AI CEO: How to build a $1B Company in 2 days | Amjad Masad @replit (YouTube Video Transcript)

(00:00:00) Your YouTube transcript will appear here (00:00:00) How far ahead you think is time when a (00:00:03) solarpreneur is going to build a billion (00:00:04) dollar company (00:00:05) >> in the next few years? Yeah. (00:00:06) >> This is Amjad, founder and CEO of (00:00:10) Revlet, an AI powered coding platform (00:00:12) that turns your ideas into apps. (00:00:14) >> Our mission is not just to make software (00:00:17) more accessible, but really make (00:00:19) entrepreneurship more accessible cuz (00:00:20) creating a business is really one of the (00:00:22) best feelings in the world. (00:00:23) >> But let's be real. Can anyone just sit (00:00:25) down with AI and build a billion dollar (00:00:27) company? Amjad says it comes down to (00:00:30) three simple steps. They turned Replet (00:00:32) into a $3 billion business. I tried (00:00:35) those steps myself and uh what happened (00:00:37) wasn't what I expected. (00:00:40) Hello everyone and welcome to Silicon (00:00:42) Valley girl. We have an amazing guest (00:00:44) today. We have Jad the founder of (00:00:46) Replet. I talked about Replet a few (00:00:49) times on this channel because I've (00:00:50) personally been using it. Um and uh I'm (00:00:53) just fascinated by your journey. I (00:00:55) wanted to start with this question where (00:00:57) you said uh you're going to empower a (00:00:59) billion software engineers or (00:01:01) programmers in the next couple years but (00:01:04) at the same time I saw you say that in a (00:01:06) couple years companies wouldn't need (00:01:08) software engineers. Can you explain (00:01:09) that? (00:01:10) >> Yeah, I I I'm mostly talking about (00:01:12) entrepreneurs like us. Um I think that (00:01:16) bigger companies will always need (00:01:17) software engineers but people who have (00:01:20) an idea and everyone has an idea like (00:01:23) you know one experiment to do is go on (00:01:25) the street and stop people like do you (00:01:26) have a business idea everyone has an (00:01:28) idea but uh for the most part the thing (00:01:31) that's stopping them is that they don't (00:01:33) have the technical skills or they don't (00:01:36) have someone you know as a programmer (00:01:39) growing up all my friends were like oh (00:01:41) hey can can you program this idea for me (00:01:43) well now you could do it. And so we're (00:01:46) getting to a point where you can run a (00:01:48) business and it's difficult. It's still (00:01:49) that technology needs to mature, but we (00:01:53) have a lot of stories where people have (00:01:54) built their dream apps and they've had (00:01:56) these ideas for like 20 years. (00:01:59) >> We're talking about um CFO at a VC firm. (00:02:03) He's a domain expert. like he knows how (00:02:06) to manage a VC fund and he never found (00:02:11) the right tools for him and he had all (00:02:13) these ideas on how to build them but you (00:02:16) know it's almost it's always hard to (00:02:19) find engineering resources so he he used (00:02:22) Replet and 3 months he built his dream (00:02:25) app and he went out and sold it and got (00:02:27) a lot of contracts and I think he made (00:02:30) um he's on his track to make 5 million (00:02:32) quit his job now he's an entrepreneur (00:02:35) and he told us every time he he's saying (00:02:37) well at some point I need to onboard a (00:02:39) software engineer and maybe he does but (00:02:42) you know it's been he got to 5 million (00:02:44) in revenue and still he he didn't have (00:02:47) to (00:02:48) >> I'm sure you know Meta and Openi and us (00:02:51) we're always going to need software (00:02:52) engineers but there could be a lot more (00:02:55) entrepreneurship in the world a lot more (00:02:56) businesses if that bottleneck that is (00:03:00) making software goes way. (00:03:03) [Music] (00:03:08) >> Do you have like a screen here where you (00:03:11) track your most important metric and (00:03:13) what's the most important metric? (00:03:14) >> So, every every team has a screen, (00:03:16) infrastructure metrics, there's like (00:03:17) product metrics. It's really depending (00:03:19) on the on the team and almost every team (00:03:22) has a has a has a screen. Um, (00:03:25) >> so what's one universal metric (00:03:27) everyone's looking at? (00:03:28) >> Um, I mean, ARR Okay. (00:03:32) >> Like everyone's responsibility. (00:03:33) >> Can you share your recent era or (00:03:36) >> Yeah. Um 160 (00:03:38) >> 160. (00:03:39) >> Yeah. (00:03:39) >> That's amazing. You've been they've been (00:03:41) growing like crazy. Like your graph is (00:03:44) dream white combinator hockey stick. (00:03:47) >> But especially if you plot it from like (00:03:49) 10 years. (00:03:50) >> Oh yeah. Right. (00:03:50) >> And it does that. (00:03:52) >> That's And also like it's so encouraging (00:03:54) for entrepreneurs who don't see progress (00:03:56) right away. (00:03:57) >> That's right. That's right. How many (00:03:58) people are (00:04:00) >> right now? Let's see. Is there about (00:04:02) 50,000? (00:04:03) >> 50,000 people (00:04:03) >> right here. Yeah. (00:04:05) >> Do you know how many apps built on (00:04:07) rapullet are actually active and (00:04:09) running? (00:04:09) >> Yeah. Uh about (00:04:12) 350,000 like paid online apps. That's (00:04:16) growing fast. That's growing 25% month (00:04:18) over month. (00:04:19) >> Do you know how many of them actually (00:04:20) generate revenue? (00:04:21) >> We don't. I think those are more (00:04:23) anecdotal the stories that we hear. (00:04:25) Pretty soon we're gonna be uh helping (00:04:28) you integrate with Stripe and monetize. (00:04:30) So we'll be able to to track that. (00:04:31) >> I think I already added my Stripe to (00:04:33) >> So yeah, the agent knows how to do it. (00:04:36) >> Uh but we're going to make it even more (00:04:39) effortless like one click. Our mission (00:04:40) is not just to make software more (00:04:44) accessible, but really make (00:04:45) entrepreneurship more accessible because (00:04:47) that's really the thing I think that (00:04:48) changes lives the most. You can make a (00:04:50) piece of software, it's fun, but making (00:04:53) creating a business is really one of the (00:04:55) best feelings in the world. (00:04:56) >> Let's take a quick break. If Amjad's (00:04:58) story with Replet fired you up and (00:05:00) you're tired of just watching people (00:05:02) talk about AI instead of building (00:05:04) something real, this is your chance. (00:05:06) HubSpot just dropped something called (00:05:07) the 102 ship vibe code prompts. It's a (00:05:10) pick of 10 proven frameworks to help you (00:05:13) turn a business idea into a working lead (00:05:15) genen web app. No coding experience (00:05:18) required. You'll get tools like viral (00:05:20) hook generator, lead quality optimizer, (00:05:22) multiplatform deployment assistant, and (00:05:24) my favorite, the replet build prompter, (00:05:26) which acts like your own AI product (00:05:29) manager. These prompts will help you (00:05:30) define your perfect customer, build your (00:05:32) prototype in replet, generate viral (00:05:34) hooks, names, and launch content. It (00:05:36) will also help you benchmark your idea, (00:05:38) and plan how to turn it into scalable (00:05:40) growth machine. No more waiting months (00:05:42) or spending 20K a month on a DAV team. (00:05:46) This is what building with AI actually (00:05:48) looks like in 2025. If you're ready to (00:05:50) launch, download the 10 zero to ship VIP (00:05:53) code prompts for free. Link in the (00:05:55) description. Big thanks to HubSpot for (00:05:57) sponsoring this video. So, actually, I'm (00:05:59) trying to build something with Replet (00:06:01) right now. And the thing is, it's (00:06:03) something that I'm encountering. So, (00:06:05) yes, it's building like a beautiful (00:06:07) layout, but then sometimes but see (00:06:10) service unavailable. That's that's the (00:06:12) recent bug I got. So, it feels like (00:06:15) It's (00:06:17) >> oh interesting. (00:06:18) >> Yeah. (00:06:19) >> On the deployment. So you can go to logs (00:06:21) here and understand why the service is (00:06:23) unavailable. So you can see there's an (00:06:25) error. (00:06:26) >> You can copy that error and give it to (00:06:28) the agent and tell it um when I deploy I (00:06:32) get this error. (00:06:33) >> Okay. (00:06:36) >> But basically what I'm realizing uh is (00:06:39) that it's still a little work, right? (00:06:41) >> It's it's work. If you have this data, (00:06:43) how long does it take to build something (00:06:45) that's actually working? Like what I'm (00:06:46) trying to build here is a tool that's (00:06:48) going to analyze my videos and let (00:06:50) anyone analyze their videos on YouTube (00:06:52) and determine videos that have potential (00:06:54) if you change their title and thumbnail. (00:06:56) You know how it works? Sometimes you (00:06:58) repackage a video and it just starts (00:07:00) getting all the new views. And I'm (00:07:02) trying to build something that's going (00:07:03) to help me do that. (00:07:04) >> But I've already spent like six hours (00:07:06) and it's it's a process. (00:07:08) >> Yeah, it is a process. you're still (00:07:10) acting kind of like a software (00:07:11) developer. You're acting like a software (00:07:13) development manager. (00:07:14) >> And so you have this um powerful but (00:07:18) easily distractable intern and you need (00:07:22) to manage him very well. So for example, (00:07:25) you type this prompt there's like only (00:07:28) one sentence. I would have like spent (00:07:29) maybe like another minute or two on it (00:07:32) and just say when I deploy the site I'm (00:07:35) getting this error but you know but I'm (00:07:37) not getting it in the preview and so (00:07:40) communicating in a more precise way is (00:07:44) is very important. So prompt engineering (00:07:46) and prompting is not that different than (00:07:49) programming. (00:07:50) we just take away the syntax from it, (00:07:52) right? Like you don't have to understand (00:07:54) the syntax and a lot of the underlying (00:07:56) details, but you still have to be very (00:07:58) precise and actually it helps when (00:08:00) communicating with developers as well to (00:08:02) be able to talk that way. So, you know, (00:08:04) an app like that will probably take a (00:08:06) couple days, whereas previously even, (00:08:08) you know, senior engineer would have (00:08:10) taken them like a couple weeks, you (00:08:13) know, before AI (00:08:13) >> $10,000, $15,000. (00:08:15) >> Yeah, something like that. It will cost (00:08:16) you something like that. But I would I (00:08:18) would spend two, three days on it. I (00:08:19) think you'll be able to get it. (00:08:21) >> How can I learn to be better at (00:08:23) prompting? (00:08:24) >> We have a YouTube channel. Uh we have a (00:08:26) great developer relations uh person who (00:08:29) creates a lot of content. His name is (00:08:30) Matt. Um and so we try to train people (00:08:33) on on prompting and and the underlying (00:08:37) systems. So Replet has a DNA and sort of (00:08:40) education. So we when we were talking (00:08:42) about a billion developers, um these (00:08:45) billion developers need to learn. It's (00:08:47) not going to come for free. So there's a (00:08:50) learning curve associated with it. And (00:08:52) you need to be resourceful. So you need (00:08:54) to go to YouTube, search like how to how (00:08:57) to prompt. You need to spend a lot of (00:08:59) time and practice by building changing (00:09:03) your style. Some people go to open AAI (00:09:06) for example and like pick 03 or right (00:09:09) now it's GP5 with thinking and give it (00:09:12) the idea and and tell it hey I want you (00:09:15) to structure it into a really great (00:09:17) >> it for me something like that. (00:09:18) >> Yeah. So I think uh you know Paul Graham (00:09:22) uh the founder of Y Combinator you know (00:09:24) the best accelerator in the world um (00:09:26) wrote this uh essay about being (00:09:29) resourceful and he talks about the (00:09:31) qualities of founders and one quality is (00:09:34) relentlessly resourceful. So they they (00:09:37) they're able to find resources to (00:09:39) unblock themselves and be able to um you (00:09:44) know not hit a wall because I think a (00:09:46) lot of what entrepreneurship is is (00:09:48) finding all these walls and really (00:09:50) driving through them. (00:09:51) >> And the way to do that is so think of (00:09:53) think of it as like a video game. In a (00:09:55) video game, especially open world video (00:09:57) games, you're often running into these (00:09:59) problems where you don't know how to get (00:10:00) to the next level. A lot is creative (00:10:03) thinking kind of moving around and (00:10:04) finding the clues or and I think (00:10:07) entrepreneurship and you know building (00:10:09) software sort of similar to that. (00:10:11) >> Can you tell me what changed when you (00:10:12) decided to start your company? What was (00:10:15) this thought that was like I need to (00:10:17) build this? (00:10:18) >> So there's starting the project versus (00:10:20) starting the company. Starting the (00:10:21) project was so obvious to me programming (00:10:23) is hard. We need to make programming (00:10:25) easier. That's like a very technical (00:10:27) sort of view on things. Starting the (00:10:29) business was less obvious because I had (00:10:31) worked at startups. I knew how painful (00:10:34) they are. Like so I worked at code (00:10:35) academy for example when I first came to (00:10:37) the states. It was based on the open (00:10:39) source version of of replet but I saw (00:10:42) how difficult it was and it was really (00:10:44) painful a lot of hard work. So replet (00:10:47) was still a side project and started (00:10:48) growing and we really I didn't want to (00:10:51) start it into a business because and I (00:10:53) actually tried to sell it to to Facebook (00:10:55) where I was working back then. I wanted (00:10:57) to stay there. I was very comfortable (00:10:58) and happy there. (00:10:59) >> It's a scary feeling to leave your job (00:11:02) and and go heads down. (00:11:04) >> But it was kind of derisked because we (00:11:06) had a platform that people really loved. (00:11:07) We had 100,000 users plus a month. So (00:11:10) after a lot of deliberation (00:11:12) >> and and really trying to think what (00:11:14) matters to you, what what what creates (00:11:16) meaning in your life and uh you know (00:11:18) serving our customers and really (00:11:20) achieving our mission, helping people (00:11:22) create businesses and all of that felt (00:11:24) very very important. So I think it's (00:11:25) about meaning partly (00:11:27) >> and it feels like with this tool there (00:11:30) are two problems in entrepreneurship (00:11:31) used to be three like coming up with (00:11:33) idea building product and marketing. Now (00:11:36) you're left with coming up with ideas (00:11:37) and marketing right (00:11:38) >> what do you think is going to happen (00:11:39) when everyone's building an app? (00:11:41) >> Yeah. So when everyone's building an app (00:11:44) when that is it's still a skill right (00:11:46) it's still like like we talked about (00:11:47) still a skill that you can you can (00:11:49) develop and I think grit is very (00:11:50) important. So resourcefulness, (00:11:53) grit, like not quitting, like not (00:11:55) quitting after six hours, like spending (00:11:56) another day or two on it at least. I (00:11:59) think domain knowledge is very (00:12:01) important. So if you're you have (00:12:05) excellent domain knowledge in on YouTube (00:12:07) and so you need to imbue you need to (00:12:10) give that domain knowledge into the (00:12:12) agent. You need to prompt in a certain (00:12:14) way so that you're downloading your (00:12:16) domain knowledge and that is your (00:12:17) competitive advantage. But at the same (00:12:19) time, what OpenAI models are training on (00:12:23) on what I know and then they're so much (00:12:26) better at defining what a good YouTube (00:12:28) video is. (00:12:28) >> I think you still have tacet knowledge (00:12:30) that is not necessarily expressed in all (00:12:33) your videos and all the content out (00:12:35) there. (00:12:36) um that CFO at the VC firm has a lot of (00:12:40) knowledge and skills he built up over (00:12:43) the years that he can make into an app (00:12:46) that you can't find on blogs and you (00:12:49) can't find online. And so I think every (00:12:52) one of us as we go through life, we (00:12:54) build up a lot of experiences that LLMs (00:12:58) do not get to experience because they're (00:13:00) not embodied. But do you think there (00:13:02) will ever be time when AI sees the (00:13:06) problem, comes up with a solution, codes (00:13:08) the app, doesn't need a human? (00:13:10) >> You know, I I I might be a bit different (00:13:13) in the Silicon Valley context in that I (00:13:15) am uh quite skeptical about the AGI (00:13:17) vision. I think we can build extremely (00:13:20) competent agents, but you would always (00:13:22) need the human as a driver. uh because I (00:13:26) think that the way large language models (00:13:29) work is they train on the entire corpus (00:13:32) of um texts on the internet, text and (00:13:35) books, all of that stuff that is text of (00:13:38) the past of what has happened, right? (00:13:41) What people have put in. But can it come (00:13:44) up with novel ideas, creative ideas, (00:13:47) ideas based on what's changing in the (00:13:49) world right now? Because they're not (00:13:51) always learning. They're not (00:13:52) continuously learning. They're within (00:13:55) this closed box that is their training (00:13:57) corpus, right? It's almost like a (00:14:00) library in a very way. It is a it is a (00:14:03) li a library has a lot of ideas and it (00:14:06) can remix and mix and match ideas. But a (00:14:10) net new idea is something that I think (00:14:12) still humans have have a special place. (00:14:16) >> I love your answer. But so the fear (00:14:18) inside me wants to push back by saying (00:14:20) that you know all the ideas already (00:14:23) exist in the world like when we're (00:14:24) talking about new movies or new books (00:14:26) they take a pre-existing idea but like (00:14:28) changing characters changing (00:14:30) circumstances (00:14:30) >> there's always a novel idea so think (00:14:32) about Bitcoin Bitcoin are you like in (00:14:35) crypto or Bitcoin a little bit. (00:14:36) >> Yeah. So, Bitcoin it was based on a (00:14:40) history of 20 years people trying to (00:14:42) build digital uh cash. Uh and in the (00:14:46) references it's referencing hash proof (00:14:48) of work was existing uh proof of work. (00:14:51) The idea that uh you know the machine is (00:14:54) solving these cryptographic problems in (00:14:57) order to secure the the Bitcoin network (00:15:00) was actually originally invented to (00:15:02) fight spam. So there was the spam (00:15:04) problem of the emails. We don't have (00:15:06) spam filters or AI. And so when I'm (00:15:08) sending you an email, I solve a (00:15:10) cryptographic problem that is expensive (00:15:12) to show you that I'm not a spam uh (00:15:15) agent. Right. So Satoshi Nakamoto took (00:15:19) all these ideas and you're right (00:15:21) existing ideas and put them in a in a (00:15:23) new package, but he added a a novel idea (00:15:26) which is how to solve the double spend (00:15:28) problem. And this is the blockchain, (00:15:30) right? So I think it's easy to think (00:15:33) about there's no nothing new under the (00:15:36) sun. I think that's the expression. But (00:15:38) I think if you look carefully, if you (00:15:40) look at what Einstein did, right? Like (00:15:43) uh with his theories, um there's always (00:15:46) one novel insight, one really strong (00:15:49) novel insight. It's almost like this (00:15:50) divine intervention is that there's (00:15:52) something spiritual about having a (00:15:55) really novel idea. And I'm sure you've (00:15:57) experienced it in the past. uh that I (00:16:00) think is is fundamentally human. I don't (00:16:02) know where it comes from. (00:16:02) >> Yeah. And also this feeling that you're (00:16:04) the one to bring it to the world. I feel (00:16:06) like AI likes that. Okay. Let's see. (00:16:09) Let's see what's going on here. Uh (00:16:13) should I go to overview and try again? (00:16:15) No. (00:16:15) >> So I hit redeploy. It's asking you to (00:16:17) redeploy. So it'll take a second to (00:16:20) deploy. I'd be curious. So fixed session (00:16:22) decentralization problem. Okay. while (00:16:25) it's deploying. Um, you said a gap (00:16:27) between a Republic user and a senior (00:16:29) Google engineer will disappear in two (00:16:31) years. Should people still learn how to (00:16:34) code or what's going to happen to the (00:16:35) engineering job? (00:16:36) >> I think the engineering job will (00:16:37) continue to exist especially in um very (00:16:40) domain specific areas is basically what (00:16:43) we're talking about. There are a lot of (00:16:44) things that are not very well (00:16:45) represented in the data. If you're a (00:16:48) platform engineer at Google dealing with (00:16:51) a billion users, there are knowledge and (00:16:53) things that you understand and have (00:16:55) learned on the job that LLMs do not know (00:16:58) because no one's written them anywhere. (00:17:00) It's this task of knowledge. (00:17:01) >> And so I think those engineers will (00:17:03) continue to exist. If you're an engineer (00:17:06) at NASA and you're building fault uh (00:17:09) tolerant systems, if you're building (00:17:11) provable systems, (00:17:13) uh like I don't want my Tesla autopilot (00:17:16) to be vibe coded, right? Like there are (00:17:18) a lot of life and death systems that we (00:17:21) want engineers that are very low-level (00:17:24) and very almost mathematical about it to (00:17:27) to exist. So there's a lot of situations (00:17:29) in which engineers will continue to (00:17:31) exist. But if you're a product builder, (00:17:34) I would say just go ahead and build the (00:17:36) product. Like don't wait like if you (00:17:38) need to learn coding along the way, (00:17:40) learn it. But but your mission is to (00:17:44) build the thing. So I would start by (00:17:45) building and and like we said, being (00:17:47) resourceful along the way goes a long (00:17:49) way. (00:17:49) >> So what should engineers who are (00:17:51) building apps like this do now? (00:17:54) >> Um let's see. Where is it? (00:17:56) >> There we go. (00:17:56) >> Oh, there we go. Okay. Sign in. Okay. (00:17:59) like (00:18:00) >> we had a new issue (00:18:01) >> and it's like (00:18:02) >> Google authentication not but that's (00:18:04) that's how software engineers work (00:18:06) >> okay this is this iteration (00:18:08) >> got it so you think (00:18:10) >> solve a problem (00:18:11) >> and then and then get a new problem but (00:18:14) >> do you think it will ever get get to to (00:18:15) the stage where I don't have to do this (00:18:17) >> yes (00:18:17) >> because it could actually run (00:18:19) >> everything and test everything (00:18:21) >> yes so we uh (00:18:23) >> is that we're working on (00:18:25) >> could you talk about it (00:18:26) >> or or not yet (00:18:28) >> uh usually Not yet. But uh let me give (00:18:31) you some some hints. Okay. Um (00:18:33) >> every vibe coding platform today (00:18:35) automates generation of code and all of (00:18:38) that stuff which is great but leaves a (00:18:40) job for you that is actually very (00:18:44) routine and uncreative and annoying (00:18:46) which is quality assurance and testing. (00:18:48) >> Exactly. QA. (00:18:49) >> So we're solving that. (00:18:50) >> Okay. All right. You're launching (00:18:52) something right? Is there a lot of (00:18:54) pressure on the team? How are they (00:18:55) handling? (00:18:56) >> Yeah. So, we have an offsite next week (00:18:58) where we're going to be going to uh LA (00:19:00) uh on the beach and u at least that's (00:19:03) like a sort of a relaxing environment (00:19:06) and we're going to be going really deep (00:19:08) and working really hard and then the (00:19:09) week after that we're going to be coming (00:19:11) here. It's called sprint week and we do (00:19:13) it before every launch and people (00:19:15) typically work 14-hour days non-stop. (00:19:17) >> You too? (00:19:18) >> Yeah. (00:19:18) >> Your wife too? (00:19:19) >> Yes. Yes. (00:19:20) >> Who takes who takes? Okay. There's a (00:19:22) question coming from all the (00:19:23) entrepreneurs have kids like how do you (00:19:25) balance this? So we, you know, there's (00:19:27) people talk about work life balance. We (00:19:30) we talk about work life harmony or work (00:19:32) life integration. And so, for example, (00:19:35) like we'll have uh the nanny bring the (00:19:38) kids to the office. We'll have (00:19:40) >> so they can vibe code. (00:19:42) >> They can v code. I sometimes sit down (00:19:44) with my kid and like do a little bit of (00:19:46) coding, but so we can see them, right? (00:19:48) Like so. (00:19:49) >> And next week for the trip, uh, everyone (00:19:52) here could bring their families as well. (00:19:55) So, you know, Replet not being super (00:19:57) young a startup, we're not that young. (00:19:59) There's a lot of people with families. (00:20:00) So, we try to like create that (00:20:02) integration. (00:20:03) >> Do you think it works (00:20:04) >> honestly or do you feel like you're (00:20:06) missing out on your kids? (00:20:08) >> I don't feel like I'm missing out. Like, (00:20:09) honestly, I don't feel like I'm missing (00:20:10) out. So, you know, just having a bit of (00:20:13) freedom in your schedule. (00:20:15) >> So, I spend like mornings with them. (00:20:17) Like I wake up at 7, 8, and I get to the (00:20:20) office by 10:00, right? But I, you know, (00:20:21) stay late here. We have dinners uh here, (00:20:24) but you know, I have these two hours in (00:20:26) the morning at least. Sometimes I'm (00:20:27) there for dinner as well. And on the (00:20:29) weekend, I'll work a little bit like (00:20:31) Sunday afternoon. Uh I'm always on my (00:20:34) phone working obviously and responding (00:20:36) and phone calls and things like that, (00:20:37) but I feel like I'm present. (00:20:38) >> What about your wife? Similar schedule, (00:20:41) >> I think. Yeah. No, she's she has a less (00:20:42) intense of a schedule. I think CEO (00:20:44) schedule's like a little more um (00:20:47) intense. And I think I think this is (00:20:50) this is you probably can relate to that. (00:20:52) I think being a mom there's like a more (00:20:53) sense of guilt that I don't really (00:20:56) >> all the time because they're back from (00:20:58) school right now, (00:20:59) >> right? (00:21:00) >> Yeah. But I also would love to do this. (00:21:01) >> Yeah. Yeah. So I I I don't know how what (00:21:04) to speak to that but I feel like (00:21:06) >> all things considered we feel fairly (00:21:09) involved in their lives. (00:21:10) >> That's good. Talk to me about your (00:21:12) mindset. Like you're coming up with all (00:21:13) these new things, but at the same time (00:21:15) we have lovable cursor. We have all of (00:21:17) these tools. We're going to let you do (00:21:20) the same thing. How do you see yourself (00:21:22) being different and how do you survive (00:21:24) mentally as an entrepreneur? (00:21:25) >> So, we're going to be the first to do (00:21:27) what I just talked about and it's (00:21:28) launching in September. (00:21:29) >> Yep. (00:21:30) >> And last year we're the first agent on (00:21:32) the market and so we're always three, (00:21:35) six, I think this feature that we're (00:21:37) building is probably going to be year (00:21:38) ahead of anyone else. And that's because (00:21:40) it's built on 10 years of innovation, (00:21:43) infrastructure innovation. like every (00:21:47) app that you're building, every (00:21:48) workspace is backed by um a cloud (00:21:52) virtual machine built on a file system (00:21:55) that we innovated. We even patched the (00:21:57) Linux kernel to to to make things work (00:22:00) for Replet. And so all that (00:22:02) infrastructure allows us to always be (00:22:05) ahead. We'll come up with an idea. (00:22:07) Sometimes it takes two three months to (00:22:08) build. If you want to build the idea (00:22:10) from scratch, it'll take you two three (00:22:12) years to build. And so I think Replet (00:22:14) will actually start to diverge pretty (00:22:16) soon. All these applications kind of (00:22:18) look the same because they all generate (00:22:19) a website. But um when you talk to (00:22:22) developers that are using Replet often (00:22:24) moved from the other uh platforms uh (00:22:28) they're saying that yeah I mean these (00:22:30) tools got me a like a pretty website (00:22:32) pretty quickly but you know a month in (00:22:35) I'm just blocked. I can't it can't (00:22:38) manage my database. I want a place to do (00:22:41) uh to store my files. Replet ships with (00:22:44) a database has an object storage (00:22:46) component has authentication component (00:22:49) like I know you're trying to set up (00:22:50) Google authentication but you can also (00:22:52) ask the agent implement replet (00:22:54) authentication has a built-in (00:22:55) authentication system. We built up this (00:22:58) massive amount of infrastructure and (00:23:00) within the next 6 months to 12 months I (00:23:03) think it'll be really obvious how Replet (00:23:05) differentiates. (00:23:06) >> Have you seen any big mistakes that (00:23:07) people make when building something with (00:23:09) Replet? cuz I know a lot of (00:23:10) entrepreneurs are looking and they're (00:23:11) probably like okay I'll build an app but (00:23:13) what's next like any tips for marketing (00:23:15) or any lessons that you've seen along (00:23:18) the way first of all what we talked (00:23:19) about with prompting (00:23:21) overcommunicate overcommunicate be (00:23:24) resourceful replet environment gives you (00:23:26) a lot of tools like the logs and things (00:23:28) like that try to overcommunicate with (00:23:30) the agent I think that's the first tip (00:23:32) that I would give even if you don't have (00:23:35) you don't want to learn prompt (00:23:37) engineering just just be (00:23:38) overcommunicative And then on uh on (00:23:40) marketing uh I think that's the next big (00:23:44) bottleneck for entrepreneurs right let's (00:23:46) say building a product becomes easier (00:23:49) and easier and easier how do you go get (00:23:51) it to market and communicate your value (00:23:53) proposition all of that (00:23:54) >> exactly and like why would someone buy (00:23:57) something like this if they can just go (00:23:58) to replet and v code the same thing (00:24:00) right (00:24:01) >> yeah uh (00:24:02) >> it's like an easy (00:24:03) >> there's definitely marketing is a is a (00:24:05) big part of the answer but also like I (00:24:07) said domain knowledge that you have like (00:24:09) just think about the things that you (00:24:11) know (00:24:13) deep in your heart of hearts that not (00:24:15) many people in the world know and the (00:24:18) other thing is like I said grit like (00:24:20) just just you know not quitting after 6 (00:24:22) hours is is very differentiating (00:24:25) actually most people just quit and so (00:24:28) just keep going and not quit like I've (00:24:30) been building this business for 10 years (00:24:32) before that back in Jordan I had this (00:24:36) idea when I was 22 and I started working (00:24:39) on it. Actually, there was an open (00:24:40) source project called Replet back in (00:24:42) 2010 (00:24:44) and I just didn't quit. I knew this was (00:24:46) kind of big. I I I still know it's going (00:24:49) to be a trillion dollar company at some (00:24:50) point in the future. Right now, we're (00:24:52) like a $3 billion company, which is (00:24:54) still huge. (00:24:54) >> It's amazing. Yeah. Congratulations. (00:24:56) >> Thank you. But it's just not not (00:24:58) quitting. I think a big part of it, I (00:24:59) mean, you know, people talk about it all (00:25:01) the time. Just show up every day. (00:25:03) >> Yeah. (00:25:03) >> Just showing up is a is a big (00:25:05) differentiator. Most people don't you (00:25:07) know it's things are hard and life is (00:25:09) full of easy things like you know you (00:25:13) can spend four hours on TikTok be (00:25:14) endlessly entertained right (00:25:17) >> um so that's one on marketing there are (00:25:20) a few few things I could say here one is (00:25:24) launch launch launch just keep launching (00:25:27) go launch (00:25:28) >> iterate or (00:25:29) >> even the same product (00:25:30) >> uh like m make another uh (00:25:32) >> another tweak (00:25:33) >> another tweak show it in a different (00:25:36) way. Iterate on your messaging. Do (00:25:38) another video. Uh try to reach out to (00:25:41) influencers to partner with them. (00:25:42) >> Podcast. (00:25:45) I launched it three or four different (00:25:46) times with different messaging and (00:25:48) different things. (00:25:49) >> Interesting. So the first three times (00:25:50) didn't work. (00:25:50) >> Uh first few times it didn't work. Uh I (00:25:52) think when we got on Hacker News the (00:25:54) first time, uh it was when I said uh you (00:25:57) can try all these different languages (00:25:58) and I listed the languages. Try Python, (00:26:01) Ruby. That was before AI. Uh, and that (00:26:04) was a hit because of the because of the (00:26:05) title change. And so again, it's grit, (00:26:07) relentless resourcefulness, and just (00:26:10) iteration, iteration, iteration. (00:26:12) >> Okay. What are your top three favorite (00:26:14) AI apps? (00:26:15) >> Of course, (00:26:17) >> yeah, we'll leave that aside. (00:26:19) >> Uh, Plexity, you know, I just like going (00:26:22) to Google and spending five minutes (00:26:24) clicking on links. I can just get it (00:26:26) really fast with Plexity. I like to do (00:26:28) deep research with the perplexity. So (00:26:31) that's something I can't live without (00:26:32) without uh chat. Um I go to perplexity (00:26:36) when it's like more research. I want (00:26:38) something from the web. Chatip when I'm (00:26:40) like brainstorming and and Claude and (00:26:42) the other ones too. Um and let's see (00:26:45) >> give me something specific like I know (00:26:47) for a specific problem if you use (00:26:49) anything maybe build your own AI agents (00:26:51) >> like naming uh like naming products. (00:26:53) >> Yeah. Nameless or what do you use? (00:26:55) >> No I I I'm using chat. (00:26:56) >> Oh are you using I like to prompt right. (00:26:59) I like to prompt it and I like to kind (00:27:00) of start different. (00:27:01) >> This is how you you kind of build (00:27:02) differently because you like prompting. (00:27:04) Uh so you just use chat GP for basically (00:27:07) anything because you can prompt it. (00:27:08) >> No, I I sometimes build rapid apps as (00:27:11) well for certain things. (00:27:12) >> Oh, what did you what did you build for (00:27:13) yourself? (00:27:14) >> Well, recently uh do you know Kindle (00:27:16) Scribe? (00:27:18) >> So you know Amazon's new notebook thing. (00:27:20) It has a web browser and sometimes I (00:27:22) don't want to use my phone. My phone's (00:27:23) somewhere else. I'm on my Kindle (00:27:24) reading. I want to look something up. (00:27:26) >> I try to open chat. It actually doesn't (00:27:28) render because the e in browser kind of (00:27:30) is like a very old school browser. So I (00:27:32) built a chatgbt that doesn't have any (00:27:34) JavaScript. (00:27:35) >> Uh that's very (00:27:37) >> and it took like an hour with replet to (00:27:40) to build. So I'm often spinning up these (00:27:42) small tools. Anytime I find that's the (00:27:44) thing about replet it becomes addictive (00:27:46) once you know that you can make certain (00:27:48) pieces of software you'll immediately (00:27:50) see a problem. It's like oh that shaped (00:27:51) like a problem I could solve. (00:27:53) >> Yeah. You mentioned hard moments. Talk (00:27:55) to me about the hardest moment building (00:27:56) this. (00:27:57) >> Oh man. Um, I mean there's a bit of (00:28:00) recency bias to this, but last year we (00:28:02) did a we did a layoff and replet (00:28:04) culture, you know, you'll get it from (00:28:06) from working walking around is is very (00:28:08) positive. Uh, people really like each (00:28:10) other here. Uh, you know, you ask a lot (00:28:13) of people why you're here. Some will (00:28:14) tell you the technology, the mission. A (00:28:16) lot of people will tell you the people. (00:28:18) They really enjoy working with the with (00:28:20) the people. And so it was very (00:28:21) heartbreaking to have to cut the team (00:28:23) because the business wasn't doing well (00:28:25) before we launched Replet agents. We (00:28:27) were in this very awkward place, you (00:28:31) know, that's a marketing observation (00:28:33) where we weren't good enough for the (00:28:35) senior engineer and we weren't good (00:28:38) enough for people like you. It wasn't (00:28:40) easy enough. We're in the middle. So you (00:28:42) learn how to code, use Rep, but then you (00:28:44) graduate off of it. Uh and so that that (00:28:47) we had to add more platform features (00:28:49) like all the databases and things we (00:28:51) added but also we had to make it easier (00:28:53) so we can have access to a larger group (00:28:55) of people uh such as yourself. But at (00:28:58) the time we we had 130 employees were (00:29:00) burning money like crazy and we had to (00:29:04) lay off the team and we had just (00:29:06) actually come to this office and this (00:29:07) office is huge (00:29:09) because I was so optimistic about our (00:29:11) future. I knew that AI is going to be (00:29:13) really big. I knew we're building the (00:29:15) right thing, but we came here, we were (00:29:18) burning all this money, and then we just (00:29:21) had to do the layoff, and we we we cut (00:29:24) the team. And I think we we cut 30 40% (00:29:27) of the team. (00:29:29) >> But then a lot of people started leaving (00:29:30) because well, the office is like empty. (00:29:33) It was like really dark place. (00:29:34) >> Was that when you were making like 2 (00:29:36) million a year revenue? 2024. Last year. (00:29:39) >> Last year. Um and you you walk around (00:29:41) here, it's very gloomy. No, no one's (00:29:43) really happy. I used to come here. I was (00:29:45) like, I can't wait to go home. (00:29:47) >> Wow. (00:29:47) >> And I think anyone most people in in our (00:29:51) place at the time would just call it (00:29:52) quits and try to sell a company and do (00:29:54) something like that. But instead, the (00:29:56) people that are working on Asian, we (00:29:58) motivated them. We told them this is we (00:30:00) think this is the thing that's going to (00:30:01) work. And we told them, look, if this (00:30:02) doesn't work, (00:30:04) there's no future and we have to make (00:30:07) this work. And so the core team that was (00:30:09) working on an agent, everyone stayed and (00:30:12) worked 12 14 hour days. And um (00:30:16) >> and you got to 144 million. (00:30:18) >> Yes. In less than a year. (00:30:20) >> I heard this story that uh Peter Teal (00:30:22) passed on investing, but then you sent (00:30:24) him your graph. Did they ever reply? (00:30:26) >> He didn't reply. (00:30:27) >> Never. Peter Teal invested in our series (00:30:30) B round in 2021. But then I went to (00:30:33) pitch him in 2022 or 23 just before Chad (00:30:37) GPT. I was trying to tell him, hey, uh, (00:30:40) AI is very important. It's going to (00:30:41) change the nature of coding and (00:30:42) programming. And he said, you know, (00:30:44) Peter is very skeptical of buzzwords and (00:30:47) he's known to be a contrarian. So he (00:30:49) doesn't like anything that's popular. (00:30:51) And so he was like, when you're saying (00:30:53) AI, it's meaningless. It's almost like (00:30:55) saying computers. uh you know don't come (00:30:58) here with these buzzwords and he (00:31:01) basically said I was just engaging in (00:31:03) hype uh you know we had raised at a big (00:31:06) valuation I'm trying to justify that (00:31:08) valuation and the entire meeting I'm (00:31:10) trying to tell him hey just like look at (00:31:12) the demo and he wouldn't look at the (00:31:14) demo and then I remember four months (00:31:16) later uh I saw him on TV talking about (00:31:19) chat and saying oh it's actually a (00:31:20) fundamental innovation I was like I (00:31:22) tried to tell you (00:31:23) >> told you (00:31:24) >> um and um you know to his credit, he (00:31:27) changed his opinion and they in uh (00:31:29) Fanders Fund invested a big amount in (00:31:31) Cognition, which is another agent uh (00:31:34) coding company. Uh but I did send him an (00:31:36) email saying after they in invested in (00:31:38) Cognition and told told him, well, you (00:31:41) know, I have a lot of respect for you (00:31:42) and that conversation was was actually (00:31:44) very hard to take in because I I felt (00:31:46) like I was doing something wrong. Uh, (00:31:48) but I hope you can see that I at that (00:31:50) moment I saw the future where things (00:31:52) were headed. (00:31:53) >> And I'm glad you pushed further and (00:31:55) built whatever you built. Right. Because (00:31:57) it's so demotivating to hear things like (00:31:59) that from people who are super respected (00:32:01) and super smart. (00:32:02) >> Exactly. It's it's both demotivating but (00:32:04) can be motivating. It's about how you (00:32:05) frame it. Right. I've become one where (00:32:10) I'm actually more motivated to prove (00:32:14) people doubters wrong. (00:32:16) haters, doubters, (00:32:17) >> transform that energy into into changing (00:32:19) the world. (00:32:20) >> It's like there's nothing better than (00:32:23) having a lot of doubters and people (00:32:25) naysayers and actually proving them (00:32:28) wrong. It It's a great feeling and I (00:32:30) recommend it to everyone. I re I wish (00:32:32) for all the entrepreneurs (00:32:34) I wish that you're going to have a lot (00:32:35) of doubters because then when you (00:32:37) succeed that's when that's when the (00:32:40) feeling comes in. Yes. I mean if you if (00:32:42) you talk to a lot of entrepreneurs and (00:32:45) you say what is meaningful about your (00:32:47) life what and they say the money is cool (00:32:50) you know money does change your life and (00:32:51) and it allows you more time to and more (00:32:55) resources to build and do more things (00:32:57) but the things that a lot of (00:32:59) entrepreneur kind of come back to is who (00:33:02) they were what they were before they (00:33:05) created this thing they keep remembering (00:33:07) it was like the struggle everyone (00:33:09) doubted me but I persisted did and I (00:33:12) succeeded and that's the best feeling in (00:33:13) the world. (00:33:14) >> Exactly. Let's talk about the future. (00:33:16) You have two small kids and yesterday I (00:33:18) unlocked another AI fear for me because (00:33:20) I stumbled upon an article that said (00:33:22) that recent graduates just graduated (00:33:24) from college, they're struggling to find (00:33:26) their first jobs because there are 15% (00:33:28) less jobs, uh, entry- level jobs, but (00:33:30) 30% more applications. (00:33:32) >> Yeah. (00:33:33) >> I'm like, oh my goodness. So, I'm (00:33:34) worrying about myself and I also have to (00:33:36) worry about my kids. What do you think (00:33:38) about your kids' future? What are you (00:33:39) teaching them? Look, I I think there's (00:33:41) deep question about where we're headed (00:33:43) as society. What is Silicon Valley's um (00:33:47) responsibility? What is the (00:33:48) responsibility of the government? And I (00:33:51) think we can have hours discussion on (00:33:52) that. I don't think anyone has a good um (00:33:56) grasp on these problems. But locally, (00:33:58) when I'm thinking about my kids and what (00:34:01) we're going to do in the future, I think (00:34:03) about this idea of being more of a (00:34:05) polymath, right? Like if you think about (00:34:08) you know even before the industrial (00:34:10) revolution the people that are most (00:34:11) memorable uh like Leonardo da Vinci for (00:34:14) example did a lot of things was an (00:34:17) engineer was an artist was all sorts of (00:34:20) things right and the the kind of elite (00:34:23) education used to be about understanding (00:34:25) a lot of things about spending a lot of (00:34:27) time learning about every everything and (00:34:29) I think this is where the education (00:34:31) needs to be headed where the industrial (00:34:34) revolution created a world where humans (00:34:37) are treated like machines. If you think (00:34:39) about corporations or factories, (00:34:42) factories are one big machine and every (00:34:45) individual person kind of doing the (00:34:47) assembly of one thing is a part and I (00:34:51) think it was very dehumanizing and I (00:34:53) think we're going to go back to a to a (00:34:55) moment of time uh where there's a lot (00:34:58) more opportunities for entrepreneurship. (00:35:00) Even when you join a company, you're (00:35:04) you're going to be judged by how much of (00:35:07) a real business impact you're going to (00:35:09) have as opposed to how task oriented you (00:35:11) are. (00:35:11) >> But what what would you say to for (00:35:13) example girls, right? I have two girls (00:35:15) and uh the way a lot of girls are (00:35:17) brought up are like find a job that's (00:35:20) sustainable, find something long-term, (00:35:22) find something that's safe. And you're (00:35:25) saying that everyone is basically needs (00:35:27) to become a generalist and also an (00:35:28) entrepreneur to generate ideas. What (00:35:31) would you say to those women watching? (00:35:33) How do you change your mindset? (00:35:35) >> Well, my wife is a is as is my (00:35:37) co-founder. I think we both uh Haya and (00:35:40) I share background of being (00:35:43) misunderstood and being um uh you know (00:35:48) uh growing up in in a in a culture that (00:35:50) we didn't feel like really fit our way (00:35:53) of of thinking and we were different uh (00:35:56) and so maybe that shaped how how we are (00:35:59) and our attraction to entrepreneurship (00:36:02) and to trying to change how people do (00:36:04) certain things or change the world. I I (00:36:07) think there's a lot that, you know, (00:36:08) culture imbuss on on sort of uh uh (00:36:12) gender roles, but I don't think girls (00:36:14) are fundamentally like sort of (00:36:15) predisposed to to a certain type type of (00:36:18) job. And so I think a lot of it is about (00:36:21) upbringing and what we tell our kids and (00:36:23) how we how we educate them. And so I I (00:36:27) think my advice will apply to to both (00:36:30) genders. I think that teaching them to (00:36:32) be resourceful, you know, actually like (00:36:34) not hiding from them the fact that the (00:36:37) future is very uncertain. We're in a mon (00:36:39) moment of time that's very different. (00:36:41) It's very different from when I grew up (00:36:43) from my when your grandparents grew up (00:36:45) or their grandparents grew up. There was (00:36:47) a lot more certainty about the world (00:36:48) right now. The world is very uncertain (00:36:50) and the way you're going to have to (00:36:52) learn and the way you're going to have (00:36:53) to adapt to this world is going to be (00:36:54) very different. It's going to be very (00:36:55) difficult, but we're here for you and (00:36:57) we're going to try to figure it out (00:36:58) together. How far ahead you think is (00:37:01) time when a solopreneur is going to (00:37:03) build a billion dollar company? (00:37:04) >> Is it a billion dollar in revenue or is (00:37:06) a billion dollar valuation? (00:37:08) >> That's a valuation. (00:37:09) >> Valuation. So, so let's say (00:37:10) >> so 100 million in revenue 10x. (00:37:13) >> Yeah, I'll say 20x. So maybe a 50 (00:37:16) million in revenue. (00:37:18) >> Um I don't think it's that far like like (00:37:20) if someone like John Cheney was able to (00:37:22) build a 2 million 2 three million (00:37:24) revenue. (00:37:25) >> It's very niche B2B high tech. I feel (00:37:28) like a 50 million company is a BTOC (00:37:30) company (00:37:31) >> which requires data. (00:37:33) >> Why not? Like he he has these enterprise (00:37:35) seats. He go, you know, who goes into (00:37:37) his companies, you know, he he has a (00:37:40) platform with a lot of content that he's (00:37:41) selling. (00:37:43) >> I I could see it be a $50 million AR (00:37:46) business (00:37:46) >> in the next couple years. (00:37:48) >> In the next few years. Yeah. I don't see (00:37:50) why not. But again, it's not just about (00:37:51) the software. It's about the domain (00:37:52) knowledge that he has. (00:37:54) >> Yeah. I love it. Thank you so much. My (00:37:56) pleasure. It was amazing. (00:37:57) >> Thank you for coming.