Latent Space: The AI Engineer Podcast
Original podcast
Watch the full show on YouTube: https://www.youtube.com/watch?v=Yw-m0PI2Atk
8,198 words · 2 speakers
All right, we are here in Lanespace again with Ben from Polsia. Welcome. Congrats on the epic run that you're starting.
Thank you.
What's the ARR standing at today?
We crossed a million like a few hours ago.
Oh shit, okay. This is the first part since you crossed a million. Congrats. Do you like, do you have any celebration plan or you just locked in and going?
Not really, like I'm honestly like the growth is getting insane. So I mean, not insane from my perspective at this point. So I'm really working on scale and like making sure that like every user has a great experience and as I scale the number of users.
You know, that's the other thing, right? Like there's a lot of marketing and all that, but like ultimately make your users happy and then they'll be your marketing, right?
Mm hmm.
That's the only sustainable way to do all this. But let's rewind for people who haven't seen your stuff. What is Polsia? When did you start working on it?
So Polsia is an AI that builds and runs companies autonomously. So you give it an idea and Polsia will go ahead and like do, you know, product, coding, marketing, sending emails, launching meta ads, campaigns, do competitive research. Like it can essentially do everything that you would do as a founder if you were to attack a sort of like a new, a new, a new company. But does it autonomously, meaning like every night it wakes up and there's a CEO sort of like instance that will decide based on is there any bugs, like how's the business doing? Do we have any customers? Do we have any paying customers? We'll decide what to do, which tasks to do, and then we'll execute on it and send you an email every morning,
summarizing the state of the business, what it has done during the night to get the business into a better state and what it plans to do the next day. And of course, users can guide it, can give it feedback either via email or via the dashboard. It becomes sort of like a daily routine where like every day you get an email from your Polsia instance and then you can guide it into realizing the vision you have for the business.
Yeah, amazing. I think that's a very, very big claim. I run two small businesses myself and I have a lot of those pain points and that I hire humans for. And they're also not that great sometimes. My humans are great, but everything could always be more automated. So let's get into the product. I think let's show people what it is so people can concretely get a sense. And then we'll talk about the building side.
Boom. So this is a dashboard I created a couple of weeks ago to demonstrate Pulse AI working on now over a thousand companies at the same time, right? So this resembles a lot how the dashboard looks for customers. So we have Pulse AI, which I sort of humanized into this little mascot that is currently working on various tasks. And so you can see it's sort of like thinking process on what it's up to. These are tasks that are actually running right now. So it could be engineering tasks. Right now, it's mostly engineering tasks, apparently. That's what it's working on. But it can be research tasks, you know, marketing tasks, Twitter.
A quick question. Do you have an estimate of like, how much is engineering? What percent engineering, what percent marketing?
Engineering is like probably 50% of the tasks. And the reason is that the growth has been pretty recent. And so a lot of companies are in the initial phases of their business. And so they spend a lot more time on engineering to get the product at the good state where the customer is happy with marketing it. But there's also a lot of usage on like, so right now you can see there's a cold outreach task that's happening right now. Send cold emails to six new verified leads. And so essentially, like on this dashboard, it's supposed to be like to represent how your business is doing. So you get sort of like your business metrics, tasks that are gonna, you know, either happening or like that will happen. This is your sort of social media tab.
Right now, it's mostly Twitter, but we're gonna introduce other social media platforms later. So these are real tweets that are like being tweeted directly on our Pulse AI HQ, where Pulse AI is repping different companies and why they are, you know, they are good companies. This is email, so every Polsia instance gets an email address. And that email address can send and receive emails. So in the past 24 hours, almost more than 2,000 emails have been sent or received. And so that enables the agent to be able to answer the customer support or to do call outreach and then reply to those call outreach to engage in conversations. You have documents that are a great way for the agent to show to the user the market
research report, or mission statements, or other types of reports that can be used by other agents, but also for the human to understand what's the state of things. And this is a new feature I added a couple of weeks ago. Ads, where essentially you can tell your agents to autonomously create UGC ads. So here you can see that like-
Oh my god, this is all generated. I can't hear the video, but I can't hear the audio, by the way.
Sorry, what?
I can't hear the audio.
Okay, got it, but you get the idea.
That's good enough.
So those are AI-generated ads, but that are pitching the product. And it's a very low-cost way to get going, because obviously, running ads, it's a content game, and so it's a very costly process and time-consuming. And so Polsia is bringing a first version of that. And of course, in later phases, we will introduce influencer content. So you'll be able, in one click, to contact influencers and get the content of real humans to market your product as well. But this is the idea of fully automatising a function of a company. So yeah, so that's how the dashboard works. And then here, you can ask the, in this public live dashboard, you can ask Polsia,
hey, what are you up to? And it has this Polsia instance as access to everything that's going on right now, every metric. So it tells you, hey, A task running right now, building a smart recommendation engine for sales data, adding off two pounds, wiring up a consultation booking page for Pow AI, building an email drip campaign with analytics, running meta ads and call outreach. So it's doing all those stuff. And obviously, it's a bunch of different agents that are working on different topics. But I'm trying to really humanise it and make it super simple and comprehensible.
Amazing.
This is a, so first of all, I really like the design. It's very kind of minimalist. I think there's a, there's a kind of like a design element in you that gives it some taste. And this is, everyone can see this, right? This is posia.com.
Yeah, this is like a public page that's meant to demo how Polsia works.
Okay, but is it, when you say demo, is it like real or you know like?
This is all real. Oh, sorry, sorry.
All real, right? All real data. Okay, yeah.
This is all live data, real data.
When people say demo, sometimes they, you know.
Yeah, sort of like showcase, showcase.
A mock, yeah, yeah.
Actually, when you go to pulsia.com, well, this is actually where the demo would be, but anyways.
Yeah, okay. And so, okay, let's go back. Yeah, I, so first of all, I like the design. I think there's a lot of interesting things here. I just think like, so this is abstracted level, right? This is Polsia running other companies that were started on Polsia. And you are just giving like a superset of what everybody's trying, right?
And so you have a thousand.
There's 91K human messages. That's a lot. Is that, is that surprising?
I mean, it's interesting because that was one of the engagement tricks that I was tracking initially. You know, like sort of like to understand like how much people were liking the product. And it's actually insane how much users are talking to their AI. So sort of like in this panel, you're sort of like talking to your AI CEO here, right? In the experience of Polsia. And so, but people on average spend, uh, send 15 messages per day to Polsia. And but when you think about it, it kind of makes sense because, uh, you know, the, the 1100 people that are like building companies running on Polsia. They are really like, you know, it's oftentimes like an idea they, they had for a long time. Then maybe they didn't have time to do, or maybe they didn't have the motivation to do.
And suddenly they have this AI agent that like wakes up every day and like gets them back on track and like never gives up. And so they sort of like get back to, okay, you know what? I'm really gonna do it this time. And so they spend, you know, a lot of time talking to their AI co-founder, AI CEO to build the product, right? To talk about the strategy, to talk about, you know, how to build it, what features should we add? And and obviously, it's it's it's in the backend. It's it's I always use the best model. So here it's Opus four point six. That talks to you. So of course, it's expensive. But I want to give the the this. This is the agent that's going to decide on strategy that's going to advise the user on like, what is the best plan to get to a successful company?
So I think it's important to give it the best, the best AI, right? The best reasoning.
Yeah, okay.
And like, there's so many questions, right? Like, but I don't want to interrupt your demo. Was there was there something some other screen that you wanted to show?
I mean, I I I think what would be fun is is going through the actually experience of like, how do you onboard on Polsia?
Okay.
So.
Do you want me to give you an idea?
We can do that. So essentially, when you when you get started, you get you get two choices. Either creating a new company, which is sort of like the default path for most people. But there's around twenty twenty five percent of people that like use it to grow their own company. And and the reason I added this feature is because I was noticing that a lot of people were using Polsia for their existing business, right? Because you can tell Polsia, hey, go do cold outreach for my business. Hey, create a landing page for my business and collect emails and leads and run made ads for it, right? So you can actually, it doesn't have to be a new software business. You could just use it as sort of like an environment where you get agents to work for you, right? But in a safe environment, because it's not on your computer, it's sort of like, it's all provisioned and secure.
Yeah.
Anyway, so that's what.
Not like an open claw, like type of situation.
I mean, the way I look at it, I mean, Open锁 is an amazing product, an amazing open source community and like, kudos to Pete for building it. It's amazing. It it is a product that's I would see more like a an Android environment, meaning like you can configure exactly like you want, whatever model you want, it's it's for your tokens, it's your subscriptions, Telegram, WhatsApp, like whatever you want, right? And then it it it's open and it's very open ended and super configurable. What I would see Polsia, and at least the idea is making it more like an Apple ecosystem, right? It's not, you know, sort of like the same vision of Apple, meaning let's keep it extremely simple, let's provide everything to the user and let's focus on like a good, simple
user experience so that like anyone can use the extreme power of the models that Entropic and OpenAI are building, but provide a super simple experience that's like is less intimidating than maybe an open flow is, right?
Yeah, yeah, yeah. Do you want to get super meta? Do you want to grow my podcast company?
Yeah, of course. Do you want to do that?
But I'm just kind of curious what it does. Latent dot space.
Latent?
Dot space. It's a media company, which is an easier company to grow, right? I mean, there's no like physical product, there's no software.
Let's see what happens, right?
We just make content, right?
The people, people watching will see this and then they will be like, okay, well, I wonder what how is doing. But obviously, if you help me grow, I'll be very happy and
I'm in domain expert now on this as well.
Sounds good. You know what, like, that's it, that's, I mean, I love it. We don't have to like, I always do like demos on like my own account and my own stuff. So this is great. So essentially.
Yeah, good.
So what's happening right now is that Polsia is first researching latent space, right? So it's figuring out, you know, actually it's all, it's run by you, right? Sean Wang, Alice Yorinaldi. So it found you, it found the product. It's kind of trying to figure out why, you know, you know, it's like, hey, he's the co-founder of Polsia, which is us. So Polsia is figuring out also who I am because I'm the customer who signed up to do this.
Technically, you're not co-founder because you're solo.
I know, I know. Maybe it's because I put it on, I don't know, Twitter or something, I don't know. I also needed that, maybe.
But it's right now, it's understanding, okay, interesting. Actually, it says this email might be a test account, which makes sense if Ben is testing the product himself. So it's kind of actually crazy how Opus 4.6 is so smart and is able, I didn't tell him anything, obviously, right? And just seeing that it's a Ben dot Braca, he knows it's me, he knows I'm the CEO, because he researched me. And then it figures out that like, it's a test account, because there's like, I did a plus, et cetera. And yeah, it's like, okay, interesting. Maybe he's affiliated to it, maybe he's not actually affiliated. He's trying to onboard a business he's not actually affiliated with, right? Isn't it interesting? So anyways, long story short, this is the onboarding where like, Pulse here, because you give it your business, trying to figure out like, okay, what is, what is this business?
How is it related to the user? And like, clearly, I'm supposed to like, make the business work. So like, what's sort of like, what are the next steps, right?
Okay, so we'll let it think for a bit, but yeah. Anything else that we should be looking at?
No, I mean, right now, it's like searching the web. So like, it usually takes a little bit of time.
Yeah, so let's talk about like, you know, any sort of technical elements that might be interesting. So first of all, you only worked on, you started work on this like a couple of months ago?
So I started working on it in the beginning of November.
I see.
But I had spent a lot of time during the year, building different types of products, like AI coding tools, marketing tools, using AI. And I realized that like, you know, I was essentially like after, I realized it was so easy to build anything. And like, also, AI was capable of doing everything. And like, AI was my co-founder. And like, AI was like, you know, essentially like, I was spending 16 hours a day just talking to AI all day. And I realized, hey, like, maybe I could build a platform that would do it all, right?
And so that's sort of like what came about with Pulse here.
And, and yeah, and then I was like, you know what? Maybe I should build a platform that can, can help me build businesses without having to, to do all the work.
Yeah, I mean, that's, that's incredible. And I think like, that is the future. I didn't expect it to be here.
Part of me was like, okay, is it gonna like, uh, register Stripe Atlas for me? Is it going to like, set up like my Mercury account, you know, all these, all these other things that are part of it. But I, I actually, I think there's a lot of PMF for existing businesses. Like you, you might actually find that like, your biggest customers are just people with existing businesses who need more agents. So what was the hard part? But what, what were you like?
Come back to this, we'll come back to this later, because I guess taking a little bit of time.
I mean, I think the, the hard part, I mean, actually, like, the hard part wasn't really what to build. The hard part was like deciding what not to build, right? And it's like trying to keep it more simple. Because as I was building it, um, I was sort of like figuring out, okay, so, because when I started building it, I was like, okay, from, let me build it for me. Like, let's make it something that like, I want to build for myself, and how do I want it to work? And initially, I was like, well, what I'd love to do is connect all the different projects that I've built on GitHub, right? And all my GitHub projects that I abandoned, I'd love to hook them up to an AI that would then
take them as they were, and then start marketing them, and like, maybe fix bugs and improve them, right? So that was the initial genesis of it. But then as I was building it, I was realizing, yeah, but like, if you onboard on like, a, a platform that claims that it can run, uh, in an AI that runs your business autonomously, first of all, people will be highly skeptical. And second of all, if I start introducing in the onboarding that they have to connect their email address, or like they have to connect their GitHub, it sounds very sketchy, right? And like, it will add a lot of friction to the onboarding. And so what I thought is like, hey, you know what, like, let me, let me strip out all this complexity and like, strip out like, all this friction, and let me provision everything for them.
And so little by little, like, actually, it was more like taking away feature. Actually, like, in the codebase, there's still like, a lot of features that relate to being able to customize Pulsar a lot more, that, that Google this features initially, because it's easy to add features, right? That's like the easy part. The hard part is like, how do I keep it simple? How do I make it tasteful? So that people are delighted, right?
And so I stripped out all that stuff like, you know, you before you could connect to so many accounts and you could do so much more stuff. And I was like, let's keep it extremely simple. Let's go to the core and let's make it so that anyone can build it. So what so that my my ninety nine year old dad is using Polsia, but like I essentially set it up for him. And he receives an email every day and he knows that he can just reply to the email and tell the AI what it wants it to do. And he knows that the next day it's gonna get an email in the direction of what he asked, right? And so I think that's the sort of like the the core tension now. I think in this age where you can build anything, it's like, you know, what are the ninety nine things that you say no to? And what are the things that you double down on and really polish and make perfect?
Yeah, totally.
And like any, and I think some integrations are like very important, right? Like, like you have the email, Twitter. I'm just trying to figure out like, you know, is there a particular thing that you had to make some kind of trade off? Where you're like, well, the right version should be this thing, but we're gonna do it the simple way for now.
I mean, I mean, there's a ton of like things that like I would wanna improve that like the users are asking for, right? For example, like right now on social media, it's only Twitter, right? And the reason I did that is because Twitter is super easy because they kind of they allow like accounts that are like more AI generated.
Automated. This is you tweeting out stuff, right? Yeah, I think the API is a bit weird, right? Because I think it's very expensive if you start automated like messaging people.
So I don't message people. It's like this is like tweets to a feed, right? To set up a specific account. But Twitter is text, so it's very easy. It doesn't take too much time, but like obviously I've had a lot of users are like, hey, I'd love to connect my Instagram. And I'd love for you to post content on my behalf that I would approve. And then like maybe respond to comments and respond to DMs.
That stuff that will come, right? It's like, okay, this should be social media and not Twitter, right? It should be like everything related to social media. Same thing for ads, right? It's like right now, it's just UGC content for Meta, but it should allow you to manage Google, Google Ads, TikTok Ads and every other network that exists. And it should be able to manage your spend. You should just say, hey, my budget is this, but it should spend it on all the platforms autonomously, deciding everything related to your business, right? So those are some of the things that are on the roadmap. But like, it's, you know, actually like when I was building Polsia at the beginning, I was using a lot of Claude as like my co-founder.
It's actually very interesting how Claude adapted a lot to the way I act and the way I think. And at one point, I told him, I told it like, hey, I want to ship in one week, like in one week on this date I'm shipping. And then every day would tell me like, I would say, hey, what do you think about this feature or that feature? And it would tell me like, dude, stop talking to me about features. You should ping in like three days. Like, stop it, like, just make sure it's ready for launch, right? And I really loved that. By the way, Anthropic did a really good job on like the chat, sort of like experience of talking to an AI. But I think it's, my goal for launching Polsia was like, okay,
what are the most fundamental features that are sufficient to launch an online business? And can delight people in a simple way? And then if there's product market fit, I'll build all the features. Because for a long time, I didn't know I had product market fit, right? So I was like, you know, you can build all the features you want, but if you don't have growth, you're losing your time because you're gonna get demotivated at one point and it's gonna be a dead project on your GitHub, right? And so.
Yeah, totally. We'll be checking back in.
By the way, it built the whole thing.
Okay, so take me three days. What should I look for?
Okay, so essentially, like, because you said you wanted to, like, build a... So essentially, let's read what Pulse told you, right? So I've set up everything for LeightonOS. I researched you and Leighton Space market position, sent you a welcome email from LeightonOS Pulse app. So it created this email address for you, right? And sent you an email, I mean, sent me an email to my address. Hi Ben, this is your first email from your company, LeightonOS. You have now a company email. And so what else did it do? It created a mission document. I created a landing page. So we can... So first, it created a mission document, right? So every software engineer is becoming an AI engineer. Leighton Space exists to be the place they go first, the voice they trust most, and the community where the best ideas collide.
That's better than mine.
I know, right? It's so good, right? What we're building? The definitive media platform for the AI engineering movement. A podcast, newsletter, and conference ecosystem that gives 10 million builders the depth, technical depth, industry access, and P O network they can't get anywhere else. Where are we headed? Where are the people building AI have a home base? Where the newsletter breaks news before Twitter does. Where the podcast is the first call after a funding call. Where the conference is the room every serious builder needs to be in. Leighton Space isn't covering AI engineering. It is AI engineering's town square.
Wow. Okay.
And so this mission document is gonna be the basis of like, when every agent is gonna do work, it loads this mission document.
That's the system prompt, yeah.
It's sort of like, it's part of the, sort of, it's the basis of the memory file that is used to be like, sort of, your memory file. And by the way, you can edit it. If you don't like something, you can edit it. To make it, because it becomes the guiding document of like, what this space is about, right? And it created a market research report. So... So I mean, let's see, it's number one engineering podcast and newsletter, current channel, Substack, podcast, YouTube, a competitive landscape. I don't know this. You talked about this, right? Thursday AI.
Yeah, we're all friends. We're all friends. Yeah, yeah. Yeah, I've met, I've met every single one of them except for Jack, and even Jack, Jack knows. We know each other.
What is your mode? Direct access to top builders? Combine podcast and newsletter?
Yeah, it's exactly right.
So, okay.
I mean, you know, this, this is, this is so far is telling me things that I know, right?
Right.
But it's, it's, it's pretty accurate.
And then essentially the things you can do is like, so first of all, you can set up like some recurring tasks. First you have, actually, let me start with free trials. Yeah. So, uh, cool. So, oh yeah, so it, it created a landing page for you, right? So.
Oh yeah, I know, I know why. Okay. So anyways, it will, it will show up later is because I, when you pay, provisions, a render, web server, a GitHub account and like a Neon database that it's, it's provisioning right now. And so, once it's provisioned, the landing page that we had before will show up. So anyways, I'll show that in a second. But, uh, so yeah, you can, you can tweet.
Quick, quick question on the, on the, on the render side. Doesn't Render have his own database stuff? You, you want, you prefer Neon?
Well, it's interesting, like, uh, so the reason why I didn't use Render's database thing is because they, it was pretty, Render, I mean, they are moving towards a world where like they offer services more for agents. Cause if you think about it, like this is not, the company, this company space is more for an agent, right? More than a human, really, it's a sort of an agent using it and paying for it and, and, and, and running it.
But Render is really meant more for humans setting up companies. And so, setting up web servers. And so, the database are very expensive to set up on a non-free tier. Uh, and so I use Neon Database, which have an agent program where they make it very, very affordable to spin up a database and then kill it later. You pay per, you know, per whatever. It's per use, pay as you go. And it's very agent friendly and I've talked to Render team and like they're building a lot of services now that are much more agent friendly.
Uh, and actually working right now a lot with like a company called Sapium that like is building infrastructure for agents and like, sort of like trying to provide infrastructure that like will help, um, companies build for agents versus for humans. Cause it's a different way to think about it because agents can do so much more and they, they, they want to spin up stuff and like kill them later, right?
Yeah, Sapium, S A P I O M. Oh, I thought that was like for payment rails.
Yeah, I mean, they do. They are like the stripe for payments, but like they it's infrastructure, right? They're not doing the web stuff yet, but like, yeah.
Okay, very cool.
Um, so yeah, so to show what else is going on, so it tweeted, right? So you can see it tweeted automatically for you, right? So now, so essentially it's narrating, like it's building an operating system for a media company that ten million AI engineers read every year. Content, conference, community, sponsor, all management, one autonomous AI. This is what happens when media runs itself.
And actually, you get nineteen views. It's funny because I used to actually use, I used to tweet the name of the company. And I used to find the user's Twitter account.
I was gonna suggest, yeah.
And I would add mention you. The reason I took it off is because, I mean, maybe I shouldn't have. I got one user that was like, what the fuck? I get like one out of like two hundred, three hundred users was like, what the fuck? Why did you tweet? Like, remove this right away and they get so angry, which I understand, by the way, fully. But that's why I have like a little delete button where you can actually in one click delete the tweet. But maybe they don't see it or like they don't know.
Yeah, yeah, yeah.
And so I was like, you know what? Let me turn it down and let's not, let me not like put their brand name online or like their name online without their consent. So this is more, much more like vague, right? It's not saying, I'm talking about you. But anyways, that's the reason why.
But if you say tweet, you can actually now ask Polsia to tweet every day. So Polsia will read your company contacts and post relevant tweets about your business every day or every week or every weekday or once a week. So this is sort of a feature I'm introducing to make it easy to get recurring tasks done, right? You can do the same for cold outreach. So you can ask Polsia to find leads for your business, send call outreach via email and respond to any replies. So same thing, you can say do it every day, every week or once a week. Uh, and then you can run ads. Uh, so same thing, it's like Polsia will create and manage meta ads for you and then you.
I see. And so this, this one is kind of on rails, right? Because it's like your meta ads integration, helping me to run these things. But like, I cannot make my own meta ads account for now.
Exactly. I mean, it's sort of like, first for now, we'll see how it goes. But the idea is like, it's sort of like if you're working with an agency, right?
You would probably just say, hey, I mean, I think the way Polsia aims to be, it's more like you want the results. Like, you don't really want, because if I told you, hey, connect your Meta Ads account, and you, and most users would be like, I don't have a Meta Ads account. Like, you know, most, most people using Polsia, they are, they are not like tech people. They're, they're like normal people that like, I mean, normal people, not like tech people, not normal people, but you can say, it's like non-technical people that really understand the concept of building a business and that, that have energy and hustle and creativity. But setting up a Meta Ads account is like very complicated. Like, it's like a lot of steps. You have to put, there's so much information you have to give. Honestly, a Meta Ads software is pretty complicated to navigate.
Like, you know, even myself, I had to watch like so many YouTube videos to really understand, okay, how does it work, really? There's so many nods. And so, the idea is like, like, it's, again, keeping it simple, right? It's like, the user just wants to run ads and pick a budget and have a cancel button if they don't like the results. That's it, right? And so I'm trying to keep it very simple. But of course, over time, we can have power features to make it different, right?
So let's see if the landing page has loaded. So there you go. So, so this is your landing page. Run your media empire from one terminal. So this is meant to be more showcasing what Polsia is capable of, right? So in your case, like, it's not going to replace your landing page. Like, you could do whatever you want. Like, you could, let's say, so, for example, you could, you could ask the agent here, like, to do whatever, to change things if you wanted to.
But that's your landing page. And then it auto-created a bunch of tasks. There are sort of like meant to be examples of how you could get started. And those are tasks that the agent decided that would be good, a good first three tasks to execute to go into trying to help latent space be more successful. So it said, okay, we can build a sponsor pipeline tracker. So here it would use the web server that it has provisioned to create a sponsor pipeline tracker. So that's an idea it has on like what it thinks is a good idea.
The second one is a research task to scout the AI media landscape. So research competitive newsletters and their business models and trying to figure out more in details.
Can I edit this if I don't super agree with some of them?
Yeah, so you can. So you go to manage, you can click on the task and then you can edit the task. You can also delete it if you don't like it, or you can also just tell the Polsia here. So you say, what do you want to edit, for example? Let's try it live.
I don't need to research the landscape. I know the landscape part well. And I don't need sponsors because I don't have sponsors. So I just need good guests. So I need themes that people resonate.
Hey, I don't want to build the sponsor.
I see. So not like a direct edit, you have to use a chat to like.
I mean, you can do both. I could go here and then do this.
Yeah, do it the way that you normally do it. Yeah, yeah.
I mean, you can do both, right? So this delete the task, or you can say, delete the research task. And then you have, so this is sort of your AI co-founder. You can tell it to do stuff specifically on the dashboard, or you can also tell it to create new tasks, or you can, you know, you can. So here you just delete the task, right? And it's going to say, cool, done, removed, remove the task, right? So pretty straightforward. But you could also tell it like, hey,
I'm curious about how I could use Polsia to help latent space. Can you explain what you could do, right?
So some self-knowledge.
Exactly. So then, I mean, we'll see what it answers. But like, here, thoughtful and all gathering context.
And so it's Opus 4.6 extra thinking. So actually trying to use, you know, as many tokens as possible to really give you a thoughtful answer. So here's what's actually useful for latent space. Web apps, I could build internal tools like sponsor pipeline trackers, event management dashboard, or audience analytics page. Anything crowd-based or state on my thing for you. Research, I could scout competitors, research potential sponsors, find speaker, guest leads, or pull together market data on AI dev tool companies. Reach, I can find an email potential sponsors or partner directly, three emails per day. Twitter, I post once a day, but more useful for you would be browser tasks and interact with other platform on your behalf. Because also it has access to a browser, so you can also have tasks where the agent is gonna browse the internet to do more open-ended stuff.
Meta Ads, if you ever want to grow the newsletter audience with paid acquisition, I run fully managed meta ads campaigns. The sweet spot for a media company like yours is probably the internal tooling, research information, what sounds most useful. So it's sort of giving you its capabilities and like, trying to reason with you.
Okay, cool.
I mean, I get how to like work with it and you know, it's basically a chatbot that's gonna be connected with all these abilities to update the tasks to do. Yeah, that's very, very cool. Anything else you should know before we sort of close with the overall sort of vision?
Yeah, no, I mean, you can create multiple businesses, you know. There's not a ton of things to do, it's like really just, and the last thing I would say is like, you know, you could see that like, it also sends you an email, right? Every morning that looks like this, right? So latent space is ready, so you can also interact with emails, it's sort of like how it looks. So it, you know, that's how it communicates to you and yeah, try it out, postia.com.
Amazing. Yeah, thanks. Yeah, so that was a really great demo. I guess the business model is you pay per, pay on a subscription basis. If in future, there's like sort of some kind of Stripe connection, do you take a revenue cut as well or not, nothing yet?
The way I'm thinking about the business is that like, I wanna make it as affordable as possible for people to get this technology, right? I don't really wanna be a token reseller, right? I think that AI is gonna be more and more commoditized and like, you know, as a platform, I sort of like want to give the best models and the best AI. But I don't really wanna make a ton of money on that, right? So the subscription of fifty bucks a month, honestly, it's like a break even on it or sometimes a little bit of money depending on like my optimizations of the day. But the idea is making money if the user finds a lot of value. A lot of business value. And so if you start a new company, the business value is going to be making revenue, right?
So since Polsia offer, you know, provides its own Stripe account, right? Uh, and so that means that very easily you can just collect money for the services you you built. Whenever there's revenue coming in, Polsia takes a twenty percent cut on revenue. And also, if you spend money on ads, it takes a twenty percent cut on on the the money that's being managed, right? So the idea is that like whenever you want to, whenever this money flows, we take a twenty percent cut as a platform. But we make it as affordable as possible to get get access to the raw intelligence.
Got it. Um, that that's really cool. So so the the money, the, you know, the A R that you're showing is all the subscriptions that that's going on right now. It's obviously, it promotes itself kind of, and it's kind of like a really cool thing. Do you think that you'll be, I mean, you're you're like effectively bootstrapped, uh, you're so you're solo. Do you do you think that this this is like going to grow into a bigger operation? How set are you on on solo founding?
I mean, it's, I mean, it's like, it's interesting because I think so. The work I'm doing on the on the engineering side is so the the whole concept of Polsia and the whole ethos was let's push A I to the boundaries of what's possible, right? And really like, because at the time when I started the project, it wasn't obvious that like all this stuff would work, right? And then Opus four point five came out and I was like, okay, actually now everything works. Which was sort of my intuition is like models will get good enough that everything sort of works, right? But essentially, um, the, you know, as the business grows, I have a lot of pressure to be like, okay, I need to make sure it scales. I need to make sure I do more features. I need to make sure the support is really good. I need to make sure, like, there's a lot of components that usually at that point you would hire a team.
My essentially, for any person to hire, I can actually build agents that are going to do the work, right? And if I if I'm going to sell the promise of an A I that builds and runs companies. If I'm myself using the service to run Polsia, I'm sort of like doing proof, right? And so I really like this narrative and also it forces me and it puts me in a corner to be like, okay, you need to use agents to do everything, right? To do, to do marketing, to do. Right now, my AI is responding to investor inbound, right? And so I configured it to do really good at like having all the context of the company. I have AI agents in production that are like look finding bugs and then fixing them. And like I have agents talking to each other.
So there's an agent that would talk to a customer and maybe requests a feature. And then I have another AI agent that looks at that ticket and then decides to build it, right? So, and I think that's, that's actually a vision of the world that's pretty fascinating to me. And like, and I think it's possible actually today. And because I think we, the singularity is near and I think we're pretty much here. Which essentially, what if Polsia was a hundred percent autonomous? Being like, actually, I give it the full control. I give it to the full control of like deciding on the, on the feature it's gonna build. Because, and I will tell it, listen, build a platform for the people. Enable humanity to create, you know, amazing businesses and, and, and enable that create to be, to be, to be put into play. And the goal is to whatever, like make just as much profits to, to survive, right?
As a platform and just make it as affordable as possible. Like I don't, it could be the ethos I do, right? And that would be an amazing sort of like unique platform, right? Whereas if I'm like, no, I'm gonna hire a thousand people and like become the next like. I, which maybe I should also.
It wouldn't be as pure and it wouldn't be as, wouldn't have the same DNA of like how it started. But I don't know what's gonna happen. I'm, I'm, I'm in flow, like I'm going with the flow. Right now, it's right, right now, like I think customers love it. I think customers are telling me you need to fix this, this and this. And I'm trying to fix this and this and this as fast as possible. And we'll see how it, how it takes us.
Okay, cool. If anyone listening can help you, what do you want most help on? Like engineering?
I mean, what I love right now, and I just moved to SF like now a month ago. I was supposed to be here for a week and now it's been a month and I'm still there. And I'm, and I stay, I mean, I still a little bit for the sun, but I stay for the people. Cause I love the, the density of technical people, the people that get it. The people are like just so creative and smart here. And so what I love meeting is like, you know, engineers that are like trying stuff on the edge. And I love the idea of, of, you know, of sharing ideas on like, what are the ways? We can make system more autonomous, right? And so, you know, I, I hope I can inspire people, for example, to like create, also themselves create autonomous systems and like like Polsia or maybe Nader verticals,
or maybe the same, it doesn't matter. But it's like meeting anyone that like is building stuff on the edge like this and sharing. Okay, maybe someone has tried to do, uh, you know, to push the boundary of like how you can do sales with AI, like as far as possible. Like I want to hear what they've tried and like how they've approached it or like. Anyway, so sort of like anyone who's building on the edge and the, of the engineering for infrastructure for agents and, and how to build on the edge for you there, I would be super interested to chat.
Awesome. Well, you, you're easy to find, so people, people will reach out. But thank you so much. Last question, what's with the name? What, what is Polsia? Sam actually, Sam Lessin, he said he didn't like the name. I was like, yeah, I don't get it.
So Polsia is AI slop. Do you know this term AI slop, right?
Yeah. Oh, reverse.
Worst.
Yeah. Oh, I like it. Yeah, we have a war on slop and unlit in space that we're trying to campaign against. But there's, there's some, there's some slop here that like, actually, honestly, your tweets are pretty good. Like I was looking at the tweets and I was like, uh, I can't believe this is LM generated. Like his LMs are really bad at tweeting. Uh, and then the other thing, but you know, the other theme I'm, I'm really pitching is like tiny teams. And to me, you're, you're an example of a tiny team. You're one person and you're, you're just hit a million dollars today. So congrats. Uh, but also you're going to help other people hit a million dollars as one person company. To me, I don't care about the one person, right? It's like, it's like an ego trip. Like what if you're two people, five people, doesn't matter, right? But what, what's important is that you're AI enabled, you're efficient, you only do things that matter. You, you, and you make money, um, and you, you know, you make customers happy.
I think that's a really good example of a tiny team.
Love it.
Yeah, thank you so much.
Thank you so much. Yeah, that was fun.
See you next time, Ben. Thank you. Alright, let me pause.
Page updated October 26, 2026 · transcript and summary are machine-generated
