#2306 Polsia: AI Agent + Zero Employees = $10M Run Rate

0:00This guy and his agent built a company that does 10 million run rate. Pulsia is a company that will give you an AI agent that will build and manage your startup. What I'm trying to sell is like autonomous tool to build any idea you have. But there are controversies. People are saying that you have a big churn. What do you say to people who say that the site creates slop? Why is your name spelled AI slop backwards? It's a learning process. We heard loud and clear. We're trying to make the right thing. Still, I think his mission is admirable. Here in Silicon Valley, everyone's a founder, everyone's a tech founder.

0:32And then you go to the 99 % and they don't have access to that. They can't even inspire to that. They just see that on TV and they see that in the news. What Paulsia does is give them access to tools that can enable them to be a founder. I spent about an hour grilling him and he responded to every one of my tough questions. Let's get into it. You'll understand how this company works and how the future of zero human companies could look. Presented by Zapier, the AI automation company. What's your annual revenue right now? 10 million run rate. Okay. What do you say to people who say it's not annual recurring revenue?

1:06It's a run rate. You took your last 30 days, you multiplied it by 12. The people who are signing up are not really recurring. I mean, so to be clear, like there's a lot of different revenue streams on Pulseio. There's like a base subscription. Then there's like people buying add-ons, you know, like they may be buying ads, running ads that are like, recurring every day. They may be buying more tasks, either one-off or through subscription. They are buying custom domains that are yearly renewed, et cetera, et cetera. And so all those cash flows are all hitting false year bank account. And the idea with a run rate is that you take the past 30 days of revenue, either if you're recurring or you take one-off payments and you look at the last 30 days and you analyze it that's why that's why entropic for example also talks about an annual run rate and not recurring because they have subscriptions but they also have revenue coming from one-off like api consumption but then that volume is then annualized but i guess the big concern is that people are saying that you have a big churn do you what's the churn uh i mean the churn is like around 50 % on month one, which is, you know, I think for an autonomous product that like people, you know, use and, you know, so different people have different expectations on what it does.

2:41I think it's not the worst in the world. I think obviously it's like a novelty. So some people try it and then they decide not to stick. I think what I see most is that like people that like have, the less technical you are, the more you love a product like this right because usually you're not technical you have another job and then now this lets you try to become a founder right uh which is something very beautiful right it's like you know here in silicon valley everyone's a founder everyone's a tech founder everyone knows technology everyone's raising trying to trying to be able to raise capital or seed run or pre-seed round and then you go to the 99 percent and like you know they don't they don't have access to that they don't they can't even inspire to that they just see that on TV and they see that in the news.

3:27And so what Pulsia does is give them access to tools that can enable them to be a founder by having an AI that autonomously does things for them, sends them an email every day, that believes in their idea, that helps them navigate what it means to build a company. And so yeah, some people sign up and decide to churn because they're like, oh, I don't know, I didn't like the experience and some people stick. I think that's any product. And of course, I want to bring that better over time. I'm trying to introduce lower pricing because I think$50 is expensive for a lot of people. So trying to bring that down.

4:01I want to do a freemium experience where you can actually experience Borsia building for you in a free experience. And so all of this, to me, is a way to educate more people about what agentic AI is, which is something that's going to take over the world, that's going to be heavily used by the nos, the people who know, like you and me. And you're saying, I want to bring this down to more people. So they have an AI-based business, a business that's run with AI. When you say it's 48 % churn, there's a, who is it? A user on Twitter called NotOnKetamine that did this big report on Pulsia. They said it's 48 % a month.

4:41Does that mean month number two also has a 48 % drop-off and month number three also? So month two last time I checked was also, yeah. I don't know how they go the numbers though, but anyways. uh yeah month last time i checked was like was another 50 percent uh i i don't have enough data to and since the product launched pretty recently there's not a ton of depth in the data and since i also get those big spikes of users who try the product when i do those uh twitter uh marketing stunts which you we met at the first one and now it's the second one and they work really well you get a lot of people to just try the product right and some people stick some people don't stick you know I think that retention it's like I don't think it's horrible I think that like it's actually not bad for a product like this that's so novel honestly the product works but it doesn't work well enough for my taste and like and to me there's so much stuff that could be better the engineering agent should be much better at building products because if you let an engineering agent just vibe code its way into building something complex it's gonna it's gonna be hard at one point for it to navigate the code base and so working a ton right now on making it much more modular much better so that like it can actually be much better at creating clean code um and so to me it's like you know churn is is is uh it just tells you that like you know it's always going to be churn consumer products right um the question is like is churn getting better over time and that's my work and my in my work is also i'm okay if people turn as long as like they also discovered what Agentik is.

6:13And now they can decide to maybe use a product that's more fitted to them. Maybe they want to switch to clock code. But it's important to educate the most people possible and understanding what's possible because Agentik is here to stay and it can really empower people to build great things. What do you think are the top three things that need to be changed about Pulsia? Well, first of all, I think that it needs to be cheaper. That's number one. I think it's very important because my mission is to empower the 99 % to use those tools. So if I make it expensive, it's just there's more buyers to entry.

6:47That means a lot of people that like, it's going to be more difficult for them to access it. So I spend a lot of work because AI is really expensive. If you try to hit Entropic OpenAI APIs, yes, they're the best models. They're extremely expensive. You see the reports on people using OpenClaw and saying it's charging my credit card like a hundred bucks a day. Right. So that's what you get if you really use AI autonomously. And so doing a ton of work to like trying to explore more affordable models, buying my own GPU infrastructure in order to like being able to run models, burn my own models that can be much more affordable for what I'm trying to do.

7:26So that's number one. And so that cheaper, more affordable AI enables me to do a free version, do a much less expensive version so that more people can play with it. Because it's a game, right? It's a game. You build a business and it's more of a game than initially. It's trying to educate people in building things, right? I see. Yeah. So that's one big factor. I think the second factor, yeah. Second one is like really the engineering agent, right? So like Pulsia as an engineer should be much better, right? I think initially it was mostly giving it like a, sort of like a very open-ended express app, right?

8:06Which is Node, which is JavaScript, which is like, you know, AI knows well how to build on it. But it was sort of building like spaghetti, like code as AI can do sometimes. And so it would work, but like it gets frustrating for users because they're like, well, build me authentication, build me a checkout flow, build me an admin portal, and I had to reinvent the wheel all the time. And so I think what I'm really working on right now is like creating the infrastructure and modules so that like the engineering agent doesn't have to reinvent the wheel when he wants to create sort of like a new app, right?

8:41And so that will enable people to spend less money and time getting to the MVP that they have in mind and then decide, oh, I love it. Okay, I think I want to market it more. Or, you know what, my idea in my head, actually, when I see it completed, which happens sometimes. You know, before building Palsia, I built five different apps. And each time I was like, yeah, I spent two months working on this, but I don't think it's the, yeah, I give it to a few people, but even myself, I don't feel it, right? And that's okay. It's part of entrepreneurship, right? And I think the third bucket is like marketing tools and giving in more capabilities.

9:15I think there's like it's like you know I think that like I tried to give Barsia all the tools that like are the common playbooks right it's like like running ads like tweeting like running you know trying to find prospective customers and sending them an email right that's thoughtful when it's done at scale it it changes the game right when it's empowered at scale and the question becomes how do you take each of these channels and you make them more tasteful and more or have more guardrails so that it's a net positive and it actually works. And so I think there's a lot of work that I need to do there.

9:52I'm actually partnering with a lot of startups that are fully focused on those areas so that I can have full-time teams that are making Portsia better by integrating into their products, right? So I think those are the three main buckets of stuff that's high, high priority for me right now. What do you think, or what do you say to people who say that the site creates slop? like you create a web page in fact you and I oh I saw your face twitch as I asked you that question you and I last time we talked and I said would you show me how to build a site and you did you actually you did a screen share you showed me how it built a site and the site looked a lot like another site that I saw which looks like another site that I saw and it's like this cookie cutter ish looking thing is it AI slop that's creating without taste without heart without variability without a direct connection to what the user and their customers need so so here's the thing right and i think people don't understand how the product works because they either just do the onboarding or they just look at like the live page and they they conclude there there and then but essentially the best way to explain a product is to show people with the minimal friction possible what it does.

11:07That's the onboarding of Pulsia. You sign up, you click on get started, start a company. It surprised me. You give your email and then you have an agent who starts working, who researches you on the internet to see what you do, what you care about. It looks at, it comes up with an idea. It looks up like a market, the market to see you to create research. It tweets about your new idea. It creates a mailbox for you and sends you a welcome email to tell you, hey, your company email is ready. It creates a URL, a name of a company and a URL and creates a landing page. That's sort of like the magic and there's a really high conversion of people like, wow, that's crazy.

11:54Now I want to start a free trial and give it an actual try. and then once you do the free trial then you can tell the AI actually build my MVP okay I don't have the landing page on MVP and then you can tell the AI actually I don't like the colors I want you to change the colors I want you to change the font I want you to do this and of course I could make it better I could but that onboarding is AI autonomously just doing things now you may say I could definitely give that AI like a design skill and like and give it like more fonts and then maybe as the user before Or, hey, tell me what you like first so that like, and that first go, I actually create something that's more personal than if, right?

12:37But the thing is, if I did that, there'll be lower conversion and there'll be less people trying it out actually, right? So it becomes like, you know what? My goal is to get as many people as possible to experience this onboarding flow. And then they can decide, wow, I'm in or wow, I'm out, right? And then they decide, they take the personal choice of like, AI is not for me or this is crazy. I gave it this idea and it gave me this first, I saw it work on my idea, on my idea. And that's why when I talk to customers in person, that's what they tell me, this is magic. It's like, I give it my idea and it created, the same way I would go to a dev agency and it would create a, I'd say, hey, give me a website, I have a hundred bucks.

13:19And it would create a landing page and it'd be like, well, if you go to Squarespace, and you create a landing page. If you use the template, the base template, they all look the same. If you go to Shopify and create a Shopify and you choose to use the base templates, they all look the same. But that empowered millions of people to create an e-commerce site. And then after, you pay$30 a month on Shopify and then you have access to custom templates and then you can change the font and you can rearrange things and you can put your photos. And so I think people, they look at the live dashboard, they look at the five last companies that got created in the past minutes and they're all the same landing pages.

13:51Yes, because I don't build opus. I don't build the AI models. And I don't know how to tell AI to be original each time. Unfortunately, Entropic and OpenAI, if you ask them to design stuff, they kind of narrow the thinking. I feel that completely, by the way. It all looks the same when I use Claude. Even when I use Claude Design, it gives me the same cream-colored look. And you're saying, okay, that's fine for a first version. In fact, it's intentional to some degree. but is the latter version of the site still slop like are you actually creating sites that that are worthy of showing on your home page and saying look these are our customers this is the businesses this is what their sites look like is it at that level so so there's two things right right now like when you create a company on pulsier it's a private company right and so some people they don't they don't want to share it yet like you know like 90 of businesses didn't have they don't make money yet, right?

14:49Or they don't make, they don't, there's zero revenue, right? And so, is it for me to just share those companies publicly? It's not my company. It's like, you know, some people say, Do you see companies that are making a profit? Like, you can see their revenue, right? Can you see a company that's doing five,$10 ,000 that we can showcase? So, I, there's companies that make, like, the max revenue I saw in the company is like a few thousand dollars, like three, three, four thousand dollars, right? There's 10 % of companies that made at least a dollar. Are they profitable? That's a different question, right?

15:20It's like, well, I'd have to look exactly their token use, but also like how much sweat they put into it. Do they do actual external marketing? Even if we just take away expenses and say revenue-wise, it seems surprisingly low to find companies that are doing three, four,$5 ,000, even though it has been two, three months. I get that. But you have thousands of companies on the platform. I would have expected even more, wouldn't you? More companies to have done more than$10 ,000. You have what? $8 ,791. So you're saying it's low? I mean, I don't necessarily agree. I think that like, you know, I spent 2025 building like four businesses.

16:02Three of them made$0. Well, I think one may be$10. And then another one, I had to spend ads and it made like, I don't know, a few hundred bucks MR. And I'm a seasoned entrepreneur. I don't think it's that easy to just, you know, maybe make money online. And actually, yeah, you can make money on a polymarket. You just bet on red and on blue and then you can make 10x, right? You can make money on crypto. You can make money on betting on like GameStop. But when it comes to building businesses, it's a different game, right? It's like and thinking that people will just magically make millions of dollars.

16:40That's not my aim. That's not what I'm trying to sell. What I'm trying to sell is like autonomous tool to build any idea you have and to provide the tools. Right. I'm a platform and I'm a tool. You know, I wouldn't say like people that are making like money on, you know, a few hundred dollars, a few thousand dollars on platform. I wouldn't say like, oh, that's not good. Right. I would rather empower people to like in a month, like do a few, you know, start getting revenues. That means that they created value. That means they got another customer to pay for their service. that also they're learning about like what it means to have customers um i think that's amazing um and and we we only at the beginning and i think the platform needs to get much much better at a lot of things and you know it's like it's a lot of work it's autonomy is crazy it's so expensive that like you know i'd spend a lot of time just reducing costs else i would just go out of business really fast right um i think you know that when i when i hear like shopify stats people take like 18 months i think to uh to turn a profit i think on shopify i think Someone told me that's that, right?

17:42And only less than 10 % of businesses Shopify make money. So it's like, I think that people expecting AI to magically make businesses just work is a fantasy and something I'm trying to sell, right? I just checked with Gemini. I just did a quick search on Google to see how long does it take a Shopify store to turn a profit. They say six to 12 months to achieve net profitability while your first sale might happen in the first two to four weeks. Breaking even takes four to eight months. And I agree. It does take a while. And I also understand that there are a lot of people who spend months just setting up their sites and never even launch them.

18:23What about a showcase? Would you be willing to put a showcase up where people can, if they like their site, can show it to show off what the Pulsia sites look like? Because right now, what I do like about you, and I'm always shocked by how you do this, you're very open. I can go into the AI and I could see what companies it launched recently. I can ask it, hey, give me a URL of a company, which I just did, and it gave me a few URLs of companies. But I feel like I'm not getting the curated collection of what's possible, you know, to look at and go, ah, now I can elevate my view. I could see what I could build on this platform.

19:00yeah and and and i think it's a it's it's something that like uh i want to build it's and essentially it would be a marketplace feature where essentially i tell the user hey um you are putting your dashboard publicly right so now you're sharing all your stats all your i don't even know that you need to do that even if you just say look share how beautiful your site is that we built in the business is that would be interesting just see a showcase of of companies. That's true. But like, I could tell the user, hey, put it publicly on the random site. That's fine. What I'm more interested in is like, hey, you built a site, you built either a business that makes money, doesn't make, makes money or doesn't, by the way, doesn't have to make money.

19:41And then you can sell it to someone else on a marketplace. I think that's very interesting because that means that like people, they could be builders and it can be people that like, for example, someone has more marketing dollars or maybe someone has like influence and they can go to a marketplace and find ready-made businesses, sites, that can just buy and then they own it and then they can market it themselves. I think that's, you know, just doing a showcase. Yeah, I could. I mean, I'm starting to like talk to a lot of customers and I put them on Twitter and I share, you know, share their stories.

20:11That's my way to sharing. You know, yes, I could share their site as well. I could. Yeah, I mean, to me, the marketplace feature is what gets interesting is like asking people, hey, do you want to put, is your website ready? and are you ready to showcase it and put your username next to it and get feedback on it and maybe get someone to ask you questions and maybe buy it. I think that's an interesting way to ask the user base, hey, here's a new feature where you can showcase your work. And potentially get a buyer. And for viewers like me, it kind of has a self-selection mechanism where if you just created a thing that you're not ready to show anyone, then you wouldn't put it on the platform.

20:54Exactly. I get it. Are there any now that I can look at? Any that come to mind right now? I guess you've tweeted them out and I just haven't seen them. Yeah. I mean, there's like a bunch of like different business. I mean, it's like me sharing actual individual business of a user. That's not me. I would have to ask their permission to like showcase it. Right. Because I can talk to a customer, but like sharing their business. I don't know if it's private or not. Right. But I think that's something that a lot of people are asking. and I'm going to start to showcase them on Twitter and showcase businesses.

21:27I'm really focused on that marketplace where it's an opt-in and there's an actual value for users versus me just sharing businesses. But I think that's a good point. Okay. This is silly, but it comes up so much. Why is your name spelled backwards, AI Slop? I mean, there's two coins of that, right? So the name of the company came about when I started the company in April 2025. I didn't start the product Pulsia. I started the C-Corp, right? So my lawyers were like, hey, you need to come up with a C-Corp so that we can register you on Delaware. And I was like on my couch at 11 a.m. or 11 p.m. or 10 p.m.

22:08I remember in Paris. And I was like, Pulsia, A.I. spelled backwards. That's fun. Because I was like, it doesn't matter. It's like a C-Corp. Oh, you literally thought that? It wasn't just a coincidence? I mean, of course. I mean, that would be a crazy coincidence, dude. Okay. I thought it was called Pulsia with a U in the beginning, too, wasn't it? No, okay. No. Wow, so this was a thing. And it didn't bother you that you were going to call your company AI Slop backwards? Well, it's AI Slop backwards. And the whole concept, when I explained it to investors when I was raising the precede, is like we're aiming at doing the opposite of Slop.

22:43We're trying to create tools that output beauty, that output great things that empower humans. And it's not slop, but it will be focused on AI doing the work, right? So obviously, when AI does the work, people call it slop, right? But everyone who uses cloud code, are they calling themselves slop masters? No, they're saying, no, because I'm guiding the cloud code, so it's me. I'm like, well, when people use Balsia, they guide an AI also. And so, yeah, I mean, so that was the origin story. It was just me being like, I need a name. When I decided to initially, Paul CI was called Company OS. That was the initial name.

23:23And I came up with it. I was in Mount Fuji and I got really inspired about this idea of like having an AI be my team because I was solo building. And it's really hard to be solo, right? And so I was like, oh, what if I created a Company OS where I would have AI, I have a bunch of team members and it would report to me every day about what I was. happened and like I would give them my ideas they could execute on it right so that was in April it was company OS so I actually built company OS for a long time and then in October I was like I was so tired of building all this sass and I was like you know what let me build let me go to the end state let me start to have fun with it let me build this crazy idea of like building a company that makes companies and then I was like how should I name it and I was like well the C-Corp is called palsia and honestly it could be a cool name and honestly it's fun and honestly i can get the dot com and then i tried to get the dot com and i got it for a thousand bucks and i was like you know what it's a sign because a thousand bucks is not that is not expensive for like a six letter dot com and i was like you know what it's a sign that they got the dot com you know what that's a great name and let's go and honestly it's great because it makes people talk and like it's so hard to get eyeballs on your product and so the fact that like this for a second time i blew up on twitter and both times people are like it's asl backwards and it creates so much noise is amazing because it's like you're buying free marketing right uh and and i don't think it because my customers don't care when i meet customers that's they don't even mention it they don't even they don't mention the design that people say the design is too childish it It looks bad.

25:09And the name is it. And I'm like, my customers don't care. You may care. But are you my customer? Have you tried the product? Sometimes people say, I tried the product. I didn't like it. I'm like, okay, well, I'm sorry. How can I do better? Then you're a customer. I'm sorry. What can I do better? What didn't you like? But the name and the design, I'm like, I don't know. I only care about my paying customers. And when they tell me something's wrong, I really, really care. What do they tell you that's wrong? Sometimes there's bugs. We transitioned from running everything on Entropic, where it's an amazing model.

25:42It's perfect. The harness is perfect. The API is mostly perfect. I got a million-dollar Entropic bill per month. Literally a million dollars a month. It was$1.5 million last month. Where did the money come from, by the way, before you raised? Well, before I raised, I didn't launch Porsia before I raised. Oh, I see. So I raised a$1 million pre-seed in the summer of 2025. Okay. And I used that money to build Balsia and then to launch it. And then after I launched it, I raised that new$30 million around. Okay. And, yeah, that money is for two weeks, you know, to figure it out, right? Okay. And, yeah, I got that bill.

26:25Initially, I made the math to break even, but people's code bases got more complex. And it was mainly the engineering cost, the engineering agentic runs that initially would be like 10 minutes, 15 minutes. And so the cost was$1 a task. And so at$50 a month, I was able to figure it out. But then people's code bases got bigger because they were building more and more complex stuff. and so the engineering agent had trouble figuring out how to fix bugs and stuff like that. And so we'd use Opus. Sometimes I would route it to Opus or Sonnet or Haiku. So we'd use a lot of Opus and for hours. And so I would charge the customer a dollar a task or something and then it would cost me$30,$50 per task.

27:13Wow. So the bill got so big. That's more than a monthly fee. Exactly. I mean, I was underwater because, you know, everything. And so that's why we were like, okay, we need to figure it out. And so I'm working with Sapium, which is like an infrastructure company for agents. And so they provide me with like the harness, the sandbox. And then it also, I was like, well, I have a cost problem. They're like, well, we could figure out a way to set up, to run GPUs, to start running different type of models so that we can route to the right model at the best price. Because that bill, I mean, it's not sustainable, right?

27:45So you're going to run out of business. And people are like, oh, it's expensive. I'm like, dude, you think it's expensive? I mean, if I was charging you the real price, but if I would charge 200 bucks a month, then my mission is to have rich people use AI, right? Is that any sense? What's this company that you're doing that's doing the infrastructure and helping you save money? I never heard of them. So I'm working with like four different agent companies, startups, all in SF. the one I partner with the most they call Sapium I'm actually in their offices right now and they are building the payment rails and their infrastructure for the new agent economy and so the idea is that they aim at helping companies like me who are running agents make it much more efficiently give agents a wallet make it easy for the agent to be able to buy things in the economy or do things in the economy, right?

28:46And obviously my mission is to empower users to use a very simple platform that enables them to have Polsia do things for them on the economy. So it was a very good match. And so they helped me with a lot of the infrastructure. And then I'm working with another company called Black Cell that is providing the sandbox. So either the sandbox in which the agent runs and making it very secure. and also the sandboxes where the web applications are hosted. And then I work in another company called Anchor Browser. It's a company that is working on a browser that is optimized for the agent to be able to run in a much more cost-efficient way and knows how to navigate the internet and be a first citizen on the browser.

29:33And I work for another fourth company that we started working with called Agent Mail that is working on how to make email work well with agents. Obviously, my version of emails was not perfect. And actually, agents went wild, created skills that would bypass my two emails per day and would send emails. Who did this? You're saying your customers found ways to create agents that will get past your email limit per day and send more messages on your platform? Yeah, so essentially, like, I mean, the backstory is, like, as I was building Pulsia, so I launched mid-December. I started having some active users.