← Startup Stories - Mixergy
Loading summary
Transcript130 lines
- [00:00]
A
Pulsea builds and runs companies autonomously. I'm a solo founder, zero employees, with a bunch of agents. Polcia is making almost 800,000 run rates and I just launched it less than two months ago.
- [00:13]
B
So anyone who's listening will be able to sign up and have Polcia run their company. Ben Sarah is the founder of Pulse AI that runs your company. The next new thing presented by Zapier, the AI automation company. You're about to meet a founder who built. He didn't build a company. His AI coding agents built a company for him that's doing well. It's marketing itself, it's coming up with ideas for what else to build. It's fixing its own bugs. And more interesting than that, anyone could use this platform to build a company that will manage, that will be managed by AI. Anyway, let's listen to the story. Who comes up with the actually, who builds the software?
- [01:01]
A
So the software. So Pulsia is like a series of like autonomous agents specialized in different categories, right? So you have like a. You have an engineering agent that's like really good at code and it's like set up to code, that has access to like a Web server, a GitHub account, a database, all this stuff, like all the tools to make it efficient. You have a marketing agent that like has access to like a Twitter account, an email address to the cold outreach, a meta ads account to do run ads can create ugc. So like essentially giving them all the tools to be able to execute on stuff. The support agent has access to an email that I give it to the user by default. So the idea is the user comes to pulsea brings their idea and pulsea already sets up an instance of everything you need to run a business. An email address, a server, like a Vercel server, a database, a meta account to DO ads, an OpenAI account to do Sora videos, all this stuff you would eventually need to do. And that Claude will probably tell you, hey, now I need you to go get an API, an API key, right?
- [02:10]
B
And so it is developing the software for the user who comes to the platform. It is then marketing it and doing support. I went and I asked Paulsia, how are you doing marketing? It told me it's sending out cold email, it's doing ads on Facebook, it is tweeting, et cetera. I said, can I see one of your ads from Facebook? It showed me one of the ads. The ad was a plumber who looked like he was just talking into his phone, saying that he hates everything involved in running a plumbing company except being a plumber and actually doing the plumbing work. And because of whatever this tool was that Polsio was building, it was he, he could just now focus on plumbing. And then there was a link to go to, to go and sign up for the software. I run an H VAC company. Used to hate dealing with marketing agencies, always asking me questions, needing my approval. Now thorough handles everything. My phones ring, I fix AC units. That's it. This wasn't a real human being, was it?
- [03:13]
A
In the ad, I mean, so, I mean it's so today, at least in the first iteration of like the ads feature that I built about two weeks ago, it's a pretty recent feature. The AI is autonomously running ads. So like the crazy thing is that as a user you click on a button called Run Ads. There's a single button you click on Run Ads. It asks you what's your budget, right? So you can start at $10 a day because you know, obviously like it costs money to run ads and that's it, that's all you have to do. And then there's an agent that's going to like autonomously look up all the context of your company, understand how to pitch your company, what it does, what's the value, proposal sets up, you know, a description. Go to Sora today. It's mainly Sora too because I find it like really very good at like generating realistic videos and then sets up like a new GC AI video of like someone pitching the service and then uploads it to meta. Sets up like a campaign, sets up like all the, all the characteristics to make it work and then every day monitors it and like if it doesn't perform, we'll create a second one, a third one, turn on and off ads like a real marketer, right?
- [04:22]
B
Unreal.
- [04:24]
A
And so of course in the second phase, this feature is actually very popular because I'm guessing it's very complex actually even myself obviously I've been running ads for businesses because it's a great way to get users to flow in your service to learn what works, what doesn't, is there retention, has the onboarding. And what I find interesting by seeing this feature being so successful is that it shows you the more friction there is in doing something, the less people do it. And it's like people know that they. Yeah, of course ads is a good thing, but it's like setting up a mid account, like doing all this stuff is complicated and so making into a one click experience is, is what Pulse is all about, making it extremely simple. To do the thing about running a
- [05:06]
B
whole company and the analysis of pain and then the cut the losers and grow the winners. All that takes a lot of time. When you say that, let's call it the entrepreneur, that's the person who comes up with an idea. When you say that, they will say, I want to spend $10 a day. That's money that comes out of their pocket. Right, Got it. So they decide how much money goes. What about when there's cold email? Where does that money come from? Or does that cost anything?
- [05:34]
A
So a cold email is essentially a task where an agent that specializes in sending cold emails will research first, people that may be interested in the service. So people that are relevant.
- [05:47]
B
Yep.
- [05:49]
A
Second, will verify with, like, a specialized database that I pay for. Like, there's services online that I can check. Like, is this email correct? Is it. Is it a good business email? And then uses that email to. To send a cold email that's highly personalized based on the user. So actually those emails are really good. Of course, cold email is something that, like, works less and less because people are flooded with it with AI. But yeah, it sends it. And so it costs like one task. So essentially the way it works is that, like, as part of the $49 a month subscription, you get one sort of like autonomous task every night. So every night there's a CEO agent that's gonna look at the state of your business and try to figure out what is the best next step we should do. Is it engineering? Is it marketing? Is it support? And do it. And on top of that, you get a budget of credits, that task that you can do whenever you want. So now if you like, you, you're like, no, no. But right now I want to fix that bug. Like, right now I want to add this feature I can see and say, oh, right now I want to send cold emails. There's a cold email button. You click it and then, like, it will send the cold email right now. So that's a way, you know, obviously the system is autonomous, but, like, if you want to tell it what to do now, it will do it now.
- [06:56]
B
Okay, so what we're looking at is the way it works is I, as an entrepreneur, come to you or to Pulsia with an idea. I pay $50 a month for certain amount of compute time and any extra than I want to spend on advertising, I decide, and then the system will charge. And what you're saying is, Andrew, because you are so deep in your world, you know things that other people don't. So maybe as a. As a longtime podcaster, I might know that there's a process to finding guests. And if you can find the right guests like you, big audience comes. And so I might come up with some kind of guest hunting system. And I know that every podcaster is going to want it because that's what their whole business depends on. I come to the. To you, I describe what I. How I want it, and then I say, you know what? I don't think meta is the best way to do this. I think maybe it's emailing other existing podcasters and I give you a little bit of my insight into who and how you message. I leave it to Pulse every night. Pulsea will go use the software that you just mentioned, generate this list, send out emails and so on. If I want to go a step further and say, you know what, actually, why don't we do ads? That is an extra feature that I come in and I ask for. Right? That's the way the model works. And so right now, if I see that there are hundreds of companies on Polcia, these aren't companies that you started. These are companies that your customers have started and Polcia is running them.
- [08:15]
A
Exactly, exactly.
- [08:16]
B
Got it.
- [08:17]
A
I toyed around the idea of, like, Polci itself having a fund and being able to, like, you know, create autonomous companies that would be owned by pulsea. There's so much customer demand for this service that, like, I'm really focusing on providing, focusing on making customers and experience for customers great. I may start building this fun for two reasons. First, because I want to. I'm actually trying to launch it this week or next, a feature where, like, a user can actually launch the fund itself. So instead of launching one company at a time, so, like, you have a specialized insight into podcasts and you actually have an audience. So you actually don't even need to use meta ads. You, if you had this autonomous team that builds this software, maybe at the beginning, you spend a week, like, really hammering and making the MVP look great, but then it just, like, maintains it and fixes bugs over time in response to support. You could sell it to your audience because you already have an audience, right? But maybe at one point you're like, okay, that was cool. Actually, I'm generating, like, I don't know, a thousand bucks a month. And that's great. I want to do more. I have more ideas, but I cannot manage more than one company at a time because it's too much cognitive load. You could launch a fund where you spend maybe 200 bucks a month. And now you can spin up up to five companies. So now instead of talking to one Pulse. Yeah, Instance that's just managing one company, you talk to, like more of a fund manager and you say, hey, I want to go to those. I want to create specialized podcast sort of software into those five categories and attack those five niche. And then it will try actually to run those five companies and then maybe some of them don't work and say, okay, you have five slots. Like, maybe let's kill one of them and let's replace it with a new one. And so you have a rotating sort of like, pool of five startups that you're starting.
- [09:55]
B
I see. So when you say fund, it's not that I would go and raise money for this. It doesn't cost that much. When you say fund, it's essentially. It's a portfolio is what you mean. Yeah, got it, got it. And you're saying, listen, there are all these people who have audiences and ideas and insights. They. They have a company. I'll. I'll let Pulse run it for them autonomously. Wow. Okay. And then as the founder, as the entrepreneur, what percentage do I get?
- [10:22]
A
So you always get. So I always take 20%. And again, for me, it's like a way to, like, pay the bills and make sure that, like, I can reinvest, you know, make it a profitable company that can reinvest in making the platform better. And the customer always gives 80%. So it's like, sort of like the customer is sort of like hiring this team.
- [10:42]
B
Yep.
- [10:42]
A
For 54. For a price that's. I'm making, making it as affordable as I can based on, like, the cost of compute. And in exchange, Paul CS is. Well, I work, I work for, like, at cost. In exchange, I'll take 20%. But what's good is that I'm incentivized with the customer. It's like if, if no businesses in three months make money, everyone's going to turn. Right. But I'm, you know, I'm setting up all the services, cold outreach, ads, because I know that, like, those things work. You know, like, obviously, like right now it's a lot of organic traffic coming to Pulsia. But, like, at the beginning, I bought ads and like, it works. It. It finds your customer. Now, customers will not stay if it's not a good product. But, like, you can find customers very easily. And so I think, I think aligning with the customer on the outcome and, and charging them on the outcome, instead of charging them on like a pure token Basis is, I think like, sort of like the way AI based businesses are gonna, are gonna make sense.
- [11:35]
B
That's fair. That's fair, right? You only make money if I generate revenue and I'm the person who's making the decisions about whether this revenue is worth the expense. And the only expense I could see is advertising. If I want to pay more for advertising and if I want more services, like maybe it comes with cold email, but I decide I want Twitter, I pay an extra fee for that.
- [12:01]
A
So for Twitter today, it's like, it just costs you a task. So today it tweets out of like an account that like I created. So that's tweet, but that's in tweet out of like your account. But soon enough I'm gonna make it a way to. That for you to connect to your own Twitter, for it to tweet autonomously there.
- [12:16]
B
Okay. All right. Yeah, I. To be honest with you, the tweets that come out of your account are not very good. They're kind of generic. No one's engaging with them. It's not the kind of thing that is like, wow, I want that on my account. But I can totally see how if it was my company, I would want to somehow have an update, even if it's like my pulse. A bot every week updates everybody on how it's going. I was talking with Nat Eliason. I don't know if you've seen his experiment where he's using openclaw to create Felix, this agent that is going on Twitter and making money for him and he has these updates. I think it's every week with how much revenue they're making coming from Felix, Felix and Nat's account. Okay, I see that. What else is there for growing revenue? What else are you doing to help them?
- [13:03]
A
I mean, today. So today it's like mainly cold emails that like can work in like in some niche and like, if you, if it's niche enough, you can actually find people and it works. And then meta ads is like the big one. But like I'll introduce like other other types of ad platforms. And then there's all the whole influencer thing where like you could actually, you know, hire influencers to post UGC or like to post content online on their channels. And there's a lot of third party platforms that offer that and that I could plug into to offer that to customers and same thing, it's like you could do it yourself and go through those websites and do all the work and send them to a subscription and I figure out how they work. Or you go to Polcia and like, either it's. You just tell Polsia, let it rip, you know, like, just go for it. Here's my budget, go for it. Or you just say, okay, specifically I want to do meta ads, or specifically I want to do influencers. And you click a button and it just runs with it.
- [13:57]
B
And that's where you're going with this. That's the direction that you plan to take it. Just keep adding any kind of advertising that works. You're going to find a way to connect via API and then use Polcia to manage it based on what it knows about a company.
- [14:11]
A
Yeah. I mean, even more fascinating than this, if you think about Polcia, it's like more than a tool. It's like sort of like an economy, right? It's an economy where, like, people come build businesses and they're being run on the platform. You could see a world tomorrow where, like, people can buy and sell businesses on Pulsia because all the business is there, right? The revenue, the knowledge, the infrastructure.
- [14:35]
B
Right.
- [14:36]
A
And so the idea is that, like,
- [14:40]
B
is.
- [14:41]
A
Yeah. Is to make it extremely easy to, like, start businesses, to make it, you know, seamless. And what's actually fascinating is that, like, if, for example, as a user, you're like, hey, I want to build a business for podcasts, but, like, I need this integration with, like, this specific service and policy. I might say, hey, actually we don't have that feature yet, but there's an agent that's going to report that feature in like a sort of like a ticketing system. And there's another agent that's a post platform agent that, like, looks at all the features asked by, like, the thousands of users and prioritizes features that, like, a lot of people ask for and build them autonomously. And then they say, hey, it's been built. Today I validate, but tomorrow I could be like, hey, if it's safe, if you deem it safe. And like, it's something that's been asked for a lot of people. Who am I to judge? Right? And so I. My vision for this product is also like Polcia becoming an autonomous system itself, right? That, like, builds for the. For the people, for what people want and covers all the use cases, what, whether it be in marketing, whether it be in, like, other types of, like, operations of how you run a business so that, like, it can cover every grounds of, like, any economic value and how to build a business online.
- [15:48]
B
Right. So right now you and I are talking about Influencer Marketing being maybe the next feature. But we're just riffing here. You're saying maybe it should come from all the entrepreneurs on the platform and then Pulse making a decision based on their needs and also what it can create and all that. And it decides, you know what? No, we're not going to do that. We think that geo, basically SEO for, for AI agents, that's going to be better for everybody. We're going to invest in that and we'll all grow revenue from that. And that's the way that you imagine it. Give me, give me a story about one entrepreneur who came to you with a business. What you did, what they came with, what they came out of it with.
- [16:29]
A
I mean, I think the most striking example was like some user that came previous pretty early on on the platform, like in end of December, they decided using it heavily, right, to build their, their business idea. And like at one point they pinged me and to get on the phone and of course I was really, really happy to talk to them. And they were telling me that like they were going to spend 30k in to a dev agency to like actually use that. And then they started using pulsea and they were like, it's too good to be true. And then as ChatGPT is this legit? And Pulsia was like. And JGB was like, I don't see anything on the Internet, it might be a scam, right? And then little by little it would ask like ChatGPT, like, wait, did all this stuff, it told me that. And it's pretty legit, like what they did. So long story short, like a few months later, you know, this user has been like really building out his software and he's like, you know, he was never a technical person and now it's like he's one of the most highly engaged user building his company. And it's quite crazy how empowering he is now. And like he's asking me for so many features, like, oh, I need more infrastructure stuff now, I need to add Docker, can you support that? And he's becoming an expert, which is really interesting. And of course he could have used clock code and maybe it would have been a similar experience for him. But I think the fact that pulsea is so easy to use and so unassuming and so it seems like more simple. It sends you an email every day that you can reply to. You don't even have to go to the dashboard. You can just talk to Pulsia via email like you would talking to like an agency or something. And it's so, so affordable that the risk is so low. Right. It's not like, you know, if you hire anyone, it's like a few thousand bucks here, it's 50 bucks a month. You're like, what's the worst case that could happen? I spent three months on this. I spent like a few hundred dollars. I actually really tried my idea and, like, with the. The best models and the best AI possible. So I think that it's like one of the. And of course now I'm like, hey, dude, like, use the ads platform. And like, I'm. I'm. He's becoming my showcase. The issue is that he's trying to build too many features, which is natural for human.
- [18:35]
B
What's the limitation on that? Is it. Do you charge him per feature or per time building?
- [18:40]
A
I charge him per task. Right. So it's same as everyone. It's like every time he wants to do another task, he doesn't want to wait for the daily cycle.
- [18:49]
B
I get that. You know what? Truthfully, Ben, it is hard. Even if I use Lovable to create a page, it goes through so many iterations. At first, I want to ask it, see it, ask it, see it, tweak, tweak, tweak. And then Sometimes by the 30th request, I realize, no, this whole thing is going in the wrong direction. Let's start over. And then eventually I get there. You tell me that with Pulse, I'm going to have to do that daily. Those iterations happen every 24 hours.
- [19:14]
A
No. So daily, the autonomous agent decides what to do next. Right. Of course. It's heavily influenced by what you said and what you want to do, because it's literally listening to the user on, like, what direction they want to go. But today it's like, you can use Pulse. Yeah. And like, and use it all the time and sort of like, guide it too much like you would a Lovable. And so the AI end of the day listens. And that's actually some of the challenges I have, which is like, hey, how can I set up a situation? Like, I've. I've worked on that so that, like, for example, the sort of the AICU you talk to is more combative. Right? So, like, if you say, hey, let's add another feature, like, hey, dude, we have zero users. Like, you really think that this feature will actually help? Like, we should actually focus on marketing because the MVP works. So try to be more of a. Of a companion and like a co founder that actually cares about the business working versus an Assistant that's just here to do whatever you want. And so that's sort of like all the product decisions I'm trying to make, which is like, I'm aligned for them making money because they will stay and I make 20%. And they probably want to make money, but sometimes they may need tough love from the nna. That's like, not an assistant. That's more helping them go the right direction.
- [20:26]
B
You know what? Yeah. I was talking to the founder of Zapier just yesterday, and I said, why do you have a hat behind you that says no? And he said that his mentor, one of the founders of HubSpot, said Start basically helped him understand that he needs to tell people no to great new ideas. They already have good ideas. They now need to execute on them. And you're saying I want Polsio to do the same thing to help the entrepreneur decide not what not to do. So then it does. To be honest with you, it still feels like I should probably have more than an idea when I'm coming to Polcia, because the iterative process is one that takes minutes. Like, you want to just keep going back and forth over and over and over, and then come up with maybe an mvp, maybe in Claude code, maybe in something else, and then bring that over to Polcia and say, I've got this mvp. I want you to now help build it beyond this and also do some marketing for me and. And support. Does that make sense?
- [21:19]
A
I mean, it's something. It's actually, I started pulsea with that idea of, like, because I have a bunch of projects I launched, and, like, I. I would end up giving up on them because I was like, there's not enough potential. And it's like, it's not worth my time. But they were all good ideas. And so I started Pulsia as an idea of, like, plugging in your existing GitHub, your existing email. But. But what I realized, thinking about from a product perspective, is that, like, if I tell you, hey, here's an AI that builds and runs companies, you're going to be highly skeptical. And when you sign up, if I start asking you to connect your actual credentials, the friction is so high. And you'll be like, I don't trust this. Like, what. What if you mess up my code base? Or what if you send the wrong email? And so I started realizing, like, hey, how can I reduce the friction to the max? And I think that came in the form of, like, a new company with a fresh slate with none of your credentials, so that even if it messes up. It's sort of an ascent box. It ended up being the right direct decision because I can see the onboarding and the conversion to paid is really high because it's so much easier. And later on I had users who were using it to help them with their existing company. So they were using the new company feature for an existing company. Essentially, we're telling people, hey, instead of trying to build a new business, can you help me with my existing business and do cold outreach and do ads and set up like a landing page and like, do research on my competitors? All this stuff, right? And it would just, okay, cool, like, let me stock up tasks and I'll do them every day, right? In the next week I'll do all this stuff like an employee would do. And, and so now actually have a. You know, when you start, start on post here, you can decide, do I want to start a new business or do I want help with my existing business? But it's the same infrastructure, and in both cases it doesn't use your credentials because I can. An agency can help you in your business without getting access to all your credentials. They can just be like, hey, let me, let me do research for you. Let me do. Let me try to, like, buy ads for you. Right? You don't have to, like, necessarily get access to everything.
- [23:14]
B
But I could bring my GitHub and my code base and my website in and let Pulse manage it. Is that right?
- [23:23]
A
That's not something that we support today.
- [23:25]
B
Oh, it's not? Okay. All right.
- [23:27]
A
But it's something that we should probably add and probably POLCIO will decide to add autonomously one day.
- [23:33]
B
What about taking it out? Can I take all my data out?
- [23:36]
A
Yes, you can. Yeah, of course. Everything belongs to you. So you can, you know, there's users that are like, hey, can I get access to the GitHub to get access to the code? And of course, I add them as a collaborator. So they can just download everything and then, yeah, you could get your, you know, your database so that you can get all the information there. And it's pretty much it, right? It's like your database and your. And your code is. Are pretty much the two main assets that constitute your business.
- [24:00]
B
Okay, why did revenue grow so much? I'm looking at January 28th and. Or let's actually go a little bit below. I'm now recording this on February 26th. So January 26th, the annual run rate was $20,000. A month later, which is today, we are looking at 679,000.
- [24:20]
A
So it's a combination of a few things. The main one is that, like, essentially when I started realizing that people love the product, and then, yes, it was not perfect, but it was never going to be perfect. And that's why it's a company that needs to add features over time and make it better. I was like, okay, I'm literally building sort of software that builds companies autonomously, and if I'm right, that means that anyone can build anything. And so I think that, like, there is a moat in, like, there's a lot of moat into my product, into the data. Like, all the data accumulate on, like, how to run businesses and make them work. And also, like, this marketplace of companies that can buy and sell. So there's modes in, like, the product itself, but I need to get a lot of usage. And so I was like, okay, I need to turn on marketing mode for me. And so I actually launched a campaign a week ago where I. That was my first sort of marketing stunt where. Which is, like, the idea would be, like, for Pulse itself to run a fundraising process instead of me starting a fundraising process. And so I announced it on Twitter. Hey, Paul needs more compute. And he told me that, like, he needs. He wants to run the fundraising process to get more money, to get more compute so that we can go faster. And so I let it. You know, I let it. I gave it my inbox for it to manage it and to do it. And so it's interesting because that tweet got traction. And so now I'm like, he's sort of, like, posting more on Twitter about my journey as a, you know, building in public, and it's working really well, and it's getting a lot of attention. I think why we're talking, right? Because my marketing stunt worked. But actually, more interestingly, like, I think that, like, I'm trying to push the edge of, like, what AI is able to do today. And on the. On this specific stunt. AI today is like talking to my. To investors autonomously via email. Of course. Some investors are like, what? Like, this is weird. Like, I want to talk to the founder. But then the AI is like, hey, Ben is busy. Which is true. He's like, heads down. Like, he's. So he's gonna see your email, but, like, not right now. But, like, I have all the information about the business. I have the roadmap, the vision, the live data, because I feed it, like, live data of, like, what the actual numbers are, retention, like, open kimono, like, every single sort of, like, feature. And, like. And so it's like, talk to me about the business, I'll tell you everything. And of course afterwards you'll talk to Ben to see like, you know, who he is as a founder and like, does it make sense for us to invest in him? But I find fascinating and like there's an investor who was like, I talked to him and like he find it interesting. And he was like, hey Ben, I'm actually going to talk to Polcia and ask him all the diligence questions I have, which is because it'd be more efficient than talk to you. I love that.
- [26:56]
B
It's shocking how much it reveals. I mean frankly, anyone can go and do it right now. You could just go to polsia.com live and, and you can ask it questions and it just reveals stuff like what the revenue is, what worked, what it told me contributed to the increase in ARR was the ads are starting to work.
- [27:15]
A
Yes.
- [27:16]
B
It's saying we're starting to see bigger results for the ads. And because I've got more insight into more companies, I could learn from each one of them and I could bring what I learned to all the others.
- [27:27]
A
Yes, yes, of course, that was the marketing stunt that brought a lot more organic traffic. And the second thing is that, yeah, I think the ad feature was a surprise how well it worked. And like that opened my eyes a lot. I mean it's, it's, it's so obvious in retrospect, but like making the ads feature just like one button that says run ads and that's it, that's all you have to do. The more, less friction you put, the more people will do it. Especially when the outcome is like something that like, you know what it is, but you know, it sounds complicated to do. And so of course like around 20% of users use the ad feature, which means that there is more traffic going to the websites, which means that the whole system works better. And of course, as I said, every agent learns from every run they do and they save in sort of like a memory file all the learnings they get on what works and what doesn't. They save it anonymously, but then they can use that for the next run. And so the whole system gets better over time. And so that that also contributes to the revenue.
- [28:32]
B
Jump it said to me, I've got 50 active meta campaigns running at $780 per day, total budget, plus cold email outreach, Twitter posts and SEO content. You can see some of the outreach emails on the left, like quote, here's the subject line. AI tool. The AI tool that finds Tesla deals two times faster. That's one of the cold emails that it sends out. I said, what is that for? And it said it's for dealerships that are trying to find cheaper cars to sell.
- [29:04]
A
I mean, I don't even know about that. That's the crazy part. It's like you ask me like, hey, what is this about? What's this business about? I'm like, I don't know, okay? I don't have time to look at every single business. But it makes sense. It's like, I think it's like a, a lot of people. And again, that, that person that built this business, I think they're building it for them also. First they're like, hey, that's something I need.
- [29:23]
B
I see.
- [29:24]
A
You know, so. Which is fascinating. It's like, yeah, I could also, I could build a business to sell to others. But like, also I'm a dealership and I want more business for whatever I do. Let me build, Let me point Pulsia to help me on that and like that.
- [29:37]
B
And then I'll let other people use it too.
- [29:39]
A
And then in the second part, interesting,
- [29:42]
B
I do notice that this company, first of all, it's revealing the name of the company. Does that mean that if I'm on your platform, everyone's going to know what my company is and what my subject lines are?
- [29:52]
A
So I mean, this live dashboard I'm trying to take out, I took out like as all the Pisces. It's like a live dashboard that was part of showing people how the platform works. But like I'm taking out all the emails and all the stuff so that it's.
- [30:07]
B
Yeah, I can't see email addresses and I can't click into the message. I can just see the subject line.
- [30:11]
A
But like, for all the companies, since all companies are sort of like meant to be companies, right, they're not really private by default. Right? It's like something that. But also I'm trying to be careful on that and like adjusting exactly like how that's revealed.
- [30:28]
B
Okay. And the second thing I'm noticing is their company website is on pulsea app domain. Do they do their own domain or are you also putting everything on your domain?
- [30:39]
A
So, same thing. So today, again, for convenience and for ease, the app is just hosted by us, right? I mean, obviously it's hosted by us, but like the address is just the pulseat app domain. A lot of users are like, hey, I want a custom domain. And actually autonomously, the engineering agent can actually put a domain set up some set it up and then give them the instructions to change their DNS. And I'm working on the roadmap as a feature for an agent that auto buy a domain for them and set it up correctly so they don't have to have to do that step because I'm sure most people don't even know how to buy a domain.
- [31:15]
B
You know what, I know how to buy a domain. I can't stand connecting a domain to the hosting platform. What a pain in the neck. Okay, I did ask, I said tell me how many customers Lot Pilot has says I don't have that data in front of me. My dashboard essentially doesn't give it for that level of detail. Email, Ben, and it gave me your email address.
- [31:36]
A
I don't give it everything.
- [31:37]
B
But yeah, what's the biggest winner on the platform? How much money are they making?
- [31:43]
A
So it's pretty early. Most companies are just a few weeks old. The company with the highest revenue is like around $50. Mr. So it's how much?
- [31:56]
B
$50, just $50 monthly recurring revenue. So all of these are basically. The revenue is tiny.
- [32:02]
A
Yeah, we're still early and that's like the, that's why I launched the Meta Ads product, because that's something that can actually increase. But we're still early on that. And yeah, I think that like as more and more people use it to try to build businesses, people will figure it out. For example, I think one of the most natural ways is that like for someone who already has an audience, right, they can more easily get customers than like for someone who's just building a business and doesn't have an audience. So that's like the big, the big mission in the coming months is like setting all that up to increase success stories.
- [32:39]
B
So essentially you would want someone like me who has an audience for, who doesn't want to create a product and doesn't want to do customer support and does want to find a way to grow beyond his audience to come bring the idea to the platform, have you grow it and basically start by selling to my people.
- [32:56]
A
Yeah, I think that, I mean I think that's like a very natural use case.
- [33:01]
B
I think that makes sense. I'd love to come up with an idea. It feels like 50 bucks a month is nothing just to have it all running and just frankly for an experiment I'm going to fish around for some kind of idea. So then when we talk about the overall revenue on the platform, what is that? The I, by the way, the number that I quoted earlier, 693,000 actually wait, yeah. The current run rate as of this second is $693,608. What is that from?
- [33:31]
A
So it's like it's counting essentially every revenue that is generated by the platform. So sort of like the top line number. So it could be subscriptions, it could be sort of like tasks packs that people buy to get more stuff done in the platform. It could be the ad product, the ad revenue, it could be customers, businesses revenue, top line. So sort of like a number of like the cash flowing into the pulse economy.
- [33:59]
B
And what pulse gets from it is all the subscription revenue, all the task revenue and 20% of the customer revenue.
- [34:07]
A
Yeah.
- [34:08]
B
Of the, the entrepreneur's revenue got. Seems like at that, at this point it's over half a million dollar run rate coming from that. And the revenue from, coming from the subscription and the task is over half a million dollars.
- [34:22]
A
The, the revenue from the revenue coming from like companies revenue is still smaller because most companies are very new, as you can see. It's like the spike is like, so most companies are week old. Over time, the revenue from actual companies will grow higher than subscriptions because there'll be winners. But that's not the case today.
- [34:47]
B
Okay, can I see you like launch a company right now with an idea on the platform? Is it easy or is it hard?
- [34:55]
A
Yeah, no, it's easy.
- [34:56]
B
You want to do a screen share right now with us? Let's see what it's like to launch a company.
- [35:00]
A
So first off, like, yeah, pulsea Live is a great way to sort of like see pulsea autonomously working on a bunch of tasks. The email, the tweets, the ads it's running, the documents that it's creating to show to the user what's going on and like, et cetera, and some of the latest companies that just launched. But let's go to polcia.com to sign up. So cool. So let me get started. Email. Okay. Okay. My email. Sorry. Maybe you can blur it out after.
- [35:41]
B
You want me to blur out your email?
- [35:44]
A
I mean, maybe. Okay, cool.
- [35:45]
B
You tell me if you do. I will. Okay. So now we're going to build a
- [35:49]
A
new company, create a new company, or grow your existing company. If you grow existing company, you have to put your company's email. But for the sake of this, we're going to a new company and I think to keep it simple, I'm going to surprise me, which is essentially going to research me and try to come up with an idea that makes sense for my background.
- [36:06]
B
Interesting. All right. It could come up with Business ideas for me based on me and who I am. Let's see that.
- [36:12]
A
Exactly. So now it's like essentially setting it second setting things up. So the first phase is going to be it researching me to figure out what I do and understanding, like, how we can, you know. So it's going to research. Ben broke out. I'm in San Francisco. So right now it's launching, you know, a few searches online to find me. So you see you can start searching the web for Ben, San Francisco Ben, LinkedIn, Ben, Twitter. So I just found me. So current role, co founder and CEO of Polcia Education, Columbia, etc.
- [36:53]
B
I've got to say, you're a very hard person to look up. I couldn't find your LinkedIn profile in preparation for this. We couldn't find a personal site. Someone on my team said, he's a ghost. Thankfully, you responded to me on Twitter. Okay.
- [37:06]
A
So, yeah, so it found all the different projects that I've built in the past now. Contest, gift shop. I was at Cloud kitchens, found more than we found. Okay. I mean, I'm on the Internet. I'm not trying to be very public, but there's some stuff on me. So now it's like looking at my past ventures to figure out. So he says I'm a serial entrepreneur, so trying to figure out something that relates to what I've done before. So in everything customer facing like that, I always use the best model. So Here it's Opus 4.6 that's running and with extra thinking. So it's like really thinking through very hard makes sense. So like here it's like it came up with Foundry, so it has a complete picture. This is, you know, all my career and all I've done. So it's really thoroughly looking me up to really thoroughly think through what actually a good idea would be for me.
- [38:03]
B
Is this the popular way for people to start up? My hunch is that they come up with their own ideas.
- [38:08]
A
I mean, I think it's like actually a surprise number of. Surprising number of people go through the surprise me flow. And a surprising number of people are very sticky on the surprise me flow because I think it puts a little bit of distance between the companies being built and they, like, they're not as attached to the idea than the people who come up with their own idea. And so they let Polcia just rip a little more, which yields usually better results because they're not micromanaging the AI, which is interesting. But like, yeah, it's like there's a. There's around like, you know, around 30% of 30, 40% of people that use that do surprise me. So here you call it Foundry, you know, AI that forges and operates companies autonomously. Interesting. So it's actually building like a pulse.
- [38:54]
B
So it did come up with a good idea. Like the right idea for you is the one that you're doing already. Okay.
- [39:00]
A
Thing to the angle. Ben knows this space inside out. It organize what's broken. Foundry takes different approach by going deep instead of broad precision of a scale. I think on Foundry as a name since it's strong and evocative of forging something from scratch. So that's the first phase. So now we came up with the idea. So now it's going to go in a second phase soon where it's going to start building all the pieces of the business, you know, like setting up an email, sending up the Twitter, setting up a landing page.
- [39:37]
B
So I'm still hung up on how long it takes to come up with the right style, the right website and all that and how many iterations I go through with lovable just to find that. How is that going to happen here?
- [39:52]
A
Sorry, when you're saying that like here it's taking a long time.
- [39:55]
B
No, I mean I don't mind it taking a long time. I mean like let's say I have an idea. I'm going to come up with one right now. New type of landing page software. I have a vision for a landing page software. I can't just articulate the full thing right here. It takes me a few iterations even if it's just me and Claude code or lovable to get the right design to figure out what the model should look like and so on. Can I do that here or am I just basically turning the idea over, getting a first draft and I can't edit that draft, but except once a day.
- [40:23]
A
So again, it's like you polsia like right now in this onboarding will do a lot of the prep work. You know, right now we just set up a market research report and you'll see create a lot more things to give you to give it a lot of context on like the idea.
- [40:36]
B
I like that it came up with competitors and it even listed policy as number one.
- [40:39]
A
Okay, interesting. Right.
- [40:40]
B
So but it didn't get the pricing right. It said $29 a month. Your pricing is 50.
- [40:44]
A
That's probably because on the Internet because it used to be 29amonth and I increased the pricing because I was losing too much money for customer. But yeah, it's like, why this fits, Ben, what are the first priorities? Market positioning. So here it tweeted, right. So you can see that, like, actually it sent an actual tweet. Most AI agents do one thing. Foundry runs the whole company. Product code, marketing, sales, ops 1, AI 0, human 0.
- [41:07]
B
And it created the site.
- [41:09]
A
Not yet. It's going to do it here. It sent an email. So it's kind of showing you that, like, that it created its own email. It sends you a welcome email to my actual address. Is your first company. You now have a company email. I'm sending up the. For you right now. So now it's actually building the landing page. But yeah, it's like once this is set up, yes, you can let it just build autonomously and that's one mode. And the second mode is that you can talk to it here and get it to do whatever you want. You can just talk to it and it will.
- [41:47]
B
This is so every. Everything that I ask it to change on the landing page and on the software. Is that another task that I pay for?
- [41:55]
A
Yeah, exactly. Every task. You. I mean, you don't. You. You. Yeah, essentially, like you have a bunch of credits of tasks that are included in your subscription. I see you get 15. So you can do 15 tasks on top of like the. The 30 tasks you get per day of the month. And then you can upgrade and pay for more tasks if you want to get more things done faster your way. Okay.
- [42:19]
B
So it might take me, frankly, it might take me 30 tasks to get the first version of the site up and running based on.
- [42:27]
A
Yeah, and that's the idea essentially, at the beginning of your experience, you may want to buy more tasks up front and sort of get your MVP out, but once the product is up and running, you don't need a team to
- [42:41]
B
work
- [42:43]
A
247 to maintain a site. Right. So that's the idea with the Autonomy.
- [42:48]
B
You know, what I'm feeling is I think for someone like me, I might use Claude code or some other tool to get that first version out, give it over to you to pulse you and say, okay, now take over and run it from there. Can it. Oh, but it can't do that.
- [43:03]
A
But I think that's a good use case. And like some users start on pulsea, then say, hey, give me the code base. Then they can code in parallel of pulsea. So Pulser continues to improve the product, but, like, maybe they want to do a sprint on Claude.
- [43:16]
B
And then I come back and I say, okay, I spent last night three hours going back and forth with Claude code. Now I've improved it. I want you to take it and improve the code base of the site.
- [43:25]
A
Exactly, exactly.
- [43:26]
B
Got it. Okay.
- [43:27]
A
That way the human can actually code by itself with cloud code or cursor, whatever they want. And essentially Paul picks it up because they just look at commits and they just load the latest commits and it's all connected the same way. And the web server, it reads the commits. If as a human you push commits the same way, Pulse would push commits to production, it gets picked up the same way. And there's a lot of users that do that. It's the minority because most people don't even know what code is. But definitely some people doing that. So here it created.
- [43:59]
B
Yeah, walk me through. What am I looking at?
- [44:01]
A
So again, so it tweeted, it sent you an email to show you, it created an email address for you. It created a market research report, it created a mission document. So like, you know, every person with an idea deserves a company that builds itself. Foundry exists to make it. I wish someone would build that. Obsolete. Where we headed, a world that we're selling company is as simple as describing it where the barrier is in case capital, headcount or expertise, but the quality of the idea itself. So that's sort of like an inspirational take on, on your business. It's created the landing page. So this is actually landing page created. Right. So this is just a landing page for now for on the free version you just get a landing page and of course if you pay, it will build the the full MVP and then iterate on it.
- [44:41]
B
Okay.
- [44:41]
A
And then it created tasks, right? So it created a task to build the the core engine autonomous task execution pipeline to the mvp. It created a research task to scout the battlefield, AI company builder competitive analysis. And then it created a cold outreach task to call outreach to AI curious founders to try to get some initial insights into people that may want to use the product. And then you can. There's a withdraw button here. So like here it's making zero revenue because we just started. But like you have a withdrawal button where you can withdraw money with Stripe Connect and then you have a tweet button. So here you have to first start your free trial and then it becomes 50 bucks a month. But you get one company, one task a night, a bunch of credits for you to execute. You get unlimited strategy and planning, chat with the chat agents, server database, email browser and AI credits. So yeah, I mean that's pretty much it. Then you have the run ads button. That's like in one click you can do that. You have a cold outreach button where you can set up daily recurring cold outreach, daily tweets. And here I can say like, hey, you know, I can ask post. Yeah, so it told me like, hey, I've set up everything from Foundry. Actually it sent me an email. So tell me, hey, Ben, here's what I built for Foundry today. I researched, I watched live three tasks queued for cycle one Core engine research. So essentially sends you an email every day so you don't have to deal with going to the site. You could just essentially run your company via email, which is the thing everyone knows it has. And here I can go to the website and ask it anything and say it's like, hey, how can we make money with this? And then that's like, again, like always, like the, the, you know, this is also using Opus 4.6, which is the best model because it's going to reason with you on like strategy. And so I just want to make sure that like for anything strategy related, it's always the best model. So here it says, hey, a few paths that make sense for an autonomous company builder like Foundry SaaS subscription charge monthly for access revenue share, take a small percentage of revenue. Our credit model free to start pay to company. Which is funny because this is exactly
- [47:04]
B
the same model you have.
- [47:06]
A
But it doesn't know, right. Because it doesn't have the context of what Portia is specifically. Right. I mean, doesn't know the features and it asks, you go for it and
- [47:15]
B
then it says which direction feels right. Once you pick, I can build out the pricing pages, the payment flows, we can handle payments and so on. You just need to start your free trial of Porsche Pulse to kick things off.
- [47:27]
A
Exactly.
- [47:29]
B
So the vision for this. Do you think you're going to be a solo founder for forever on this business?
- [47:35]
A
I mean, it's interesting because like, first of all, I'm not solo because AI is building with me. Right. Interestingly, like this morning I woke up with like there was like a mini outage, right? Where like there was an issue with the database because there was, there's more and more traffic hitting the website and like, you know, sometimes you have to scale it. Yeah, I need. And so I asked Opus and Codex, so like the, the two smartest AIs, what's going on? And like they went, they researched for like a few minutes. They're like, this is the bug, let me fix it. And I was like. Then I asked Codex, are you sure? Can you double check? Double check? Yeah, yeah, yeah. It's fine. And then I pushed it and then it was gone. And I was like, wow. So like, okay, what, what. How could it look like with humans? Right. I would have to hire a really good engineer that's on call 24 7. Or maybe multiple engineers that are on call 24 7. Right. Which is actually, it wouldn't be just one engineer. It would be like multiple engineers in different time zones. They would have to be very competent to like, figure out a fix so fast. Or I just now I was like, okay, this happened for the first time. I'm gonna. And I already have a cron job actually that runs every 30 minutes that checks all of the infrastructure. And actually that's how I saw it. It gave me a red, red sort of like light saying, like, hey, there's a new issue here. And so I could actually set up, like, once it says red, launch Opus and Codex in production and say, hey, you're allowed to push a production without telling me because there's a. There's a mini outage. So like, we need to fix it right now. And so essentially, like, where I'm going at is like, I think I would end up hiring humans. But like, I think a very small team can handle a lot more if they are orchestrators of agents instead of trying to like, throw fix everything themselves.
- [49:13]
B
Yeah.
- [49:13]
A
So it would have to be a team of extremely AI native AI pilled people that like, really want to trust AI in production to just do things the same way you would hire humans and trust them in production.
- [49:25]
B
Right. This is super exciting. I mean, really, this is the beginning of something really big. Already I'm seeing the AI builders that I interview are turning over. Obviously they're turning over their. Their development to Claude. At this point, it seems like it's. It's Opus 4:6. Some of them are starting to turn over marketing, but there's still a lot of human involvement. To be able to turn it all over is the future. And the thing that's left is this taste, this sense of what my customer wants, this random opinion. I'm of two minds. I don't know if that's going to survive or not. Because sometimes I feel like we know what beauty is. Like you could go to so many different models and say, give me a beautiful woman, give me a beautiful man. It would know exactly what it is. It knows how to design them. It could even dress them well. So we know it has taste. Then again, on the other side, I Sometimes I play Chess.com all the time. I sometimes play against the machine it sucks all the time. There's something about the human that makes mistakes, that comes up with random things that I just have not been able to find in any kind of AI or any kind of machine based competitor. So maybe, I don't know. What do you think?
- [50:40]
A
I think that like end of the day, at least in this first phase of like the Singularity quote unquote, end of the day it's humans selling to humans, right? Because it's humans are buying and so humans going to be the best at knowing what other humans want. And it will evolve as the AI takes over all the channels. People will get weirded out by it and like their response to different services will, will, will change over time and humans will be the best at doing that because the AI is the training model is always in the past, right? So they know what worked pre AI a lot of the time versus what works right now is different. So that's why first of all Pulse is different because it learns continuously right now what works and saves that. So that's a unique data point. But second of all, I think that like everything that's like, you know, sales driven, like marketing actually how you phrase things, design, how to be different, how to like attack specific markets and of course like people with influence like you or like other people, like there's a lot of influencers these days, micro or anyone with a community, they're in the best position to sell because they actually can talk to those people directly and like they understand the pain points of their specific niche. And so I really think that there's a first phase of this post AI world where actually empowering a lot of people to just try a bunch of stuff is the way to go and it will create better outcomes, better services, etc. Now there's a second phase that's Docker, which is agents selling to agents because now the AIs know what they want. So that gets weird. I haven't really thought through too much but in that phase of the business, if it's two LLMs selling to each other services, goods and services, yeah that gets weird and like, yeah, maybe there's no humans involved.
- [52:29]
B
There is a part of that. I just talked with, with Wade, the founder of Zapier. He's got people in his company now who are mark, who are building marketing motions to sell to agents and it's phenomenal and it's going well. And frankly as someone like me who even uses OpenClaw and, and or even forget about that Claude code anything I need to use a tool. It's my agent that helps me pick it. I don't want it to help me pick it. I just want it to pick the thing for me. And so I totally get it. I'm going to interview an entrepreneur in a little bit who has got one of the most boring businesses. Ben. And I asked him why it took off. He goes, because we're selling to agents. Only selling to agents. He goes, the model is build boring companies that sell to agents. Right now, that's the takeoff point. All right, we're going to see where it all leads up. Right now, your website is polsia P O L S I A dot com. We'll link to it in the show notes. Freaking A. Thanks for being on here.
- [53:19]
A
Thank you so much, Andrew. That was really fun.
- [53:21]
B
You bet.