creating the uh the conditions for your own disruption. Yeah. because you found a certain success with certain customer base and that customer base asking you for certain features that opens up the down market, the lower end of the market for a disruptive technology that when you first look at it, it feels like it's not a threat. It's just a toy. >> Yeah. >> Uh but you know, pretty quickly >> and all disruptive tech is like that. >> Yeah. pretty quickly it'll you know because it has more users it'll have economies of scale or some other advantage network effects whatever it'll move up market and it'll start to disrupt uh your business when uh Clay Christensen uh which was a Harvard business I love that yeah he's awesome >> what is that what is it >> jobs to be done >> no he has that but he has the other book he has which is like >> what is the meaning of your life yeah >> yeah yeah yeah um so yeah I recommend people go look at his stuff uh the the great lecture about like your your me meaning and and your work. Um but when he was writing this book it was in the 80s or 90s and it was in the hype era or when storage uh became so cheap that you had a lot of companies like SanDesk and all these companies come up and sell sell you know storage consumer storage devices and he says every six months we're getting disrupted and he says that when I was when I look at you know you look at biologists genet geneticists and they want to study genetics and DNA they study fireflies because fireflies uh get born in and die in a matter of of a couple days I think or or maybe 30 days or whatever it is >> lots of cycles so you can study and he's like I like to study the hard disk business because it was it was like a fruitfly and I think a lot of AI companies are like fruit flies and how >> that are getting born and disrupted very >> quickly die right yeah exactly um so just to to get into uh replet a little bit and and and broadly uh just software uh you built so software businesses. We just released agent 3. >> Yeah. >> And the main idea behind agent 3 is it refactors the work. So refactor is a word for uh when you're creating a mess. Like imagine you're cooking. As you're cooking, you create a lot of mess and you need to clean it up in order to keep cooking. >> Uh and so with programming, you need to be continuously cleaning up. Uh and non-programmers do not know to prompt the agent to do that. So now we have um we introduced actually it's a multi- aent system. Now we introduced an agent called the architect >> and the architect >> matrix. >> Yeah, exactly. The architect will look at the code that the like main coding agent built and like that doesn't look good, that's not secure, that's not very good, that's not stable, kicks it back and the architect loops again. And we also have a testing agent that opens the browser window so that you don't constantly test it. >> You have to sit there every 5 minutes. That was my that was why for me putting 30 hours in it was probably only like >> uh 2 hours of actual prompts and then like >> you know 10 or 20 of just sitting waiting for the thing and you know >> let me show you the update actually it would be pretty cool. So we we played around our team before we got here. This is only like a couple prompt. This is actually just one shot this app but we wanted to create a money models app. >> Okay. So the prompt is um self-funded uh funnel simulator AIdriven calculator to model money model metrics. This app would let entrepreneurs input their offer price cost expected conversion rate then simulate if their funnel meets alozi selfunded criteria whether in 30-day gross profits per customer is more than twice that the cost which >> uh so this is what it created. So you can, you know, create the uh CAC here, you can create the offer name, um, and then start to simulate it via visualizing the flow. >> So that's cool. >> So by the way, it innovated a lot of that stuff. So we that wasn't exactly in the in the prompt. >> So here's where the offer is. So you can see the the sort of the funnel. You can see the attraction, upsell, continuity. I guess it doesn't have the downell, but we can probably add it here. Um, and so here's a metric dashboard. You see like a big red thing. It's like, okay, this bot selfunding. >> I don't know if ACQ does this. Probably does, right? Does this sort of thing? >> Oh, we just do it. I mean, we we do do it by hand, the old fashioned way, you know. >> Uh, here's the AI feature where it says uh where you can just like click analyze my funnel uh and your funnel needs 80% to achieve. Um, so >> consider raising the price by 20 30%. Okay, that's it has a nice little suggestion there. >> So, I I think what we can do is we can change that app or we can start a new app uh if you'd like. So, just to show you some of the new features, what what should what feature should we add or some of the fix? For example, we can make this interaction to a chatbot. That could be interesting or >> Oh, you know, I wonder if um like, you know, add another offer or something like that to the flow would be >> something we could do here. So, we can we add another You want to add an offer? Okay, so it has that. >> And so, the issue is we need to have something that's way more expensive. So, it probably be like, how do we put a $1,000 thing on here knowing that only like 5% of people will buy it? >> Like an upsell. >> Yeah, like an upsell. here. Um, miniourse upsell. Let's change it to like full course upsell >> price 997. Yeah, sure. >> Um, and then conversion rate. >> It's five. Yeah. Let's go five. See how that goes. >> Now we look. >> All right. >> So then we have >> it's at point4 now because it was at 0 2 before. So that added So we need another So we we're we're so far behind. >> That's interesting. Why are we so far behind? Maybe because the initial price is so low and conversion rate is low. >> Could rate 15. That's pretty that's pretty high. 15 on a on a front-end offer. >> Okay. >> And then the upsell. So that's the 997 upsell. Okay. And the monthly membership conversion, right? We replaced one of the offers, right? >> Yeah, we replaced one of the offers. >> Okay. So we replaced it. So, we need to uh I mean, if we want to make we we could always just change reality and be like, "Hey, we're at 20% conversion on our on our $1,000 thing." And then >> Yeah, let's let's just see make sure it it actually works. >> So, we're closer. Um but what's our Okay, so there is definitely a bug here because if if cost of acquisition is one No, no, there isn't a bug because 150 is cost of acquisition. So that's what's off is that I don't know where the the 150 comes from, but >> Oh, it's right here. >> Okay. Yeah, because if you're if you're, you know, buying a $10 customer, your cash probably >> that's that's an awful business. >> Yeah, that's that's a pretty bad Yeah. So if you're if you're C, you're probably calcing like a true $10 thing on the front end. >> Okay, cool. >> There we go. Now we're >> Yeah, we're crushing it. Now we're crushing. >> So what what kind of feature should we add uh add to this? >> Maybe let's make this into a chatbot. >> Okay. Okay. >> The analysis because it just gives you >> it should give fun. I think it should give offer recommendations. >> Okay. Um make the a AI optimizations suggestions. I'm just going to go into plan mode here just to make sure that they >> Oh, cool. There's mode. So that's new. Yeah, cuz you kind of want that. You're like, "Hey, before you go break all this stuff, >> let's Yeah, let's brainstorm." >> I don't know if you saw this meme. It was hilarious. It was like when you fix a chip on a car, like on paint. And so it had like this tiny little chip and you'd put like tape on it and then you'd you know white out the little chip so that you know the white matches the little thing and then you pull off your tape and then the whole all of the paint comes off and it was like fixing a bug with uh with AI%. This is what we spend most of our time working around. uh make the optimization suggestions about the actual offers um and provide concrete and actionable uh advice >> and I would say using the uh attraction and upsell mechanisms inside the money models book >> uh and attraction mechanisms inside the money models book >> and I would say come with one alternative yeah I say like come with one alternative per offer in the flow that could potentially improve it. >> Come up with one alternative at least at least one alternative to one of the offers. >> Sweet. So while that loads, um, how do you see where Replet fits within the universe of lovable cursor? Like do you see Replet going after a different avatar? Like how do you how do you see where you guys fit? you know, Replet is is more expensive. Uh, and the reason it is is because we're really trying to get you to production. I think a lot of these AI tools are really kind of toys and Replet was a toy for a long time, >> but there's a few things about Replet that are really interesting. >> Uh, for example, >> the feature sets are is a lot more complete. For example, we have a database. You can create an actual database and you can do migration the database. uh here I'll just create a database here you have a development database and you can add a production database as well so once I deploy this will be so this you know it is a lot more uh production complete environment we also have um uh this deployment interface is actually a full cloud system so if you open the advanced setting you can pick autoscale you can pick a virtual machine now you don't have to worry about that until later on. But once you have a million users, you're going to have to worry about it a little bit. And you can ask the agent to help you with that. You can change the machine configuration. >> So no other system out there, none has that kind of depth of the technology has a development environment, but also the deployment environment, >> also the database, >> storage. It's like a full like cloud system. Like we have app storage, we have all of that stuff. So um to mirror back the question you asked earlier about like an example of you know an agency or whatever using you know money models or client finance acquisition in order to cash flow their growth um what are some I mean you don't have to name names if you don't want to but like what are what are some of the biggest apps that have been built on replet? What are their revenues? >> Yeah. So um >> can you see it? >> Yeah. So there there's there's one they tell us. >> So we're actually we're going to get to a point where we're going to start to see it. That's another thing that makes replet different is that we're building a lot of services for the customers. So for example like you know you don't have to go set up off yourself. We can give you off built into replet so you can manage your users here. So we're going to add a way to like charge your users. So we're going to be able to have like a macro view on how how much people are making. >> But a lot of the businesses that are making a lot of money tend to be exactly what you talked about which is >> domain specific knowledge that no one else has. doesn't have >> there's a European VC firm the CFO >> he was using like a hund different software systems like Quickbooks and all these different things >> uh all these finance tools in order to manage their portfolio it's like you know I want to centralize I know the the business of VC it is not PE it is not a b it is not like an internet business >> and so I'm just going to go to replet and build um a piece of software >> that really embodies my my my my full knowledge >> before he quit his job he had millions of ars and commitments >> because he went to every CFO friend that he gets dinner with anyways and says like would you pay for this >> so by the time he quit his job few weeks in he had signed 5 million euro AR >> um and so that that's I think an extreme example of like really fast takeoff success >> uh there is um >> it's clear product market fit which is still the fundamentals still still matter like you can build whatever you want if no one wants it doesn't really matter Um there there's this uh product uh general artificial intelligence proficiency institute geni uh that's also one of the fastest uh scaling uh businesses on replet um and this entrepreneur uh John Cheney what uh he did there was two kind of catalysts for him by the way he's he's a guy that went the Silicon Valley route and didn't like it raised tens of millions of dollars all that just didn't like it >> he he's just like He wants more control. He wants to I get it. >> Yeah. He doesn't want to like manage. >> Come over to my world. >> Yeah. Exactly. Um and so, uh he gave himself a challenge to build like a business on LinkedIn to build like a business and get the first customer like in a week or something like that. >> Uh and he was out to dinner uh with with with their neighbor and um he asked his neighbor uh who's who's selling tractors. So he's like lives in uh farmer land and he asked him how to use chachi and I was like what is chach tea right >> and then he showed him what chachi is and he like the next the day that guy is using it for everything >> I was like wow there's so many people that just needs a little bit of handholding so he created this >> uh platform to have certifications live courses um live tests you know there's a test here question answer >> and he got I think he got to 200,000 in uh in revenue in the first two weeks. >> I think he's closing in on a million dollars a month right now. >> Oh wow. >> Uh and he sells to individuals, but he also sells to enterprise. A lot of enterprises just don't know how to implement AI. >> Speaking of which, what's headcount for you guys right now? >> Uh by the way, I forgot to to kind of kick it off. So here's uh >> um >> it's a robust plan. This sounds good, but I need you to make the to prompt the AI to do this. Create a plan to do that. Um, yeah, it's like, oh, here's all the suggestions I have. Well, no, I want you to prompt AI in the app to do that. Um yeah so headcount uh for replet um last 2023 24 were actually very tough for us because we had built all the so the reason our platform is so deep in technology right like it has the virtual machines the APIs AI all of that >> is because I spent 10 years trying to build this >> uh and most of my time was not spent on AI most of my time was spent on the platform >> I was like if you want to make it so that anyone can make software yeah >> you need to make a consumer grade cloud system. No one will know how to use AWS if you don't have a coding background. So, we basically built like like consumer grade AWS. >> And then in 2024, we weren't making a lot of money. We're burning a lot of money. The team was really huge. We had to lay off half the team. >> Oh, wow. >> And I told the team, look, we we just have to solve coding. >> At that time, there was no lovable bolt, none of that. There was it was not a market. >> And I said, we just have to go solve coding. have to go build an agent that can do the coding for you. So we spent six six months built that we spent six months struggling with it because the the open eyes of the world still haven't really started optimizing for agents. Really felt like we're going to get there and actually what's interesting is replic catalyzed that. So we launched in September the system was very rough and bare bones and all that >> but it inspired a lot of people including the AI researchers. It's like huh this is possible. Yeah, >> it's possible to have an agent not just write the code but create the database, run the SQL migration and do the deployments. >> Okay, let's go optimize against that. >> And so it is sort of this feedback loop where we inspired the researchers to go optimize for our thing. And then a bunch of uh you know competitors and copycats came in and of course that's what happens. Um and so everything uh like you know we went from you know two three million to like 10 million in by the end of the year. >> Uh and still there was disbelief inside the company. Actually people continue to leave uh after they you know you mentioned when you get a down round that kind of destroys the dream. Same thing with a layoff. It really >> Oh totally. It it really kind of destroys that dream that people were like living through. And there was this belief that we're going to make it >> despite like having a revolutionary product. It's kind of really insane like how human psychology is. >> But the revenue kept going up. The feedback we're getting is is >> where did the big was there a big inflection point at some point? >> The big there two inflection points. Um in uh in February we uh got out of beta and the and the and the product started being more stable and also we reintroduced our mobile app and now a lot of people started coding from their phones. So that increased the the market. >> Uh so >> I did that too. >> That was because I could watch I could like watch TV at night and just like just give it the prompt and it comes back in five minutes and then you just give the next prompt. >> We know our audience that they're busy. They don't want to be in front of the computer all the time. So that was one and then April was version two. So version two was a lot more stable. Uh it started actually working and that and people started spending a lot more on replet and so that scaled our revenue. >> Okay. With tokens and all that. >> Yeah. Do you think um can replet build something like a CRM? >> Yeah. Yeah. A lot of people build CRM. Like there's a private equity guy that I saw. He wrote a story the other day where he's like you know there's the hotspots of the world. They're all great, but I need like a private equity >> CRM >> CRM. And so there there's a lot of >> and the market is so small and because they don't do transactions, no one's really like wanted to go after 1,000 customers that and they're not really going to say, "Oh, like >> for the economics to work with a thousand customers, you got to be charging like a million dollars a year." And they're just not going to justify it because they're like, "We're only going to make eight eight investments or whatever it is." But you still want it. >> Yeah. So you end up like for us we had to we basically retrofitted you know a HubSpot or Salesforce whatever like out off the off the gate just made it look like all of our deal stage pipelines just worked like enterprise sales but we just had to have like more steps of like diligence one diligence two team diligence you know had to work work work our deals through that pipeline um but yeah no I mean it makes sense because because then once once you're there it's like what does customer success look like >> when you make an investment >> right >> and see it's a little bit of paradigm shift but you know some of the some of the the fundamentals still apply Yeah, it's sort of like you go from you go from like a world of like large malls and kind of that that try to cater to everyone or large food chains to local uh food >> local coffee shops. You Starbucks is great >> and it's going to be okay for most people. >> But uh so now it's booting up the app. It built the feature. spinning out the app in a in a in a browser window so it can save us from the QA. That's great. So, it's interesting what you bring up though because there's this this uh it's kind of like an accordion effect that's really common in business is >> things just do like this. They like they centralize and they decentralize and they centralize and they decentralize as these new layers of kind of the internet, you know, first is like the internet came out and then there was all these different forums and blogs and then the issue becomes finding them. So, then the search, you know, search becomes this this big. >> It looks like it it implemented it. All right, let's see what we got. >> You could tell >> when you're high impact. They even rated it high impact. What we got? >> Or is it still testing it right? >> It's still testing it, but but it's good to to >> to kind of look at it when it's testing it. And you can read its thoughts and you can read what it you can look at what it's looking at. >> Uh so that's its eyes, this like purple thing. And >> okay, >> yeah, >> I've successfully triggered AI analysis. It looks like it's generating some helpful suggestions. You know, it's like I'm going to click reanalyze and see if it gives me another thing. Did you click them or Oh, it said that >> it's like a robot, right? And this is this is what we're talking about. >> The action agents now. We're going to the world where they can use the computer >> and we can just sit back and eat popcorn to look at it. >> Oh, good move. You know, >> but and now it's restarting the server. I think it it it passed the test. So, you can see a check mark. And so, it's going to pass back to the main agent and says, "Okay, I tested it for you." Uh, and and it passed. and it'll yield control to back to us in a second. It's interesting that you went to the multi-agent approach because um you know with the the STR thing, one of the things that we were considering was um instead of having an agent that was master at this you know conversation especially from especially from a text perspective not voice is still has some latency which is a pain but like just from a texting perspective um it it started to make more sense to have like an AI that was only trained on first message >> and another AI that was trained only on second message and like the more specialized we can make them like and all of them are optimized around a singular goal which is how do I increase the likelihood that someone responds as the like you're not going to get someone booked on the first message. So then we need to optimize for a different goal. We have to get them to respond. >> One thing to try there is >> you might actually find that Gemini is better at the first one. Yeah. >> Open is better the second one. >> Interesting. >> And even GT5 is better the third one. So we use every model on the market including the open source models >> because like you know for example G G5 yeah >> it's actually very kind of anal and annoying >> and and it's just like so it's the architect >> in this case it'll it'll be like it looks like cla is like that sucks and claude tries to negotiate with it. It's like really funny to look at it. So you end up like that's why sub agents and and multi- aent system >> that's another advantage you can have over open AI open AI can only use open AI models >> app layer companies can use all of the models available including the Chinese models >> right yeah >> um so I I think the multi- aent approach is the future >> what do you think about local because I know you know GPT came out with their our openi came out with their like locally hosted version like how do you see how do you see that interaction with normal mainstream business. >> I don't see much of an interaction with it. I think that >> people might still use open source but they will not host it themselves. >> Okay. >> Partly because there's so many good inference systems. You've heard of Grock with a Q. >> So Gro with a Q is this um inference uh company that hosts open source models for you like Llama like things like that. Another one is Sirius. It's about to go public. So that there's a bunch of these companies that they will do a much better that job than you at hosting these open source. So I think open source has a place and it's mostly because of cost, maybe because of fine-tuning, >> but the idea that I'm going to download a model hosted on my server inside the company, I don't think it's going to happen. >> Interesting. This is where like the the training becomes so important. And that's like with ACQA has so many of my inferences and so many times where I'm like >> the the thing that has been hardest for me to replicate and it'll just take time and iteration is like I have a very good idea of benchmarks across hundreds of industries of like what would uh an in-person close rate for a plumber be? And is it low or high based on where they currently are? And >> based on that number, is that the constraint of the business right now or is it something else? Or is it just their lead cost is too high? because like CAC being high could be one of six different problems. And so being able to walk through that decision tree and then use benchmarks as really quick litmus test or triaging the answers is what I've basically spent my whole career doing. >> Um and trying to transfer as many of those things over because I've just had so many experiences and repetitions and like a lot of that data is not public. Like no one know like no there's I don't know any blogs that are like these are the average conversion rates for plumbers. And also that's different than roofers which is different than doortodoor which is different than Google search which is different than meta ads. And so like if I have a a pest control company that comes in you know for for our headquarters and they're like hey we do door to door we have 100 sales guys who deploy them to every new market. This is how we you know uh canvas an area and then and and deploy it or build our build our subscription base. It's like okay we want to have some sort of uh coverage after we move away from the market to at least maintain above churn so we can just keep that market being you know profit or whatever. And so it's like, okay, well, we could we have a Google search strategy, just AdWords, and then we have a meta strategy. And I'll have to walk them through the fact that like, yeah, you might be able to convert 50% of your Google search leads because intent's super high. Now, your lead cost will be 10 times that of your meta leads, but your metal leads uh will be onetenth the price and you're going to convert 10%. But they'll come in and say like, meta meta doesn't work for us because we're only converting like 5% of leads. It's like, okay, well, slow down. you only need to double this for this to be effective and the ocean of traffic that's there is so much higher because intent there's only so many people searching once you once you max that out in a given market that's it there's nothing out there's only so many people searching but from an interruption based you know marketing perspective it's like you open up the entire market because you can run Tik Tok you can run YouTube you can run Facebook Instagram whatever and so but that nuance of knowing I mean this is what I mean this is what I've spent my whole career doing >> what you just said and it will learn >> what you just said is is a is a prompt essentially you need to structure it a little different. >> You know, I I'm you know, the other day I was trying to come up with this analogy of, >> you know, how the Airbnb guys looked at an unmonetized asset, which is a bedroom in their >> uh and unused bedroom. Yeah. Um unus capacity. The cloud, >> what basos did with AWS is also thing same thing. You know, >> Amazon needs all these servers to handle Black Friday, but most of the year they don't need any of them. >> Yeah. Uh and so access capacity you can monetize. I think domain knowledge is the same way right now. >> Yeah, it's a really great perspective. >> Yeah. And so I I think people watching should like really uh introspect and I bet you have a domain knowledge that perhaps >> super valuable >> super valuable and very few people have >> and that is a prompt that you can embed into an AI >> and now a lot more people can >> you can you could be consultant at scale essentially. I mean and that's that's what we're I mean that's what we're trying to build here you know is how can we help businesses grow and we do we have developed kind of like our very like it's a decision tree that has probably like 20 paths and then within those path there's another six more path you know so there's it's just a massive decision tree but fundamentally like >> but it's finite >> but it exactly it's not infinite like there's only I mean because you can always chunk it all the way back up which is like we either need cheaper leads we have to convert more of them we're mispriced right or we have a cash flow issue or we have a supply issue like we can't deliver on whatever it is that we sell. There's the issues, right? Right. Then each of those has trees underneath of them, but like that's what it is. And so figuring out having the larger that's why the context layer was step one for us of like building this. So it's like, okay, well, at least we have these. Now part of the issue that comes into it is a lot of people who are smaller businesses don't have good data. >> And so then that then you have to start asking follow-up questions that are you have to be >> in exactly questions about it. Yeah. >> And even sometimes it's like I don't know what my CAC is. So we have to say like okay well you know what was your marketing spend last year? How many customers did you get? Okay. Well we can back into CAC that way. It'll be broader and it's not going to be segmented by channel. But >> how much I have a question. Um how much of of CAC could be um you're getting the creatives wrong. you're like how much is CAC is set in stone versus like you're actually doing something wrong on because you know my team tells me right now is like Google ads is like on on rails right now >> like is that true like you can't be creative about the way you present these things >> I mean you totally can be yeah I mean yeah 100% I mean you're bidding for keywords but you still have all the headlines and the copy that's going to go there and you have endless variations of headlines copy >> and how important is that to k >> super yeah >> super important >> is the difference between $10 CAC and $100 CAC. >> Um I would that's a that would be a big one, but you could definitely get doubles, triples, sometimes 5xs. 10x is a big one. >> Um so for that kind of thing, the difference would probably be closer to offer than copy. >> Mhm. >> But the copy describes the offer. So these are, you know, kind of interrelated. Um and with keywords uh and search, it's going to be it's text based predominantly. >> Uh whereas where the creative will absolutely can have 10x 100x differences, it's going to be on video based, image based. That's where you can have the huge alphas. And where where this gets really interesting is when companies are trying to scale their advertising. So like school, we spend, you know, several hundred,000 a day in in advertising. >> Um, and what's what's what's interesting about this is that I was having a conversation yesterday with um a friend. He was like, I can't scale this offer past 2,000 a day profitably. As soon as I get above 2,000 a day, it stops working. M >> and so I was explaining to him I was like well right now the way that you have the offer and the copy structured is targeting a very aware customer and so you know there's stages of awareness Eugene Schwarz old school direct response marketer he talked about five stages awareness you've got unaware uh problem aware solution aware product aware and then most aware which is just your existing customers and so your most aware customers and product aware customers you can simply just make an offer and they will buy but if you just say here's my offer here's my offer you're not going to be able to scale to mom who's across Ross the world, she's not going to she's not going to respond to that. And so CS can get really high. And so it's kind of having this layered approach from an advertising perspective where up to 2,000 a day just demonstrates that there's at least with your existing creative um you can only really reach whatever 2,000 of let's say it's $20 CPMs. It's like you can reach a million people a day that um that will respond to that level of advertising. Now there'd be a separate campaign that then hits probably a 10 times bigger trunch when you go to just uh you know solution aware right and it's like okay uh what are the different you know um I don't have you have the product in mind but like uh if you're trying to lose weight it's it's like five things five curious habits that um are are are wrecking your you know energy >> that might be at the most unaware level because I that could lead to a pill that could lead to a workout that could lead to a gym >> because it's just pure curiosity when you're at the top of the funnel, right? And as >> that's what you should aim for. >> Well, if you can get there because as you go up the funnel, you go to I kind of think of like watermelon seeds where like at the base at the base of the the triangle, you have lots of seeds. So, there's lots of high likelihood conversions and the the platforms are optimized to get you those conversions, right? >> As you go to, >> you know, thicker and thicker slices, the seeds are more disperate, >> right? And so, it costs you more because it takes more eyeballs to reach another seed. >> Yeah. But if the money model is right and the creative is structured so that you can attract that person and then move them through the process through education, indoctrination, whatever you want to do, um that's where you get these these funnels that can overnight go from like zero to a few hundred,000 a day and spend a million a day and spend um and do it profitably. Obviously payback period is also a factor of that because I you know I design all my stuff so that we can do it in 30 days but if we can do it in six months I can spell I can spend a hell of a lot more right as long as you have revenue retention. And so um but that's that's typically when we have a CAC issue it's going to be offer driven >> you know number one like number two would be it would be a creative issue >> which can be a factor of just not making enough creative which is really the symptom or sorry the cause of not good enough creative. Mhm. >> You have to make enough to be good enough because you just need more variations, right? And so then you have >> better creative overall. Um there's obvious CRO within the funnel, like does the page load quickly? Um is the is the CTA above the fold? Like simple things, but people miss them, right? I I had a business the other day that came in that was in the dating space. >> Um and he had this whole funnel. >> I was doing three million bucks a year or something like that. And I went through it with him and his his like his final call to action to get them scheduled for a call. It had a video and the scheduleuler was at the bottom, but they had just watched a video. They just watched like a 30-minute video to get that pitch to schedule. And he said, "Watch this video before you book." And I'm like, "They just watched a 30inut video? Like, let them book." And it was crazy cuz I I told him, I said, "As as silly as what I'm going to say sounds, I'm going to give you five different things to do today." I was like, "But the one that could triple your business is going to take 5 seconds." Right? It's just like it was just below the fold and and I I'll bet people just didn't see it. >> And so sometimes it's tiny hinges like that that make huge swings on the doors. >> Um and so you know I I think of this as like I have this gigantic bag of tricks, right? And it's just this gigantic checklist that you just learned through doing this. >> I wonder if we can we can put like Hermosi bot in in like a >> in a funnel like it could crawl a funnel and then >> Yeah. And it can look at the visual. See, that would be a very valuable thing if you had a bot that could crawl a funnel. Yeah. >> And then make suggestions on like kind of like almost like like it takes a picture of each page. Exactly. >> And then just doodles. Yeah. And then just says like >> I think you have an offer issue here. And it could crawl your ads library page and say like oh your ads the the hooks are off here or the hook isn't congruent with the headline. Yeah. >> And so again there's all these different check marks that you have to go through. >> Yeah. So when we publish here, uh we actually let you uh do a security scan >> um before before you publish >> uh and um uh I don't know where it is important. >> Yeah. But I I I wonder if we can also have funnel scan. So maybe this something and we on the security scan, we didn't build it ourselves. >> Yeah. >> Uh we collaborate with a company called SDRP to to to run the security scan. So, I wonder if we collaborate with you to have like a funnel scan. >> Yeah. No, that would be cool. >> Yeah, that would be awesome. >> It'd be an amazing uh an incredible lead magnet. >> Yeah. >> Yeah. Yeah, because I think what so my opinion when we're talking about the software stuff, I see one of the big opportunities is the amount of free tools that people will start being able to give away because I sure there's ways that we could, you know, charge for subscriptions, but like >> a funnel scan for example is probably not a tool that you would use on a on a on a consistent basis. You probably use it one or two times. You give it a huge amount of data to get an inference. Mhm. >> And so those types of products I don't I don't those aren't the types of things I would want to sell. >> Those are the types of things I would I would want to give away in exchange for tons of data. >> So that then we could, you know, sell them other stuff that that we'd find. Okay. Well, now that you've optimized your funnel, what about your traffic, you know, like and that's something that would have a more consistent because once you have the the funnel down, it's kind of this asset that you build and once it's there, it's like you don't really you can always tweak it, but like >> n you know, 8020 is already there and so the the iterative process will be less so whereas the iterations you're going to make on traffic uh are going to be so much uh more consistent and higher volume that that's where you'd probably put the more subscription related product. >> Right. Right. >> Yeah. But I think this is just interesting from a business perspective of thinking like >> which of these different problems can we solve? what is the nature of the problem that's being solved? How how consistent is that problem need to be attacked? And then that kind of separates your free level from your paid levels and whatnot. >> Yeah. The the thing about AI that that is tough right now for us especially >> is how expensive it can be. >> Um and so like a herozi funnel scan bot um might actually end up being like a couple bucks, >> three bucks, something like that. >> Yeah. >> So yeah, I just >> like our AI costs probably a million bucks a year to run just in server cost. >> Yeah. So like the calculation on free is cuz like you know software systems Silicon Valley internet businesses for a long time the marginal cost is zero. >> Yeah. Now it's not that >> it's not zero. Now it might be two three four $5 sometimes more. >> Uh so it's changing. I mean I'm hoping that a lot of the Chinese models and things like that that are coming on the scene >> that are like good enough to run some of these free experiences and upsell them on the like more advanced AI. >> Yeah. But yeah, that changes how I I'm thinking about because I I I scaled rapidly to 40 million users, but just like really optimizing the heck out of the free experience and just like making it so easy. >> Oh, it was great. I mean, the upsell was so so easy cuz you you get your first 10 prompts in and then it's like, hey, used up your tokens and you're like, damn, now I'm halfway in, you know, I want to finish this thing. No, it's great. It's super good. >> It was seamless. I paid for annual. >> Oh, awesome. Well, we did your CA. >> Yeah. Hopefully we'll get you back. We'll get you back to try it again. >> I have to get my my uh my fitness app to to to work on my phone. >> Agent 3 will be able to refactor it and test it and I'll trust agent 3. >> Yeah. Okay. Great. Great. >> Appreciate you. >> Awesome. Yeah, this is this was great. This is really a master class. I think I'm really excited because as Replet is moving to support a lot more entrepreneurs like the ones we talked about, this this this stuff is is a gold mine. So, really thank you. >> No, thank you, man. I appreciate. Congrats on the success. >> Thank you. >> Yeah. Remember us little people. You know, goes both ways. At Replet, we believe everyone should be able to create software without limits and without knowing how to code. Our vision was never to just build a tool. We set out to create a teammate, a technical partner that runs longer, thinks smarter, and works like a builder. An autonomous agent for everyone. Previous agents, while great in many respects, were far from autonomous. They needed constant attention and handholding. That was yesterday. Today, we're proud to announce a major breakthrough. It's a moment we've been building towards since day one, introducing Agent 3. Agent 3 delivers a 10x increase in autonomy, redefining the user experience from the ground up. That means more progress, less micromanaging, and more time for your ideas. Let's take a closer look at what makes Asian 3 so powerful. Asian 3 runs on its own for hours, handling full tasks autonomously. We went from 2 minutes to 20 minutes to 200 minutes of autonomy. You can track your project's progress in real time from anywhere with live monitoring right on your phone. Asian 3 can now test and fix its code, constantly improving your app behind the scenes. We invented our own validation system to speed up quality control, making it at least three times faster and 10 times more cost effective. It checks every button, every API, and every data source, ensuring every part of your software works. For the first time ever, Agent 3 can build other agents. agents that can automate complex workflows using natural language and connect to your data sources and services, allowing you and your users to interact with your own agent. Agent 3 isn't just an upgrade. These new capabilities along with everything else Agent 3 brings will unlock entirely new frontiers, redefining what's possible. We're so excited about what Agent 3 can do and we can't wait to see what you're going to build with it. [Music]

**Alex Hormozi x Amjad Masad: Building in the Age of AI Agents** Alex Hormozi sits down with Replit CEO Amjad Masad for an unfiltered conversation about business fundamentals, mental resilience, and building software in the age of AI. From CAC/LTV economics to autonomous coding agents, this is a masterclass in first-principles thinking for founders. **What You'll Learn:** - Mental toughness frameworks: the difference between fortitude and resilience - The $100M money model: how to eliminate burn rate and build capital-efficient businesses - CAC optimization strategies that drove Hormozi's gym empire - Why vertical SaaS is facing an existential crisis - How AI agents are transforming software creation - The real breakthrough in AI: from information systems to action systems - Why Replit Agent 3 changes everything for non-technical founders **Key Topics Covered:** **Business Fundamentals** - Why there are no rules for entrepreneurs - The only CAC/LTV math that matters: get customers to pay for the next customer - Drive CAC as low as possible, drive LTV so high you can outspend everyone - Case study: How float tanks achieved negative customer acquisition costs - The counterintuitive power of adding friction to increase conversion - The "down sell" strategy that crushed churn and created the 2nd highest LTV customers **Mental Frameworks** - Mental toughness vs mental fortitude vs resilience: the 3-part model - Emotional regulation without behavioral change **AI & The Future** - Why AI learns the same way humans do: reinforcement learning explained - From information systems to action systems: the next paradigm - Why reinforcement learning from the real world creates good agents **Replit Agent Deep Dive** - Hormozi's 30 hours building with Replit: from fragile to production-ready - Why other AI coding tools are toys compared to complete feature sets - Real customer stories: $5M VC CFO got to revenue, $200K in 6 months **Industry Shifts** - The innovators dilemma: AI companies born and disrupted rapidly - Multi-agent future: specialized sub-agents working together **Live Funnel Audit** - Hormozi analyzes acquisition strategies in real-time - Moving customers through the funnel: the watermelon analogy - Scaling from initial revenue to $1M+ in profitable ad spend **About the Guests:** Alex Hormozi is a serial entrepreneur and investor known for scaling Gym Launch to $100M+ and author of "$100M Offers" and "$100M Leads." Amjad Masad is the founder and CEO of Replit, building the platform that lets anyone create production-ready software using natural language.