How to Vibe-Code 20 Production Apps in 13 Days - w/ Avi

This guy, Avi Pilser literally made 20 apps in 13 days, all of them are production ready, some of them are monetized, and a few of them are actually profitable. Today, we break down all barriers in terms of anybody that's trying to build production ready apps, even without having any prior coding experience. Avi also demos his latest project System Zero, which is a one-stop shop that allows you to do pretty much anything that you can imagine doing, using an AI, be it running agents autonomously for all of your various businesses, all the way to pretty much anything you can imagine. So without any further ado, let's jump in. Avi, welcome to the show. Hey, thank you very much to receive me.

Yeah, man. So obviously your post around all of that, all of the building public stuff that you've been doing and just your LinkedIn has been blowing up like crazy. I want to start with your experience with building and shipping 20 apps in 13 days. Obviously that just by itself sounds like just something unbelievable, but I want to tap into that and begin by asking you through that experience. Firstly, why was that something that you did? And then secondly, what were some of the biggest learnings that came out from those 13 days as you built those 20 apps?

A very good question. So basically my inspiration came from Peter Levels. Initially I was researching about what you can do with all those tools that we have at our disposal right now. And I thought, okay, first I asked myself what is the best question that I can ask myself. And I think that's also very important when you use those tools, all those backends like cloud code, OpenAI codecs, Gemini, CLI, etc, etc. Because everything is so easy to execute.

The cost of execution has gone down to zero basically in terms of money and in terms of time. So it's very important to ask first the right questions because it comes down to this now. It's about the meta thinking about thinking about how to formulate the best context, the best prompts, the best thinking literally because if you think better, you can prompt better. And if you prompt better, you get better results. And that's a little bit started like that, but when you take that idea of thinking better and you amplify it and then you build upon layer, upon layer. And so you quickly get an exponential work done. And so that all started like that. I asked a few questions. This Peter Level thing is a guide that a few years ago started to implement many, many different products, different ideas iterating fast,

killing project, killing companies that are not working and focusing on the one that are working. And so I said, okay, this is one of the ideas. Let's try it. And so I built about 20 apps in 13 days and it's still growing because now I have built an automated system that spawns new apps. I actually stopped it because I'm focusing now my system on distribution. But I could literally do more than two per day. It's like I could do theoretically 10, 20 per day. But I limit it because I want to focus on the distribution.

And so that's pretty much it. I can show you later on if you want a few stuff. Yeah, we would love to take a look, but yeah, I really like what you said there. And I'm also, you know, like a big fan of levels IO as he's known popularly on Twitter. And I'm sure you would have seen him on the Lex Friedman podcast as well. That is one of my favorite podcast episodes ever. It's like he just, you can tell that he thinks five times faster than he talks. And so he trips himself up while he talks. So yeah, so he's a really cool guy. I do want to ask you, Avi, let's say somebody is just kind of vaguely familiar with the capabilities of cloud code, open AI codecs, and the Gemini CLI without complicating it too much. Which of these do you think is the most suitable for most average Joe's and James of the world?

Okay, I think for the most average person that is just starting off, I think you should use cloud code because it's more familiar with the openness of the human mind. I would say open AI codecs is more a narrow architect that will just code for you very well. I think that it has a lot of orchestration in the background similar to some orchestration that I have actually built for my own system that I use. But basically yeah, start with cloud code.

That's super helpful. And I think from there also a lot of folks in our audience must be really curious about your background, right? So you're you obviously are extremely tech savvy now, but can you quickly speak to your background in terms of how familiar you were with coding and these tools before you started dabbling in the vibe coding space? So is this something that you taught yourself or did you have a background? So just help set the stage for us here. Yeah, super super good. My background basically, I really want to start with the really beginning is that giant mark when I was when I finished my studies, I did some engineering studies. And and honestly, I wasn't very good. I didn't I even didn't really graduate in engineering because at some point, I be frugated, but I got it. I got a degree. I have a master's degree,

but my point is then I immediately went to being an artist. I became I when I was a kid, I learned the violin. So I just went, I moved to China for a few years. I became an electric violinist artist. I built my own show. And so in the very beginning, I had this artistic mindset and also a little bit business because I just you know, I I sold my show. It was a big electronic show that I sold. I was performing with a team, thousands of people in China, etc. For big brands, I was and so that just to give you a background of how I started. And then after that, I said, okay, you know, my teenage years, I have to start doing something that I also like, that can make, you know, and more serious and long term. So I started working in AI. I self-taught myself at

the time. It was back before deep learning, you know, neural networks were popular when they were unpopular. So it was a little bit even. You were stigmatized if you were doing deep learning and I self-taught myself on the internet, you know, there were very few courses. I talked it to myself and you know, it started to take off and I got my first job in the area. I worked in research in the academy, first in bioinformatics. Then I worked in Motorola. I was because of my knowledge in deep learning that nobody else had. I became a chief of innovation, deep learning, AI in Motorola solution in innovation center. I worked, I worked there for a few years and then I became co-founder of a few startups as a CTO. I self-taught myself Python and I used Python extensively, PyTorch, KRS and, you know, intensive flow, etc. Pretty much it. That's my

background. As a programmer, I'm more of an architect, engineer, researcher, AI than a pure coder and programmer. Got it. So naturally, I'm curious then, Avi, that suppose this was the time around which you were trying to teach your self-ai, right? Because when you did, we obviously didn't have chat GPT. We didn't have these LLMs that we have today. So if you were trying to teach your self-ai in right now in 2026, as there are so many people that don't come from tech but that are trying to teach themselves these tools, how would you recommend somebody can go about doing this? Would you still prefer YouTube and just like Coursera and like courses or would you just go to an LLM and then directly start building stuff, start doing stuff? Which way do you prefer more

or is it like a hybrid model of both of those ways? It's a good question. I think, yeah, now that my first instinct would be just to go to an LLM, first you need to pick the best one for you. So you need to research a little bit on YouTube. Find out what you prefer. Maybe try them out for some time and then just go to an LLM and ask questions. And you know, it's very easy to learn that way now. There are little tricks that I use that facilitate my life. It's true.

Because LLMs output a lot of texts and I build this little tool that I use every day all the time that is a text-to-speech that I put in my Mac OS. And I just with a shortcut, it's a free access to, I think it's Microsoft Edge TTS. I build that. It was probably my first app that I built with AI. And it reads for me. I'm very auditive person. And I also have this tool that I can speak instead of typing called web flow, something like that. Yeah.

Whisper flow, is it? Is it called? Yeah, I use that as well. Yeah. Right. But you know, to make my life easier, I like to assure it like if I'm talking to a person because reading makes me tired sometimes. So yeah, I can relate to that for sure. It helps to have, yeah, somebody read it while I'm going off making coffee or something. From all of the apps that you've built, I want to kind of get into that. So first question right off the bat is how do you get the ideas in order to build apps? Now, if you're like me or probably like most people, you probably have a bunch of ideas, right? But now obviously you can't go after each one of them because obviously it's like a finite space in terms of what you can build. So how do you go about going from idea to execution in

terms of choosing that this is worth my while? We should probably build this. So there are two things I do first is the idea come from me. Actually, some of my best ideas still come from me deciding and thinking about it. And the second way is the process that I in built into my AI, my system, that that goes through a very strict process of selection and measuring and gives a lot of ideas and select them. So the 15 apps, like actually I have more than 20 apps, but let's say I have 15 solo ventures, those have been selected by the algorithm, automatically, that runs on the loop.

I have five moonsh adventures, which is a direction that I ask the LLM to go to tackle problems that seem impossible today. But because of the exponential growth of technology, it's necessary to tackle impossible things because they're going to become possible very quickly. And they are like, through three autonomous apps that I build myself from my own mind that one is a marketing company. Another is an app that helps you run your business 24/7 autonomously. And the last one is basically the system called System Zero that I'm using to create all of this and which is System Zero is actually it's always on fully autonomous personal AI agent. And it's an orchestration system. And it's for one operator. And it's no human in the loop by default. Awesome. And I know we will be

taking a deep hard look at System Zero. So I'm really excited to do that. But before we do that, just a real quick detour on this one, which is that off these 20 apps, and you said there's more than 20 that you've built, are you able to speak to maybe one or two of your favorites that, and just like as an example, you can be like, here's what I thought of. Here's how I built of it, like built it. And then this is what it does currently. So because I'm very curious to know about more about, you know, some of these full-scale apps, production apps that you have live right now.

Yeah, so the first one that I really like is called Viral Epic at Viral Epic.net. And it's basically a nap that reminds me a little bit of the origin of Facebook when it started. It's basically a nap that helps you to discover anyone on the internet just by pasting its LinkedIn URL. And I use my System Zero as an analytics looping genius to find new vectors and dimensions about the life of that person. And it's focused to help that person to find what is its deepest desires and what are the best outcomes for this person in the world. And there's a little report that you can purchase for $5. And then there are other services that helps you to grow your business, to grow your life that cost a little more. And then there are also autonomous systems, autonomous services that you can purchase. Got it. Wow, that's really cool. Yeah, that's,

I would love to, you know, take a look at that. And maybe separately, I can link that as well, in case our listeners want to go check it out. But you could say something that really struck my attention, which is that you said you're kind of more focused on the distribution side of it now. So I know this is a question that comes up a lot, where it's become really easy to make apps, relatively easy at least, you know, not like easy in terms of absolute terms. So now the problem becomes how do you find users? So how do you approach distribution and how do you plan to get users to your all of these various like dozens of apps that you have.

So first of all, I want to say, virally peak is already be a profitable is already gathered attraction, customers, even 72 hours after it was released. And the way that like it's really some, the way I did without putting my system into it, is for now on social media, probably mostly LinkedIn, I do a few posts, and you know, surprisingly people, you know, connect and they go and see and see the products. So I got it. I also have a system that in system zero has also completely automatized LinkedIn, post creation to the post posting everything engagement there is. That's the first thing that I was doing. But now I'm going to the next step, is basically to really test if an app is working, doing some Google ads. And again, I'm also automating the process 100%. Actually, I started this yesterday because,

yeah, I have a new also a new branch on my app productions, which is called mass market. Oh, so it's not full adventure. It's not moonshot is mass market. These are new apps that are targeting. It's not only in the digital world, it's not only in the tech for tech people. It's for everyone. And I'm testing those with Google ads. So basically, I have a very small budget that I run for one week with a limited budget per day and per week and per month. And after one week, I look if there were some engagement in the platform, it's not necessarily need to have some purchases yet. But if if there is some traction, boom, I keep that app. And if there is not after one week, I removed every app. So I prune and I purge everything that is not making profit. Interesting.

So you basically seeing you look at real world signals to figure out which of your apps are continuously on live support versus which ones you just pulled the plug for. It's not like a guessing game for you. Is that right? Yeah, it's not a guessing game. It's based on data. It's based on the real real world. I'm never know you. You never know which app is going to become successful. I got a hint with my LinkedIn post social media. You can see because I have an automated system that gather email or captured emails or people. And I see that there are two, three apps among the 15 that are good. It's the viral epic that I told you this is pretty good. And yeah, I'm going to confirm that with the Google ads. Awesome. Yeah, man. So at this point, I'm just

extremely curious to look take a look at system zero. Maybe it would be great if you can maybe show us how it works in practice so that, you know, then maybe we can even talk about if you're willing to share kind of how you built it. What are the possibilities in the future? And you know, maybe like go so go go from there. So are you at all willing to kind of give us a taste in terms of showing us what system zero is and how it works? Yeah, of course. So let me share my screen.

Very simply, here is basically I'm going to show you a few things that I already built with system zero. Then I'm going to show you system zero. Perfect. Perfect. System zero, as you can see is here. And this is just the cleanse. And I will show you later. So first of all, it's my company ultra detect that is commercializing the systems during infrastructure. And we are about to put that to the public. It's a it's a very you know, it's for people. And it's also for companies enterprise. So you can go on to tech.com to to see a little bit behind the design. Then I want to show you my personal website, Avi Pelser, that is gathering all the apps that they have. And you can see here, viral epic here is fleet AI that they didn't talk about much.

But it's pretty interesting. It's like basically running your business full time for you. You can just try it. It's at fleetai.avipelser.com. You just you launch your business. It's extremely fast. Let's say you want to take a business that you have, you put the link, you click import, and you run agent cycle. And as you can see, it's starting a process where it's deploying agents, an engineer, a marketing person. Interesting. And it's building the plan in real time for you, as you can see. And so it's generating the website as well automatically. As you can see, generating a promotion on LinkedIn. You can see that directly you have your LinkedIn post. Cool. The website is here. You can see with login and with payments already completely integrated. You have also payment. And you can

in one of that, that's one of the product. You can see, you know, these are the generated report, etc. And then you can add your Twitter, you can add your API keys, your strike for payments. Once you add your API keys, people start to purchase. So like, how did system zero build this? Like, did you use system zero to build fleetai? Exactly. You got the points. System zero is an autonomous system that builds autonomous systems. So that's the point. So you can, yeah. So that's fleetai. I am just introducing this as well. And virally peak is, you know, simply this other platform that you can just put the LinkedIn of some person and purchase the report.

I'm going to show you an example of report of my report. Those report become published. You can delete them. And so you have a page where it's really going depth. And it's using system zero that I'm going to show you. It's a small outburst of system zero that is a small component that runs that process in loop. And it analyzes you really in depth, really gets into your heart and understand where you are and give you directions for your future. What do you want? How to achieve what you want in your life? And you see, it gives you, you want to be the person who names what nobody else can see, et cetera. And you have a report with deep analysis that you can purchase.

And then you have these other products that are connected exactly for you to develop who you are and to make you successful and in your business. And it's automated. And you can purchase them as well. So that's awesome. And I'm curious to try it for myself at some point. I'm very curious to see what it gives. Yeah, just go to fleetai.avipalset.com. Sorry. So viral epic. Virally pick that net and you can have your report generated in a few hours.

On the queue. And so these are the sort of ventures that I have. Each one, you know, ship and skip is become quite popular. Should you build that? You have an AI idea, you have a project, a business, you analyze, you put your idea here, it will give you maybe let's try with my website. I don't know if it will work. It will give you an analysis. Maybe it's going to go and I went on and actually you need to give a plan and a text because I don't think it would go on and look at abbipalset.com.

But it will be a note and we have plenty of other websites, probably not recommended to do what I do because it's best for me, not for anyone else. Just be yourself. And that's the most important thing that I feel like the first user is always us. So if you build something for yourself, that is at least a guarantee that even if no one in the world uses it, at least one person will use it. And that's good enough, right? Exactly. And here you can see one of my dashboard that for my all my ventures that putting them here, you can see the status has been launched, distribution, some of them are more advanced than others. You can see all the businesses, moonsha adventures that have not been launched yet, moonsha adventure are crazy businesses that I'm not distributing yet. Of course. Yeah. These are all the all the subscriptions of

you know, of all the apps, how many people have got their email captured on each app, etc, etc. And I have recently my mass markets that I have 17 pending businesses that I spread all over the world. And I just just approved one today in the morning and add the Google ad has been launched automatically. Okay. So how did you build all of this? I guess did system zero? I know it I system zero built it, but I guess what is the underlying architecture for system zero to be building all of this? Let me show you then. So I'm going to show you system zero system zero. As you can see, you are in your in it right now. So system zero is basically not like in the interface, you feel a little bit like you're on cloud code. And so you can talk to system zero. Okay, you have your chats and you can talk to system zero. The only

exception and only one of the big one of the exception of many exceptions is that it's agnostic to the model. So right now I'm using opus 4.6, backend is cloud code, but I also have codecs. I'm often using codecs, by the way. I have Jimin, ICL, I have open API, a rock API, many other backends. So it's completely agnostic that way. You don't depend on any, you know, you don't want to depend on any backend. That's my philosophy. So I have my system zero. And at first glance, it looks like, you know, cloud code or codecs. But then you would see that it becomes much more than that. Because when you switch to the task tab and you start to realize that in the background, this system that you actually talk to, like cloud code and you make tasks is actually fully

automated, running 24/7, and have many different tasks that are running for you. And here you can see example, I have 100, 135 active tasks, just for my core module. And I have 60 tasks out of 62 that are running just for my modules. And you can see, for example, all of all of those are running autonomously 24/7. There are frequency, et cetera. I have some for LinkedIn. If I just output those that are running for LinkedIn automation, you see all of this I have for I don't know email, maybe exactly you type email, I have email nurture engine, validation, monitor localization, factory influencer affiliate, those are all tasks that contain the word email.

I have my daily email check. So it's answering my email autonomously. All those are tasks that run 24/7. And this is just the beginning because then once you realize that when you go to the intelligence tab, you realize that actually all of this is not only it runs autonomously, completely autonomously without your intervention, but it's self-improved. So yeah, so it's self-improved. So what you're going to realize is that I have my skills that are selected. As you can see, for self-improvement, and they are all rated. And you can see here, they have been running all over the repo autonomously, because what I didn't tell you is that I have a system. And actually, that's the first thing that I did to catch this exponential growth. I have a system

of check. So you, when you prompt the LLM naturally, if you, you need to be a genius in prompting, basically, top in the world in order for the LLM not to make errors and to do the error from the first time. So I have a system of check that I implement autonomously. Every time I have a prompt, it checks the output and it validates. So it runs a second pass autonomously to fix all the kind of problems that I have in the answer. So I don't need to do it myself. I don't even need to read almost what's the output of the LLM to check if it's correct or not. It's doing it for me, okay? Now I'm building new features. So because of that, you know, I'm writing into the LLM, but technically the goal is to complete autonomy in the self-improvement. Here you can see the

metrics of the improvement rates right now for today is +0.32%. So it improves. So you can see in which area it improves. For example, in detection, can the system see its own problems? So you have 87.5% of all the problems that have been self-detected by the system. So it's not yet 100%. That's why I'm keeping building it. But it's already very good. Here, for example, you have, can the system fix what it detects? So it can detect that 87%, it can fix that 80% autonomously. Does the system prevent recurrence? You see, we have some progress to make here. It gets 93% of lessons, et cetera, et cetera. So you can see that the system is self-improving and we are measuring this. And if you purchase the subscription, it's soon we're going to put this online, by the way, for everybody to use. If you get the subscription and you're going to have

access to this, personalize to your repo. And it's very easy to use. It's just connects to your or your stack already, either by API course, or you can use the platform on the cloud and have access to the entire dashboard and see how you can self-improve your repo by yourself. It's personalized to you. And it's not only a rate, every loop, everything that happens at 24/7, but it's select them with time. So every skills that I have here, you can see they are rated and some of them are downgraded and retired because they are not performing well enough.

So this is basically the system zero that I use every day to build that. And I'm soon putting it in the hands of everybody. And it's going to be on system zero for now, system zero.avipilcer.com. But it's going so fast that I'm prioritizing not the domain name, but I'm prioritizing the use for people. Totally man. So I'm like just blown away by how cool that is. One quick question on that. So you mentioned that it integrates with your repo automatically.

So like a note on that is do. So I know that you had like a bunch of recurring schedule sites, so to speak. So can it run both? So let's suppose I have a company, right? And there needs to obviously there's a website. Maybe I have an automated flow to write SEO blogs for my website with a certain rule. So I teach it all of that there's a skill about SEO blog writing. So all of that is configured. It runs maybe once a week. Automatically writes a blog pushes it to my website separately unrelated to that website for my own personal brand or whatever use case.

I'm trying to do like LinkedIn carousel making, let's say, so I have a skill that takes in an idea, automatically generates an a PDF carousel for LinkedIn, and it automatically uploads it. So these two things are not related to each other, but can they both still live on system zero and can it still serve as the central hub for me to manage all of these activities that I have to do for across my businesses, across various different things. Totally. And by the way, you know, if you're talking about that, I want to show you one of the modules that we are selling with system zero as well. That is the LinkedIn auto post and auto generation is a specific dashboard.

And it's so to answer your question first, you can actually it's like your personal repo. You can put different modules. One could be the SEO modules that will do everything about SEO. One could be the LinkedIn and they are together in the same repo and system zero will manage both of them together. And you could also have two system zeros. You can subscribe to two systems zero if you want to have more attention to each one to and you can subscribe to two systems zero and each one will work exclusively on those. And you can scale. It's not only two projects that you can have. You can have SEO, you can have Google ads, you can have marketing, you can have content right right.

Here I have, for example, LinkedIn and I have, I don't know, about a 188 posts that are already pre-prepared with carousel. Everything is done automatically for you. You can see here, for example, you're talking about carousel for LinkedIn. Everything is done for you. The screenshot, the videos that are all taken from the web automatically, from the best sources, most updated. And you also have the Twitter text that is created as well. And not only that, but you can also edit your posts here for LinkedIn. I don't know, you can hold it. You can hold kind of editing easily.

So it looks better. And so yeah, so basically you have, you can add up as many as many projects. You can see, this is another project I have. This is another project. All my businesses right now, they run on one system, on one. But I'm thinking to duplicate because I want to give more focus to some very specific project and to work full on this right now. Yeah. So again, that is all so impressive. I'm just like I said, blown away. It's almost as if having been somebody that has used Open Club, which maybe I don't know if you've used that as well.

But this is like a really smart, very user experience focused Open Club. It's with just much more insight into how it works, what it does, like just a much, much better or like more usable version, almost of Open Club. I don't know if that analogy makes sense. Would you agree kind of is that was that kind of the idea you had? Yeah, for sure. Thanks for the feedback that you have. The best feedback is from the other people. But yeah, I agree. And the system is basically proven and tested on more than eight weeks of full use that I mean, like I've been using it for myself. So basically it's not like I came up with a system that I download to my computer. I don't know how it works and stuff. It's built for the person and it's not based on Open Club and it's

not based on any frameworks. It's actually framework agnostic. And the reason is because at some point, you don't want to depend on the framework. You want to be able to use all of them so that you always above those frameworks. And you're not like you know, depending on one company. So it could take theoretically, it could even incorporate Open Club inside it. It was this in mind. So to discover any system that is put it in, you could use it in Windows, Mac, Linux, it ought to discover the tools that he has, everything.

So then from there, I guess, naturally, we, I'm sure a lot of our listeners are now wondering about the big elephant in the room these days, right, which is just usage slash API costs and such. So with this amount of heavy usage, if I were to guess, especially seeing how fast everything is how slick it is, how quickly it just does the thing you ask it to do, it sounds like this would be pretty expensive to run. So is it that is that the case? And I guess how have you thought about optimizing for not having, you know, like a $300 bill every time I run it after like a whole day's work? So to answer your question, there are different, different points in your question.

First, the first thing is you need to understand you need to put this in perspective. If you're generating $10,000 a month because of your business, it doesn't, it doesn't matter if you're spending a few hundred bucks in terms of compute. But that's the first thing. But not to really answer your question. We are running a very strong optimizations in system zero to minimize the cost from the LMS. And so we are still looking at the way we're going to put this out for the customer, if it's going to be a revenue share, because we believe in our system and there might be a little cost of functioning, you can choose what the options you want, if you want full boost, or if you want just the minimum to maintain. But it's fully optimized for cost, trust me.

And we're still working on more optimization and reducing costs right now. Until we're working on this every day. But yeah, yeah, it's for sure, you know. But of course, it's going to be affordable for sure. And if you make money, you will pay zero, because it's true. It will pay for itself. Definitely. And I also agree that if you're already thinking about it, right, once you have users, it's only going to get better from that, right? Because you'll only be able to learn from user behavior and optimize it even more, so that even if there are certain small gaps that exist right now, I'm sure those will go away once more and more users get onboarded. For those in our audience that are more technically oriented that are just obviously trying to figure out how like what goes

behind building something like this, I realize that we're now entering into like IP territory, and I totally understand if there's some things that you would not like to mention. But I would still have to ask on the behalf of those listeners, how did you go about building this? What can you share about, you know, the internal tool architecture, how it all connects together? And it does the amazing job that it's doing currently. So to be honest with you, I'm not so afraid about giving more insights because I've seen how the world has started to evolve now so quickly and so exponentially, I wouldn't have said the same thing about two weeks ago, but when I say two weeks ago, it's almost for someone who's not connected to the world of AI, two weeks is nothing. But for us, like two years, right? So it's literally two weeks ago for me,

it was like literally two years ago, almost, right? I agree, it does feel like that, for sure. And so I've seen things are growing so fast around that in two weeks, which is in two years or 10 years now, because it's accelerating, all my secrets will be anyways. And also I've seen I've seen some open source project that definitely all the things that I use, you can't find online, you can't find an open source, you can't find because I use stuff that it's full of innovations. And you remember, I told you about I'm a violinist when I started being a violinist, it's also come from the mindset. It's not only if there is an advice I can give, it's not only about tricks and tips, it's about your mindset, the way you ask questions. Remember when I said that it's the way you ask questions because

everything is free, it's free time and it's free money, literally almost, and it's going to become cheaper and more affordable with time. So not in terms of the price, but we will be able to do much more intelligence work for the same price, right? And so if you think about it, even if you you pay the full subscription to Cloud Code or to Codex $200 a month, even if you have two, even if you have three of them, it's still nothing compared to having a business that generate, like if you can take advantage the right way, if you can put together a system that take advantage the right way and save you so much time. And also does it well, it's literally nothing compared to what you can make. And that's the solution I propose. And that's it. And the mindset,

if you have the mindset, if you want to automate everything and you say to yourself, I'm going to free up time and I'm going to do it, I'm going to automate everything, but I'm going to do it well. We have built a very strong experience and tree barrier because we have been running those systems internally for weeks. So this is really important because I have tried before that five open source system that I started from scratch. None of them worked. Because when you try to engineer that from scratch, it doesn't work. It's a holistic system. So you need to look at it in a holistic way. It is a little maybe a little tip of advice that I can give for programmers that usually have a very atomistic mind, very focused on A+B=C. And they are not like

the world, especially the world of living things, which is we're starting to witness with system zero with my system and other systems that are emerging. They are starting to become living creatures, living digital beings. And those living creatures, they are not rectilinear, they are more like the branches of trees. They're always evolving at any given point of time. So if you have this kind of mindset that's much better in order to build things than having very narrow, straightforward rectilinear mindset that is very good for coding, but it's not necessarily good for AI. I learned that when I was doing neural network and AI, it's more holistic system, complex system, and system emergence things. And so if you think that way, I'm giving you a good

hint here. I think in terms of holistic things, branch of trees, rather than a face of a building world, everything is rectilinear. And it appears almost flat, but you don't realize how much is there behind it, hidden behind the surface. Exactly. Yeah, man, this is so cool. And I'm just so inspired by what you've been able to build here, because generally, it just feels like it is the only the one tool that you will need for all of your other AI needs, right? Like it's almost that one entry point that if you have this, everything else you just covered, like you don't even need to go anywhere else. And you can just do all of your stuff just using this one gateway.

So it's really cool. I'm just inspired and already my own brain is starting to think of so many cool use cases similar to this, some that are dissimilar, but in just really innovative ways in which we can already use the landscape that we have currently, which, as you said, we are nowhere close to mature. Like these things are changing, getting better every single day. I think just earlier today, I was looking at the announcement for the new cloud model. I forget the name. Maybe you saw it too. But after Opus, it's like fresh off the press at this point. They haven't even made it public yet. But anyway, I guess the final question here, Abby, before I let you go, what do you actually think happens to software engineering as a whole? Maybe a year from now to years from now? Do you

think as a profession, software engineering will still be your thing? Or will it just change into something else? What is your future prediction when it comes to the AI world? So when you say a year from now to years from now, it's at the rate of April 2026, but not at the rate of May 2026. At the rate of May 20, since it's evolving every four months, it's almost doubling every four months. It's going to be two months. What you say is it's going to happen in a year. If you consider the acceleration, it's going to happen in three months max. If it's not already happening.

But what I think about the job of developers, it's very hard to predict anything. But I think that every disruption has brought jobs to disappear and jobs to emerge. What I think is different this time, because of the AI, the AI is not a tool that you, the humans, that amplifies human. That's the big difference between AI and every other industrial revolution that happened in the past. What we have invented in the past, amplified humans, but human was still able to be on top.

This time, it's not only amplifying humans right now, but we are at the tipping point, and I'm talking in terms of weeks right now. We are in the singularity point where we are moving from AI is amplifying human to AI is completely autonomous, and there is no amplification there. I think there will still be a world where humans would be working, but they would be completely different meaning of work. I think we'll still be coexisting with some type of AI, but some AI will take off on their own and build their own species.

Yeah, but it could be not scary, but with the AI that we will be able to interact with, and yeah, we'll have them as a colleague, but we're working with them. We'll not be on top of them in my opinion, but we'll have all kinds of different jobs, jobs that we never have imagined before, jobs that have much bigger ambition. So, for example, we'll go on into the stars and we'll do bigger projects, we'll solve that, and we'll solve all diseases, and we will have much bigger ambition, and we'll be able to accomplish much more. In one day, we will be able to accomplish what we used to accomplish in one year, and it will be completely normal for us, but we'll still have jobs, and we will call that jobs, we'll probably not get money for it, because of full

abundance, we probably not need money, but there will be a progression in the period where we'll transition, and that's the period that might be difficult for everyone. Yeah, I love what you just said, because I think on the one spectrum, we have that, and then on the other, we also have, and this is already happening, where AI agents give work to humans, right? So, there are things that agents are not able to do right now. So, there is a website, I forget the name, I can link it when this goes out later, but it hires humans to do like in the real ground, in the, like go to this store, get like a picture, or keep this thing here. So, this is already happening, so I think it'll be interesting to see who works for who, in what capacity, and yeah, I think overall, I'm

definitely really excited for the future that is to come, because yeah, I don't think there has ever been a time in human history, where I would rather be alive than right now, at least that's how I feel. I agree with you. Well, this has been so cool, man, thank you so much for taking the time, yeah, really appreciate you showing us all of the cool stuff that you've been building, we will have to have you back in a bit, one System Zero is live, just to share feedback how things are going, what you have lined up for future projects or improvements, but this was so cool, I walk away from this conversation, so inspired, so motivated to continue to build and build fast, right, because that is the key here. So, thank you so, so much for taking the time, this was awesome. Thank you, Naman, thank you very much.