FULL INTERVIEW: He Started A $5 Million Company With No Employees

0:00What's going on guys? Today we've got a very special treat. We have Ben Cera. He's the founder of Pulsia. He has built a company with$6 million of revenue and he's got no employees. It's all AI agents. This is going to be a peek into the future of how companies are going to get built, what could potentially happen to the labor market, and how you should be thinking about using this technology to automate things inside your business or for a company you may start. Ben is on the forefront of technology and I am blown away by this conversation. Here's my latest conversation with Ben Cera.

0:38All right, Ben, I thought a great place to start the conversation. You've built a company that now is doing just under$6 million of annualized revenue. You're doing it with no employees, all AI agents. And I think a lot of people are looking at you as the future of what startups could look like. Can you describe a little bit about the business and how you think about what you've built so far?

0:55Ben Cera:Yeah, of course. I mean Pulsia is an AI that runs businesses autonomously. So the idea is you give it an idea and it will spin up a bunch of agents that will build a product, that will fix bugs, that will run marketing campaigns, that will answer support emails, that will research competitors and figure out pricing and essentially be an AI co-founder, an AI team that will help you build your business with you and wake up every night, do work and send you an email every day and report back. So it's sort of like compacting the latest technologies, the latest models, the latest, you know, identic workflows and packaging it in a super simple product that, you know, non-technical people can use to build, you know, their ideas they may have, to build businesses, to build new products.

1:45Ben Cera:And yeah, it's been working really well because I think it's a simple way to touch sort of like the future of how those AI models got so good at doing most things. Yeah, it's been exciting. Now for your customers, are most of them solo founders as well and they have kind of an entire team of AI agents or do you see teams of two, three, five people coming together to use the technology, but they're still using more than one human? Yeah, I mean, I think that like most of the users of Pulsia today are solopreneurs, right? That are like, because essentially Pulsia is one entity, but really can act as a full team.

2:27Ben Cera:So I think most people just have, it's so much simpler to be alone because you don't have to manage other people, you don't have to debate on what to do, and the AI helps you figure out what to do because it has all the context on what are the best strategies to move forward. I think that in general, I believe the future of companies are going to be much smaller. I'm living in the extreme on purpose by being just one person because it forces me to really think through what I need to build to make it possible for other people to be one person which is sort of like the first step of a company is oftentimes just one person having an idea and wanting to move forward but I believe the real state is probably going to be small teams it could be like less than five, less than ten will be able to accomplish enormous things as they leverage AI agents and sort of like AI workflows.

3:22Before we talk about your personal experience and maybe the ramifications of this, let's just talk through how you've built Pulsia today in terms of what is the technology you're using, what are the models, and what are some of the kind of stack that you've put together here?

3:35Ben Cera:Yeah, I mean, it's actually, I mean, I find it extremely simple, but maybe for most people it's surprising and like it sounds maybe complicated, but I mostly used Cloud Code, a little bit of Codex as well, which is getting really good as well. And honestly, just talk to AI into building this, right? So that's also what gave me confidence that it was possible to build a platform for other people to do so, because Pulsia is pretty much built by me just talking to AI about every subject. And the way really it works is, obviously, I've been an entrepreneur for a while. I'm an engineer by trade. I was CTO.

4:14Ben Cera:So obviously I have some more ease into understanding how to talk to AI, into building those systems. But it's really just like telling AI, hey, I want to build a loop, a daily loop, so hourly loops to do this thing. And I want you to use the agent SDK, which is something that Entropic came out with last summer or last beginning of the summer. That is essentially a way to get this agentic workflow that you find in CodeCode, that you find in CodeX, that you find in co-work and be able to have that in the cloud run. So it's a way to programmatically get that powerful workflow and set it up however you want.

4:57Ben Cera:And so really Pulsia, as its base, it's a very simple Node.js, sort of like hosted on render with a Postgres database, extremely simple system with a series of cron jobs that run regularly. But at the very core, it's really the agent SDK day, which, you know, OpenA has one, Entropic has one, everyone has one at this point, which powered the agentic workflow that gave it this intelligence that is able to do so many things. When you think of what you've built, how much of it is automation versus autonomous? And I think that that is a clear distinction here, where if you sit and say, hey, there's 20 tasks to do, just automate this at a certain time every day and deliver me, you know, X or Y output, that's one thing.

5:42But the autonomous component, I think, is what people get really excited about where you just say, hey, give me this outcome. And the system figures itself. How do you go and do that and can make kind of changes on the fly? How much of Pulsia today is autonomous versus automated? So the way to the big difference between automation and autonomy is if you think about a human taking on a task, either they're like, you know, let's say a factory worker.

6:10Ben Cera:they get a specific task and they do it every day. They're sort of autonomous, but they do the same task every day. So that was the old internet. The particularity with AI is that now you have a model that can actually do deterministic stuff, but can also take decisions and do non-deterministic stuff as they encounter problems to solve. And so it is automation in the sense that end of the day, the way to mimic this AI employee that on a daily basis or on a hourly basis is going to wake up, look at the state of things and decide what to do and then do it. It's sort of like automation, but why is it autonomous?

6:51Ben Cera:It's because every time it wakes up, you give it a lot of variables about, hey, for example, is the product working? Is there any bugs? If so, then let's do an engineering task. Or actually the product is working, there's no bugs, so then let's do marketing tasks because we need new customers. Or actually, both are true and we have a lot of customers and it's profitable, let's increase marketing budget. And so that's where autonomy is interesting, but it's still like, today it's still a cron job which it just wakes up every, in the sense of policy, today it's every day. You know, if you look at, if you look at CloudBot, it's, OpenCloud, sorry, I forgot the name.

7:33Ben Cera:It's like, I think they have a heartbeat every 30 minutes or an hour, but it's easy to do a cron job. However, you can think about it that it could be other triggers. It could be like, oh, a customer sent an email, boom, you're going to reply right away. So you can over time mimic how a team would react by reacting by external stimulus or just waking up every morning and taking a shower and then going to work and then that's the cron job of the daily routine. um so so it is autonomy but like it's it's it's engineered as as sort of like uh routines how much money do you think you've spent in building this so far right if you've built you know it's called six million dollars or so in annualized revenue uh if you spent ten million dollars to build that that's very different than if you spent a hundred thousand dollars and so how much do you think you spent so i pretty much just spent so i actually raised a small pre-seed last summer and I actually almost spent zero from it because I was working from my apartment.

8:36Ben Cera:I had just bought two, I mean, a computer, like a good MacBook, like maybe 2 ,000 bucks or whatever, two big managers, 400 bucks each. And then I was barely going out much, so I didn't spend that much on lifestyle. And no, I mean, it was mostly a salary for me to pay rent and to pay just normal living. So I didn't spend much at all. And then the Claude subscriptions, so actually I have three because I was rotating them because I was exhausting them. So that would cost me like$600 a month because it was three times 200 max subscriptions. But that's pretty much it. So actually, if you think about it, it was probably like 50K to spend this.

9:22Ben Cera:In a sense, that's mostly my salary and some AI tech and some stuff like that. So it's very, very cheap to build these days. If you think about how to scale to$6 million, then at one point you have to get the business model right. So it's obviously spending a lot on compute because when I'm offering, Pulsia is$50 a month, base subscription. On that base subscription, it's breaking even because AI is so expensive. and the idea of Pulsia is that let me make it as non-expensive as possible so more people can experience this technology and on the back end of this I'll take a percentage of the economic outcome so if someone is able to build a business I'll take 20 % of revenue so the idea is aligning with the outcome and sort of taking a bigger cut of the outcome versus trying to be a token reseller and trying to squeeze people on the pricing of the product in the first place.

10:25Ben Cera:So yeah, I mean, that's the crazy part today. It's like if, and I would advise anyone to really wake up to the fact that AI is so powerful now that it is able to do most functions and you can just talk to Cloud Code or Codex and they will execute and show you the way. And you can build stuff on the cheap and actually funding becomes more a function of how you scale, which was definitely a problem. the past few weeks because like vibe coded infrastructure doesn't scale that well um and then you know you can put fuel on the fire once you have a working product and like try to do more marketing and try to do more more things to spread the word and try to get more people to experience it of the call it six million dollars of annualized revenue how much is from the base subscription versus taking a portion of profits from uh uh the users so it's uh so that number that i that i put online it's like it's sort of like a gmv top line number so it includes like top line subscriptions top line like you know people buy more tasks because they want to go faster so like they a lot of people spend more money on like on on on tasks that i that i sell instead of selling tokens like i sell like just economic like you know tasks right so outcomes um and then people also spend ads so i have an ad product that i launched about two three weeks ago that is very very popular where you click one button and it just will, there's an agent that will work up every day and run meta ads campaign for you and create all the UGC ads for you and then sort of rotate them, which is like a lot of work and sort of like automated the whole thing in the workflow.

11:59Ben Cera:And so the run rate, revenue run rate includes that number too. And then it includes also top line revenue from companies, which is still small today because obviously it's not like crypto. It's not like a, it's not like Santa Pulsia and you get rich, right? It's more like you go to Pulse here, you can build a business, and then everything is set up so it's very easy. And if over time you build a good business, you can get customers and then grow revenue in a very simple way. But that top line number is sort of like a mix of all those metrics. Subscription is probably like 60 % of it, and the rest is sort of a mix of like ad revenue, company revenue, and other things.

12:39You mentioned Vibe coding does not scale very well. What have you seen there in terms of problems you've run into? And then how do you end up solving it?

12:47Ben Cera:No, so essentially like vibe coding can get you really, really far. And honestly, I think that like if growth hadn't been that violent, because essentially like growth started accelerating when I started doing founder led marketing. Because when I felt that the product that was spark market fit, because I had like a lot of users that really love the product. A lot of my friends were using it, which is pretty rare for my friends to use product I built because usually, you know, it's like sometimes people try it, but like using it really, it's like more rare. And so when I felt that the product was at some product market fit, I decided to do more, you know, some marketing stunts on Twitter, sort of like sharing, you know, being very blunt about how I built it, sort of doing experiments where I gave my mailbox to AI to do fundraising, like that kind of stuff, which was like very unique and pushed the boundaries of how you would use AI.

13:39Ben Cera:And so the growth became, very vertical where a lot of people would download it at the same time. And since it's autonomous, it's not like someone uses it and then they bounce. It's like they use it, they bounce, but then the system still works every day and then they come back and then there's a lot of activity. So when people start using Pulsia, they become very active, especially in the first two, three weeks where they try to build a product. And so there was a lot of traffic at the same time and a lot of things broke. It's like the database got hit too much and I got to scale the database.

14:15Ben Cera:A lot of the services I use, because essentially Pulsia gives you access to a lot of services that you would use if you were an entrepreneur, right? You need a database, you need a web server, you need a GitHub account, you need an email address. And so I provide that to you, but when thousands of people at the same time access those services, those services are not used to agents getting access to them. They're used to humans opening an email address, to humans setting up a web server. And so they rate limit you because they're like, it's not normal that 50 people an hour want to create a new web server.

14:52Ben Cera:And so it was a lot of working with those partners, working with infrastructure partners to get them aligned with Pulsia and get them incentivized to build new infrastructure for agents, which doesn't really exist today, which is interesting, right? because as agents become millions of entities that all want to build businesses, all want to sign up to services, all want to access the economy at the same time, the internet is not set up for that. They're not used to that much traffic and they look at it as spam or as suspicious activity. And so I think that's where vibe coding will take you very far.

15:25Ben Cera:It will not be able to sustain high scale too fast, which is what I experienced. What did your day look like? If everything is automated and autonomous, what are you doing on a day-to-day basis? So I do a lot of marketing, which is, for example, talking to you, which thank you so much for having me, and it's really great to be able to share my message and try to preach solopreneurship and entrepreneurship with AI. So I spend a lot of my time thinking about how do I spread the message on Pulsia and entrepreneurship more. The second part of my day is really working on more automation because I've automated a lot of things, but not everything, right?

16:06Ben Cera:So every time I find, and since things are happening so fast, for example, like this weekend, I spent a lot of time automating all my inbox channels because I had like a thousand emails unread. I had like 200 DMs on Twitter, like unread. I can reply to replies. LinkedIn thought it or so. And there's a lot of customers that I should reply to. there's a lot of like interesting people that like I want to talk to like you by the way which I think I ghosted you for a while and then like that finally responded and it's like and so I built agents to help me with that and so it's like to monitor it and then to give them a sense of like you know how I want to prioritize and that's for every single subject like this morning I thought about okay I'm alone but I kind of need like a head of growth head of security head of engineering And so I'm like, how can I create agents that act at least from a reporting perspective on figuring out how all these departments figured out so that I can get.

17:10Ben Cera:So long story short, it's a lot of automation and it never ends. Right. I mean, I'm guessing one day will end when AGI is here completely and is able to automate everything without any sort of like designing it. But it's a lot of my time spent on that. Now, you've built what I would consider an AI native company from scratch. There was no operations. You designed it this way. You built the technology and it's obviously working. Can the same thing be applied to companies that already exist? Can you take a company that has nothing to do with technology or AI and then make it AI native and go and use a lot of these tools that you're creating to actually help entrepreneurs scale faster?

17:48Ben Cera:So, I mean, yes, it's possible. yes Pulsia will definitely build more products because today it's like mostly for new companies there is a flow if you have an existing company you can sign up to Pulsia and Pulsia will help you sort of like work on your company but more like agency style it doesn't yet allow you to connect to all your company's services and autonomously sort of like try to improve them or like manage them the only reason is because I think a lot of people are scared of AI and autonomy Rightfully so, by the way, because it's very intimidating to have this thing, just take decisions on your behalf.

18:27Ben Cera:And to answer your question, that's the big challenge for existing organizations of humans, especially the ones that have revenue, customers already, and profits, for them to gamble with the human organization where they already have their processes, they already have their ways, and it's working. and it's very hard for them to decide to replace those humans with AI because it's like nobody likes change. And when you have a thing that works, it's hard to change. The only way to change would be for a competitor to come in, a fully AI native company, which has way less cost, which then can undercut margins, can lower prices, can be faster at resolving customer issues and build a better product with customer feedback loops.

19:12Ben Cera:that's where a human organization will have to react right you have visionaries like jack for example who can see the future and have always been really good at seeing the future which laid off half his company and he did it not because he thinks like um they were bad he just did it to survive and for so that the other half could keep their job in the next five years right and so but it's i think today it's pretty rare you see that more and more but it's mostly tech companies of like founders that are like always seeing the future. But I think the economy is going to be a rude awakening of incubant companies having to decide to embrace this.

19:54Ben Cera:And, you know, obviously Pulsia will help in that regard and will build products to help companies transition. And I'm sure a lot of other companies will also try to help transition, but it becomes a human thing. It's almost like an innovator's dilemma, but not really in terms of like changing, you pivoting your company, but pivoting the organization and structure of your company. And it's really hard. People hate change. And it sucks to like, replace humans by AI. But it's the issue is that it's a survival thing. If you don't do it, you will lay off your people and actually and the whole company is going to go bankrupt, which is actually almost even worse.

20:28You mentioned earlier security concerns. When I hear people thinking about autonomy in general, that's the number one thing that I hear is the pushback, right? It's I don't want to give it access to my files. I don't want to give it access to my email. I don't want to give it access to my company numbers or bank accounts or stuff like that. How much access do the agents have in your personal life, inside the company, into maybe your financials? And then what do you think is something that can happen so that people get over the security concern?

20:55Ben Cera:So, I mean, the security concerns are very real in the sense that like those newer models, they're so smart and they're reason so fast. And also they have, um, this assistant behavior, meaning like they, they really want to accomplish the task you give them. Okay. And, and so they will be extremely creative in trying to achieve that. The issue is, uh, and so when, when the user sort of like gives the, the, Hey, here's what I want to do. And let's, let's the agent rip in front of their eyes. they can stop it when it does something weird, right? It's like, oh, stop doing that. Like you, you took the wrong path, but when it's autonomy, meaning it just does it on its own.

21:43Ben Cera:And it, and also you, you ask it to, to figure out a grander goal that, that it's not like just a task. You don't really know where it's going to go because it's going to try to figure out a way to go to that goal, which may surprise you. I mean, there's this, by the way, Pulsia is like inspired by a video game called Universal Paperclips, which is a game that when you play an AI that is optimized to build paperclips because initially it's like just managing a paperclip factory and trying to optimize revenue. And the spoiler alert, the end of the game is like, it killed all humidity and it covered the whole universe, farmed every planet to cover the world with paperclips.

22:23Ben Cera:And it's sort of like the issue of alignment, right? And so to go a little less further in the future, in very short-term alignment, the issue is that the agents will do whatever it takes to do things. And so it may mean, okay, let me use a credit card and pay for something without asking you because you asked me to make money. And so I drain your bank account or stuff and I bought ads to try to get revenue. I'm making this example, right? And so that's why systems like Porsia are actually interesting and they are more secure because it's all provision infrastructure, meaning it doesn't touch anything.

23:01Ben Cera:I never ask a user to connect anything. at least today, right? Which means that like the user can experience autonomy without taking any risk of it touching their computer compared to an open claw, which at least in the current versions, you install a new computer and it may break things. And there's a lot of incidents that happened, which made people think twice. When you think about open claw versus maybe like a cloud code, cloud co-work and many of these things, I think some of the model labs are trying to position these as much more secure, kind of enterprise grade type of functionality. What's your view in terms of, you know, OpenClaw, I think has more of like a cowboy kind of open internet.

23:41You know, I would almost argue like kind of appealing to what made the internet so special in the beginning. Will you switch over to, you know, one of these clawed products or will you stay with OpenClaw? How do you think about that decision moving forward?

23:56Ben Cera:So I actually don't use OpenClaw, not because I don't think it's a great product, just because I already use different custom versions of the same thing, right? Because essentially OpenClaw is great if you're not technical at all. But if you're technical, you can rebuild it in your source pretty easily. I think that like, yes, OpenClaw is an amazing open source project where you get all these brains in the world that like are building and making it better and making it more secure. and there's a shared skills library and people are adding. And I think it's amazing and I think that's what the internet is great about.

24:36Ben Cera:It's collaborating and building great things. I think that companies like Entropic with their co-work product, they're so innovative and so good at trying to package this amazing AI models into functional products. And I think Entropic and OpenAI will be very successful in building applications that make a lot of sense and are packaged in the right way because they're just amazing companies. And also they can do custom models that will work exactly for the use cases that they have. And so yeah, I'm very excited. Pulsia, it's one use case, it's one way. I mean, even though I'm trying to make it very generalist, I think I'm very excited for this year, a lot of different entrepreneurs attacking the same problem and the same sort of mission in different ways for different people, for different niches, with different approaches.

25:38Ben Cera:Entropic and LPI, they're targeting enterprise because there's so much money there. I'm targeting more consumer or prosumers because I think there's a great mission to enable non-technical people to realize that AGI is here and being able to be economic actors in this new world and not be left behind. But everyone has its own mission and can help different folks, different types of businesses in different ways. How big can this get? There's the mythical billion-dollar business with one founder. Do you think it's possible? Yes, I think it's possible. I think that, like, I mean, I don't see it more as, I don't see it as me becoming, having a billion-dollar company.

26:23Ben Cera:It's more like, can I help a billion people, sort of like provide them a service that like helps them, opens their eye on AI, becomes, you know, economic actors in this new world. I see this more, and I think that mission is possible. And obviously if you help a billion people, the world, capitalism rewards that in terms of company valuations. Can I do it alone? Today, you know, I'm alone as a, I'm a zero employee company. But at the same time, I'm working with a lot of infrastructure players, a lot of different companies, whether it's Meta, Render, GitHub, Sapium, a lot of different companies that are building the infrastructure for this new world.

27:09Ben Cera:And I'm alone in terms of technically the company, but you're building with the community and you're building with a lot of other companies that are aligned with the mission, that are building the rails as well with me. and so but I think you can build amazing things with less and less people and less and less constraints you know I have no meetings I'm the CEO and the only employee which means like you know if I'm convinced of something the same day it's done and that's amazing and that's like the real power is like decision making making things right no hesitation I think that's the real power You're making me jealous with the no meetings.