The New Startup Stack: One Founder + Agents | Henrik Werdelin (Audos) and Ben Broca (Polsia)

Transcript

I think Entropic recently is ripping. They are just dropping new products and new capabilities so fast. And that's what makes them win. Not because they're the model companies, because they are just in this flywheel. OpenAI needs to catch up on that, realizing, okay, they're not trying to do a social network, compete with Facebook. Let's actually go and just do the pragmatic things first. And a health product, and a shopping product, and a hardware device. They're still doing advertising. But the other thing is this relationship capital. Humans understand humans best. And I do think, can there be a founder customer fit where they understand the customer at a level where they can almost predict the problems that they're going to have? More or less. [SPEAKER_02] Why, hello, and welcome to a very special More or Less. Britt and Jess are in charge. Buckle up, folks. [SPEAKER_02] That's right. How it should be. How it is, actually. Let's not lie. [SPEAKER_01] We're under no illusions how this show operates. But this week, we actually have two very special guests. So I'm going to turn it over to Britt, who, by the way, her hair looks great, everybody. Check out the video. It's a special taping. But Britt, who do we have with us in the... I should say hot seats, because we're very gentle interviewers. [SPEAKER_01] You know, we're missing our husbands, Dave and Sam. Are we? Are we missing? Should we talk about that? [SPEAKER_03] Frankly, we're not missing our husbands. [SPEAKER_02] Having a great week. I get to go to bed on time. I get to watch whatever I want at the end of the night. It's pretty special. [SPEAKER_04] Amen, sister. It's pretty special. And we thought to ourselves, well, we do like having camaraderie on the show. I mean, as much as Jess and I could chat for an hour, I'm sure it'd be better to have some different points of view. So we found two other gentlemen to join us today, both of whom are near and dear to my heart, and both of whom are building some wildly exciting things in the world of solo entrepreneurship, AI running your company, the future of the internet, and so much more. And so today we have Henrik Werdelen, who's the founder of Autos. A-U-D-O-S dot com. I do have some people wondering if it's a car manufacturing company. And Henrik is someone that I've known for 20 years now. He is a serial entrepreneur himself, has built things like Bark, which he took public, and through a studio he ran. [SPEAKER_02] Wait, wasn't it BarkBox? [SPEAKER_02] Yeah. Was that a rebrand? It started as BarkBox. [SPEAKER_04] Okay. Well, I think people might know it as BarkBox. So we have to give it... [SPEAKER_04] It started as BarkBox in the Box era. [SPEAKER_04] Yeah. [SPEAKER_04] Oh, I see. And then you dropped the box. Got it. [SPEAKER_04] It was when Snapchat was not Snap. [SPEAKER_04] I got it. Okay. [SPEAKER_04] Yes. Anyways, has built and scaled so many companies in his life, and I think is at the epitome of what he does best, which is helping other people build companies. And then we have our friend Ben Broca, who is the founder and CEO of Pulsia.com, which has gotten a lot of press lately for helping people build the AI that runs your company. Ben was actually formerly at Cloud Kitchens, worked with Uber founder Travis Kalanick for a while, and then ran a company called Gift Shop, and now is doing this. So both of these gentlemen are doing incredible things. And again, thinking about the world in which we might be living very soon where there are maybe millions of donkey corns, as Henrik likes to call it, single-person companies making millions of dollars. [SPEAKER_04] Donkey corns? [SPEAKER_04] Yeah. We're rebranding it. [SPEAKER_04] They grind like mules, but they party like unicorns. Oh my God. Guys, I'm already confused. This episode is going to blow my mind. Okay. I'm paying... I'm taking notes. [SPEAKER_03] All right. All right. All right. All right. So what I'm so excited about is this space in general, right? And I guess, Henrik, we'll start with you. You spent 15 years running a venture studio, building all kinds of companies. What made you think that now you could help other people build and scale hundreds of thousands, if not millions of companies all by themselves? [SPEAKER_02] Well, as you mentioned, we spent a lot of time trying to figure out how do we help somebody go from nothing to something. And when OpenAI came around, we were asking, well, can we atomize all this process that we've built? And can we create agents that can help people do that? So we've spent quite a few years now trying to make that happen. And I think really the dream is to democratize entrepreneurship so that we can get a new class of everyday entrepreneurs out there that can take their relationship with a specific customer group and then use AI to build an AI startup for them. And so this is like an agent builds a company for you. [SPEAKER_02] Yes. So with our platform, people come in, they might not even have an idea. It helps come up with ideas, help them build the product, the landing page, help them connect with customers. And then, what we have, which we're announcing soon is a new business model where we then help finance their business against their royalty. Super interesting. Yeah, it's the... you know, we've talked about this a lot on the pod. I think Dave was actually one of the first to point this out, almost 18 months ago, which feels like 12 years ago at this point, which is, by coding is coming. Now people that never knew how to code or run a business are all of a sudden going to take these tools into their hands, be able to build things overnight. And we're going to go from 10 million people in the world who know how to build software to potentially hundreds of millions of people in the world who know how to build software and by the way, profit off of it. Right. And so that's what I think is so exciting about this. And, you know, Henrik, you're coming at this from a team at Autos of many people who are building all these tools for your burgeoning donkey corns. But Ben, you are a one man show right now. You have built Pulsia by yourself with a team of agents, no humans involved. And your philosophy is slightly different. And that I think you believe that the entrepreneurs of the future will also be solo entrepreneurs like you and don't need full teams. And ultimately the one man or one woman company is the future. Can you expand on that? [SPEAKER_04] That's what I think is so exciting about this. And Henrik, you're coming at this from a team at Autos of many people who are building all these tools for your burgeoning unicorns. But Ben, you are a one man show right now. You have built Pulsia by yourself with a team of agents, no humans involved. And your philosophy is slightly different. And I think you believe that the entrepreneurs of the future will also be solo entrepreneurs like you and don't need full teams. And ultimately the one man or one woman company is the future. Can you expand on that? [SPEAKER_04] In my opinion, AGI is here. Which means the AI models are super intelligent. They can reason on anything given the context and they are fluent at using any tool, which just means that if you push the argument, AI can run the economy now. The issue is that we need to educate everyone as soon as possible to become part of this new economy, which is going to be AI augmented. Every company will be 80% to 90% of it completely fully autonomous. And honestly, the fastest way to start a company is just one person. If you start with a couple of people, then there's already friction because they have to find each other. They have to agree on equity. They have to debate. A one person company is always going to be more efficient, but it used to be so frowned upon. It would be like the bootstrapped person. It was not ambitious and maybe they have to do compromises because they can't build as big of a platform as before. But now that you can take a human salary and literally offer that salary to an AI that costs one thousandths of the price, and runs 24/7, you can have one person with an AI team that can go really far. And that constraint I put on myself of staying one person is a testament to pushing the boundaries. And so feeling the pain—because it's painful. Meaning for a long time I was waking up every two hours to check that the platform was still alive. Now I found the solution. Your AI can shake your sleep. [SPEAKER_04] I like that. That's really smart. [SPEAKER_04] We solved it. But can I push you on AI can run the economy today on March 2026, because my Claude can't remember what I told it to do yesterday. Now my Claude code can, but my Claude co-work can't. So say more about what you're actually seeing that gives you the confidence to say AI can run the world? [SPEAKER_04] Well, to answer your question, I would say skill issue, right? No pun intended. Meaning if you think about what—so first of all, I say AGI is here. It's part of a provocative series of statements and marketing to wake people up because however you want to define AGI, it doesn't matter. Superhuman intelligence is here. We're two weeks away from superintelligence. You heard it here first. I'm living in the future. So AGI is here. Maybe it's here in two weeks and I'm exaggerating. It doesn't really matter. Jensen said it the other day, right? He said that AGI is here. But I was just going to say, maybe he's also being provocative or maybe he's trying to pump his stock. [SPEAKER_01] I can't imagine why Jensen Wong would want people to think AGI is here. I can't imagine how that would do to GPU sales, but maybe I'll ask the question differently. The technology you're building, what are the limitations at the moment? Because there can't be any limitations. You're still here, right? So what do you see as—and maybe that's a way of saying like in six months, what do you think you'll be able to do that you can't do now on the platform? [SPEAKER_01] I truly believe that there are no limitations today. If engineered the right way with the right guardrails, with the right infrastructure players that are agentic friendly, meaning they allow an agent to actually do anything in the economy, which today is not the case. They have great limits. They are not used to agency. They see agents as spam. So that's the limitation today. It's not really the capabilities of the model. It's more like the infrastructure of the internet that's not meant for agents. So there's a whole series of companies that build that. And we are building that at the platform level. And I'm partnering with a lot of companies at the infra level to build that. What are the limitations of AI? AI can run the operations, but in a world where humans control the wallet share, humans decide where their wallet, what goods and services are worthy of attention and spending. Then another human on the other side is actually the only one suited to understand in March 2026 what are the marketing techniques, the branding techniques, the price point, what resonates with people? For example, the marketing I'm doing—I didn't do a launch video on Twitter. I'm just screenshotting blurry screenshots of a graph that goes up. That was my intuition. It's provocative. It's a little annoying, a little douchey, but it wakes people up because it's like, what do you mean? And repeating one AI, one founder plus AI, zero employees every single time. That's my intuition on what I think will resonate with people right now. And so far I'm right. Now I'm working with a lot of super talented marketers. And I'm asking how can we reinvent marketing in a world where people may have different perspectives on how they react to marketing. So to make a long story short, AI is ready and it's ready to execute on the core work of the operations of running a company. AI is not good at really knowing what people want and really understanding the subtleties of what moves people. It may get there, but maybe we'll never get there because when it gets there, the market will be flooded with that perspective and humans will be like, that's AI. I'm not interested anymore. And then humans will be like, actually now the new thing is offline. It's person to person, but let's say you have a new company. It's emails. [SPEAKER_04] grant work of the operations of running a company. AI is not good at really knowing what people want and really understanding the subtleties of what moves people. It may get there, but maybe we'll never get there because when it gets there, the market will be flooded with that perspective and humans will be like, that's AI. I'm not interested anymore. And then humans were like, actually now the new thing is offline. Okay. Now it's like person to person thing, but let's say you're like, okay, what? I have a new company. It's emails, but it's human to human. And literally you send it. It's like someone writes a piece of paper and then they go to another person. It's an email, whatever. That's a great business. [SPEAKER_00] Post office. Yes. By the way, it sounds like a great idea. Yeah. Yeah. It's post office, but you rebranded as, you know, email, email 3.0, .com. We have a fun phrase that we use about offline ventures. It's actually gone more meme-ified globally, which is offline is the new luxury, right? Works for our brand. But I like what you said about brand, by the way. And I've noticed this in your marketing with Pulsia, it's always like one human, one, you know, a team of agents. Like you repeat the same thing over and over again in your marketing, all your tweets and you're right. Like in a world that is so crowded across social media, like part of the brand itself has that could be really relevant is to repeat something every time you talk. And I don't actually see that many brands doing that. Jess, you should take note of this for the information. Well, we're doing the first in the information. We're trying this. We're just saying first, first, first, first. Yeah, we're trying. There you go. By the way, I was recently with you when you bought note cards to write handwritten letters to people. Offline is the new luxury. How is this going? And where is my note card? You're on the list. Don't worry. We went to a stationary store and I bought Japanese stationary and I told Jess, you know, I'm going to sit down every morning a few times a week to actually send handwritten notes to people because in this world today, we've gone so agentic and AI everything. But I think people will really respect that. I just want mine written with a quill. There we go. Okay. I've never learned to write with a quill. I'm not from like the 1700s, but my terrible handwriting will get you something. Okay. I want to flip it back to Henrik too, because Henrik, hearing what Ben said, what parts of that do you agree with or disagree with? And how are you coming at this differently with your philosophy for Autos? [SPEAKER_00] Well, I think in many ways, I agree a lot with Ben. When we started, we thought we would need no humans. And so the first version of our platform basically went out and tried to find problems on the internet. Then it created landing pages. Then we had conversion funnels. And then it basically pinged us when we had our first paying customers. And it was crazy stuff. Like literally crazy. Like people wanted to buy fecal transplant from Olympic athletes and stuff like that. Which turns out to be illegal by the way. And so we got so obsessed about that. [SPEAKER_00] That is what the AI identified as a great business opportunity? [SPEAKER_00] I mean, that. [SPEAKER_00] Did the human ask for that? [SPEAKER_00] That was where the cack was low. [SPEAKER_00] I mean. [SPEAKER_00] Cack was low. [SPEAKER_00] Tampons for born women, now men that didn't want feminine products. I mean, it was some pretty advanced stuff that came out there. [SPEAKER_02] But what we got obsessed about was, to Ben's point, that you need a human in the loop. You need a relationship, a customer founder fit. And so we spent the first year and a half not being good as Ben of tweeting, but writing a book about what we call relationship capital. And so everything that we're focused on now is like, how do we get somebody in who has authenticity and authority in a specific affinity group? And then how do we help them build a product for that group? And what we can see is that we have people like Sarah who lost her dad, and then she built a business about grief counseling or Matt who took up golf and now is really good of understanding how to improve your golf swings. And so the businesses that are coming out now, to Ben's point, is people-centric. But then obviously, it used to be incredibly difficult to build these things. And now, specifically after Cloud Code, you can have machines help you do it. [SPEAKER_00] What makes this defensible? This is Sam's question, by the way. If Sam were here. [SPEAKER_00] Oh, I was just going to say, where's Sam? [SPEAKER_00] Do you want to be Sam right now, Jess? [SPEAKER_02] I think you could do him his best impression. [SPEAKER_00] I mean, it's also my question. [SPEAKER_02] Well, yeah, it's a good question. What makes this defensible if the agents are going to— [SPEAKER_00] He would also say, is this just all like narrative hype and you're winning the market based on this narrative? [SPEAKER_00] I know, but he also thinks it's not defensible. [SPEAKER_00] Yeah. [SPEAKER_00] He would think both of these things aren't— [SPEAKER_00] Defensible slash VC backed. [SPEAKER_02] I mean, I think what we will see is a lot of people trying to go after this space. You know, there's obviously Ben and us, and there's Paperclip just came up, and there's Base 34 with a super agent and Lovable. And I think a lot of people, and I think that is a good thing, will now try to help people build a business. And so, you know, hurrah for that. I think what will happen is that there will be somebody, there are people, that will have a really good relationship with a segment of an audience. But I think, and the reason why we talk about Donkey Kong is that what we might see is not that we're building tons of unicorns. I'm sure that we will have that too. [SPEAKER_00] But also that we'll see, and I hope that a million people becoming internet millionaires by building something on top of it. And I think for them, that is an incredibly good living. And we shouldn't forget that over half of the U.S. population worked for a small company that I think it's 89% of all new jobs last year came from a small company. [SPEAKER_02] a really good relationship with a segment of an audience. [SPEAKER_02] But I think, and the reason why we talk about Donkey Kong is that what we might see is not [SPEAKER_02] that we're building tons of unicorns. [SPEAKER_02] I'm sure that we will have that too. [SPEAKER_00] But also that we'll see, and I hope that a million people becoming internet millionaires [SPEAKER_00] by building something on top of it. [SPEAKER_00] And I think for them, that is an incredibly good living. [SPEAKER_00] And we shouldn't forget that over half of the U.S. population worked for a small company [SPEAKER_00] that I think it's 89% of all new jobs last year came from a small company. [SPEAKER_00] And so I think in many ways, what I hope is what we saw with Roblox and YouTube and Airbnb [SPEAKER_00] is that we can foster this whole new class of entrepreneurs that we might not have been [SPEAKER_00] talking too much about because we were very excited about unicorns. [SPEAKER_00] So I've heard the analogy, this is the next Shopify kind of category where [SPEAKER_00] small businesses can come and build a business on top of a platform. [SPEAKER_00] Ben, is that the analogy you would use? [SPEAKER_00] And how would you think about the overall defensibility of the space? [SPEAKER_00] Defensibility of like companies, like both our companies, what do you mean? [SPEAKER_00] AI, AI built companies. [SPEAKER_00] Yeah. [SPEAKER_00] I mean, Pulsia, I can only talk about Pulsia specifically, but we're not building [SPEAKER_00] a tool as building an economy. [SPEAKER_00] So if you think about entrepreneurs going into Pulsia, building a new business that maybe [SPEAKER_00] starts making a hundred bucks MR, 500 bucks MR, and then we're going to build an investor [SPEAKER_00] layer of investors that don't have to be your level of investors, but maybe middle class investors [SPEAKER_00] that maybe invest in the stock market. [SPEAKER_00] And then they go to like a platform like Pulsia and they can have an agent sort of invest [SPEAKER_00] or acquire companies on Pulsia. [SPEAKER_00] You start creating a much more liquidity in an economy where you have [SPEAKER_00] people, builders that have new ideas and create new ideas. [SPEAKER_00] And then you have investors that are either invest or acquire and grow and roll ups of [SPEAKER_00] that. [SPEAKER_00] And so if you believe that's possible, meaning you believe AI is able to enable millions of [SPEAKER_00] people to get new ideas out there and build them. [SPEAKER_00] And if you believe that there's still going to be investors that are going to be [SPEAKER_00] interested in investing in new ideas and funding them and acquiring them. [SPEAKER_00] And so essentially you believe that capitalism is going to be more and more vibrant. [SPEAKER_00] You start building a stock market, you start building an economy, you start building. [SPEAKER_00] And then with that economy, you can add more players. [SPEAKER_00] You can add loans. [SPEAKER_00] You can add a marketplace of customers that can discover companies and services. [SPEAKER_00] I think that's very defensible because there's a lot of network effects. [SPEAKER_00] Now you have to believe AI is here because if you don't think AI is here, this [SPEAKER_00] whole economy is just slop, right? [SPEAKER_00] It doesn't work. [SPEAKER_00] But if you believe it's here, I think [SPEAKER_00] there's going to be many Pulsias. [SPEAKER_00] I'm really struck by what I've seen happen in my business and the news information [SPEAKER_00] business with blogging and everything else. [SPEAKER_00] Right. [SPEAKER_00] So we saw the internet came along and all of a sudden it democratized the tools for being [SPEAKER_00] a content creator or journalist or any kind of content creator. [SPEAKER_00] And a lot of people pin their hopes on building an independent business with these tools, whether [SPEAKER_00] it was traditional journalists who left big publications for Substack or new writers who were emerging. [SPEAKER_00] And what's happened is there are a few hits, right? [SPEAKER_00] There are people who are making a lot of money on Substack, but it's really just a few hits. [SPEAKER_00] And then other people who are doing it, but they're not making money off it, right? [SPEAKER_00] They're doing it for expression or to feed into some other business. [SPEAKER_00] And everything, content got a little more commoditized. [SPEAKER_00] And yes, there was a pool that was able to make it big. [SPEAKER_00] And that's cool that they use that tool. [SPEAKER_00] But the vast mass majority of the people who came to these platforms for the promise [SPEAKER_00] of building new businesses just haven't. [SPEAKER_00] And largely because it was deflationary in some respects. [SPEAKER_00] So why won't that happen with AI? [SPEAKER_00] I think that platforms like Substack, all the social media platforms, [SPEAKER_00] they enabled more people with voices and creativity to express themselves and get an audience, or [SPEAKER_00] maybe you think about platforms like Etsy, people that build craft and sell. [SPEAKER_00] Now is everyone on Etsy a millionaire? [SPEAKER_00] No, there's some people making 10 bucks a month. [SPEAKER_00] Some people making a hundred, some making a thousand. [SPEAKER_02] It depends on how much energy they put, how much taste they have, [SPEAKER_00] how they think about distribution channels. [SPEAKER_00] I think about all this. [SPEAKER_00] I just think that if you think about how all the internet collectively enabled more people [SPEAKER_00] to be more independent in their choices of how they make money and how they [SPEAKER_00] choose to work. [SPEAKER_00] Because some people say Uber is so bad, the gig economy, it's so negative. But if you think about, talk about the people that actually have three jobs and are able to stack incomes and decide, I'm going to turn on Uber this weekend because I just need cash right now and then turn it off. Now, is it profitable for them? I don't know. I'm not in the Uber business. Maybe it's not. Maybe if you account for insurance, it's not great. But still people have a choice to decide across all these different ways to make money and make a living, how to do it. Right. And so I think AI is going to, AI and platforms like Otos or Pulsia or CodeCode or whatever, enables people to create more things. And so does that mean that all those things will be profitable or blockbuster successes? [SPEAKER_02] Probably not. [SPEAKER_02] But is it going to grow the economy? [SPEAKER_02] A hundred percent. [SPEAKER_02] Because there's going to be more economic actors, more things sold, [SPEAKER_02] more things bought, more diversity, more unique things. [SPEAKER_02] Because if you create a thousand things, you probably have a couple of, [SPEAKER_02] maybe 10 or 20 things that are great and amazing. [SPEAKER_04] And so I think AI is going to, AI and platforms like Otos or like Pulsia or like CodeCode or whatever, enables people to create more things. [SPEAKER_04] And so does that mean that all those things will be profitable or blockbuster successes? [SPEAKER_02] Probably not. [SPEAKER_02] But is it going to grow the economy? [SPEAKER_02] A hundred percent. [SPEAKER_02] Because there's going to be more economic actors, more things sold, more things bought, more diversity, more unique things. [SPEAKER_02] Because if you create a thousand things, you probably have a couple of, maybe 10 or 20 things that are great and amazing. [SPEAKER_02] If you can create a billion new things a year, maybe you have hundreds of thousands of things that are unique and amazing. [SPEAKER_02] And maybe when you buy a candle today, you have a choice between 10 brands that resonate with you. [SPEAKER_02] And maybe tomorrow will be a hundred. [SPEAKER_02] And maybe you'll be so happy to buy the candle that you would have bought at whatever, Deep Teach. [SPEAKER_04] There's a demand for a hundred candle makers. [SPEAKER_04] I'm just playing devil's advocate. [SPEAKER_02] I'm just arguing also. [SPEAKER_02] I think the reality is we don't know, right? [SPEAKER_02] Because this stuff is so new. [SPEAKER_02] Well, I know, but you don't know. [SPEAKER_02] Well, I don't know. [SPEAKER_02] We should check in in a couple of years, but I think, and Henrik, I'm going to go to you in a second, but it is really a question. [SPEAKER_04] There's no doubt today that I think the people on your platform have found that discontinuity. [SPEAKER_04] If you were like, they found these opportunities where there's some market need, they aren't being met and you can automate serving that need in some way. [SPEAKER_04] But if you really believe in HDI and us all having a gazillion agents, don't those opportunities collapse because everyone at all, I could just tell my agent to go build me this business, but go figure out every business that could possibly be built and go build it. [SPEAKER_01] And everyone's agents are telling them that. [SPEAKER_01] Right. [SPEAKER_01] So. [SPEAKER_01] Yeah. [SPEAKER_01] I mean, I think, and this might just be an overly philosophical question, but I mean, all of us was part of building the social web and I think I remember how excited we were when we took pictures of stuff and we could show it to each other. [SPEAKER_01] Little did we know that we helped destroy the headspace of a generation of young people. [SPEAKER_01] Right. [SPEAKER_01] And so if the best way to predict the future is to make it, then I also think in many ways we do have responsibility of trying to say, what can we actually do with this technology that could be pretty useful. [SPEAKER_01] And I think you're very right, Jess, that there is a chance that we'll all end up working for the foundational models. [SPEAKER_01] And then, you know, that is what it is. [SPEAKER_01] Right. [SPEAKER_01] I think for what I think Ben and what we're doing is that we're trying to say, Hey, if 60% of Americans say they'd like to build something and only 8, 9% go out and actually do it. [SPEAKER_01] And so there's this gap between the intent of wanting to go out and serve community and then the people that actually do it. [SPEAKER_01] Now, some of the people that maybe just say that they can't do it, they can't do it. [SPEAKER_01] But for a lot of it, it was the capability issues, right? [SPEAKER_01] Like they didn't know how to code. [SPEAKER_01] They didn't know how to do marketing. [SPEAKER_01] They didn't know all these different things. [SPEAKER_01] And I think over the last six months, they now have the ability to do so. [SPEAKER_01] And so there could also be this beautiful abundance universe where people who truly care about somebody else and understand their problems on a very granular level can now say, I can understand how to build something on top of AI and I can build a meaningful business doing that on your behalf. [SPEAKER_01] And for me, this idea of a million people making a million dollars is incredibly exciting to the alternative of all of us working for Sam Altman. [SPEAKER_04] Or Dario or Demis. [SPEAKER_04] By the way, guys, I have a, sorry, this is a good point. [SPEAKER_02] I have my advanced copy of the new Demis biography here by Sebastian Malaby called The Infinity Machine. [SPEAKER_04] What's the, what's your initial thought? [SPEAKER_04] Get all your post-its. [SPEAKER_02] I know. I really earmarked the information references. [SPEAKER_04] No, no, I, there are way more than that in here. [SPEAKER_04] This is an amazing book and I think it comes out soon and I actually have no idea if I have permission to even share this, but Demis gave Sebastian Malaby 40 hours of interviews or 30 hours of interviews. [SPEAKER_04] So it's this huge, very in-depth biography. [SPEAKER_04] And honestly, and I haven't read every part, but the takeaway is we're in the hands of the AGI. [SPEAKER_01] I mean, it's a little bit of, why wouldn't you think that? [SPEAKER_01] Feel the AGI, you know? [SPEAKER_01] And this is obviously Demis from, is the earliest believer in all of this. [SPEAKER_01] For the purposes of our conversation, it definitely paints the uncertain, it dials up the uncertainty about the future piece of it. [SPEAKER_01] That's where I was going to go. [SPEAKER_01] Jess is, you know, everyone that believes in, a lot of people that believe in this AGI future believe websites don't even matter anymore. [SPEAKER_01] The internet is dead. [SPEAKER_01] You're just going to have your agents surrounding you. [SPEAKER_01] You're going to have wearables all over us. [SPEAKER_01] Everything's going to be always live recording on, et cetera. [SPEAKER_01] And I guess for either of you or both of you, do you subscribe to that future or do you see it differently? [SPEAKER_01] And how does that influence where you're taking your products today and over the next few years? [SPEAKER_01] I mean, maybe just to answer a quick thing on the website. [SPEAKER_01] I mean, I believe that we will have a lot of agent to agent kind of stuff happening. [SPEAKER_01] And I think that will probably come sooner rather than later. [SPEAKER_01] The internet is dead. [SPEAKER_01] You're just going to have your agents surrounding you. [SPEAKER_01] You're going to have wearables all over us. [SPEAKER_01] Everything's going to be always live recording on, et cetera. [SPEAKER_01] And I guess for either of you or both of you, do you subscribe to that future or do you see it differently? [SPEAKER_01] And how does that influence where you're taking your products today and over the next few years? [SPEAKER_01] Maybe just to answer a quick thing on the website. [SPEAKER_01] I believe that we will have a lot of agent to agent stuff happening. [SPEAKER_01] And I think that will probably come sooner rather than later. [SPEAKER_04] And so we are all thinking about what a post-web is going to look like. [SPEAKER_04] But one thing that we've observed is when we started, this was before the first wave of AI code. [SPEAKER_04] And so early days, we were basically DMing through Instagram because that was the technology that was available. [SPEAKER_04] And then we were text-based. [SPEAKER_04] And what we learned pretty fast, what we had to invent was basically a just-in-time UI, because there are just things where you need some kind of visual interface. And so we had to build websites, even if they were just made in real time for that specific user. And so I think, yes, you're right, that websites as we know it today, where you sit down and drop down the menu and stuff like that. I don't imagine that to be around forever. But I do think having visual representation for a user is often needed when you are trying to solve some kind of problem. Ben, do you have any alternative thinking on that? I think to me, AI is doing 80% of the grunt work. So if you want to sell a candle, if you want to sell a t-shirt, it can just do the grunt work of signing up the websites, setting it up, giving you a few options. But end of the day, it's a human selling to another human. At least that's the framework that I think both of our companies are in. If it's agents selling to agents and humans become an enclave, you know, in the matrix, of course that world may exist, but let's just talk about the human to human economy. I don't know if people still would want websites in two years. Then it's going to be websites and humans will know if it's websites or if it's only offline. It's the old thing: 99% of the economy is offline because people hate the internet. And now you can still have an AI help you set up that offline thing because it can rent a space for you. If you think about it, okay, I need to build this offline experience. It's still a lot of work. You need to find a space. You have to rent it. [SPEAKER_02] You have to find stuff. [SPEAKER_02] You have to decorate it. [SPEAKER_02] You have to find someone to paint it. [SPEAKER_02] And you can have AI go research the internet, figure it out, get quotes, figure out your budget, find the right person, manage them. [SPEAKER_02] Because you should think about how, if I'm a contractor and I say, hey, I need business. [SPEAKER_02] What I do is I'm a painter and it's 50 bucks an hour to paint. [SPEAKER_02] You put it on freelancer.com and you have an AI that says, hey, okay. [SPEAKER_02] I'm not a human, I'm an AI, but I'm working on behalf of every customer. [SPEAKER_02] We'll pay you 50 bucks a month. [SPEAKER_02] You have to go here. [SPEAKER_02] Are you available? [SPEAKER_02] And they'll say, yeah, I'm available. [SPEAKER_02] And it's going to be 150. [SPEAKER_02] Okay, cool. [SPEAKER_02] I'll pay you on time. [SPEAKER_02] I'll be your best manager. [SPEAKER_02] If you do the job, I'll book you again for my next customer. If you don't do the job, I will blacklist you, but that's fine. You can go somewhere else. I will not book you again. I'll be your best manager. I'll explain everything and I'll give you a bonus if you do it in two hours instead of three. And I think that world will come and actually will be a net positive because people prefer talking to an AI than talking sometimes to another person when it comes to sharing weird ideas they have. And the same way people will maybe appreciate an AI that pays them on time and doesn't have bad breath and doesn't yell at them when something goes wrong. Do you have to disclose it? Is that the key? Is it disclosing like I'm an AI, I'm writing on behalf of Brit to see if you want to paint her wall for 50 bucks? I think whatever works, right? Whatever the human on the other side accepts or is fine with. I cannot predict. What I cannot predict is how humans will react to anything. What I can predict though is that humans will be the best at figuring out what other humans want. Well, Ben, you just had a flood of investors getting really excited about Polsia recently and emailing you and your AI responded and did all of your investor relations basically. Oh wow. I emailed Ben to connect him to someone and it responded to me and was like, hey, I'm Ben. I don't know. [SPEAKER_00] I think it said it's Ben, not your AI. I don't remember what it said. You shouldn't say Ben, but you know. Okay. It might've said it was unhinged. And it was like, hey, here are the terms of the round I'm raising. Here's the timeline. [SPEAKER_00] Like doing all the fielding of investor interest, which I thought was hilarious. [SPEAKER_00] But on the flip side. [SPEAKER_00] So I'm a user of both Polsia and Autos. [SPEAKER_00] Obviously I've built many companies on both of them. [SPEAKER_00] And I currently have one company that I'm building on actually on both at the same time. [SPEAKER_00] Because it's been fun to play with both. [SPEAKER_00] Which is making more money. [SPEAKER_00] It might've said it's a, it was unhinged. [SPEAKER_00] And it was like, Hey, here are the terms of the route I'm raising. [SPEAKER_00] Here's the timeline. [SPEAKER_00] Like doing all the fielding of investor interest, which I thought was hilarious. [SPEAKER_00] But on the flip side. [SPEAKER_00] So I'm a user of both Polsia and Autos. [SPEAKER_00] Obviously I've built many companies on both of them. [SPEAKER_00] And I currently have one company that I'm building on actually on both at the same time. [SPEAKER_00] Cause it's been fun to play with both. [SPEAKER_00] Which is making more money. Well, it's only two days in so far. Okay. Fair, fair. Okay. So I'm just going to tell everyone that's listening what I'm building. [SPEAKER_00] If you want to copy me, fine. So I think there's something really interesting about finding the best spiritual teachers. So a lot of women in particular are in my community. I've gotten really deep into finding the really cool psychics or astrologists or whatever. And it's an underground market. These people, you have to have a network. [SPEAKER_02] Find them on the internet. [SPEAKER_02] I have to ask you. [SPEAKER_02] Just wait. Okay. I'm going to ask you guys to do this in charge of subscription potentially, even on both sides. And so I just started best this weekend. Fast forward a day or two. Pulse is emailing out. He's identified all of the spiritual teachers starting to just email these people. [SPEAKER_01] And I'm wondering, huh, is this psychic medium going to think this? [SPEAKER_01] I'm totally insane for having my AI reach out to them or are the psychic mediums psyched? [SPEAKER_01] Cause they knew this was happening. [SPEAKER_01] This was going to happen in the first place. [SPEAKER_01] And it's like, they're down with it. [SPEAKER_01] But I do think this part of disclosure and this AI acting on your behalf to even just reach out to people. [SPEAKER_01] It's sent 28 emails yesterday. [SPEAKER_01] I'm okay, I'm kind of letting it run and see how it goes. [SPEAKER_01] But I don't know. Are normal people going to be okay with that? Or again, is there going to be some form of AI etiquette here? Bot etiquette is totally, I think you put it very well, Ben, where there's just, I mean, I called the village pub, fancy Michelin star restaurant in Woodside to figure out if I could bring my kid. Cause I don't know what the kid policy is. Right. And they pick up right away and it says something like this is an AI, maybe I even said AI because in Woodside you can say that. Right. But it was like, you're talking to the digital assistant of the Woodside pub. Right. And I hear that voice at first. I'm like, oh God, this is going to be a series of dead ends. But I thought, you know, we're in Silicon Valley. Maybe this one actually works. I said, can I bring my nine-year-old to dinner at the bar? [SPEAKER_01] And they said, absolutely. And we'll see you at 5 PM. [SPEAKER_01] And I thought it was such a great experience and I'm glad it was so efficient. So I think we're fine with bot communication. [SPEAKER_02] We're good with that. [SPEAKER_02] Okay. [SPEAKER_02] But do you think the psychics are cool with it? [SPEAKER_02] With the AI? I do not pretend to speak for the psychics. [SPEAKER_02] But I think you know what I really think? [SPEAKER_02] I think most of the psychics are bullshit. [SPEAKER_02] And so they'll say whatever. [SPEAKER_02] That's why you got to find the top 5%, Jess. [SPEAKER_02] I know. [SPEAKER_02] I know. [SPEAKER_02] Also, once again, you and Sam are building the same company, which is so funny because you guys are polar. [SPEAKER_02] Sam's building a psychic marketplace. [SPEAKER_02] He's building something that could be used to do this. [SPEAKER_02] So I cannot say more. [SPEAKER_02] He's fun. [SPEAKER_02] Yeah. [SPEAKER_02] My co-founder, Nicholas, he did the same thing. [SPEAKER_02] He built something called Badger that does all the outreach on his behalf. So a business where I can just give a call. [SPEAKER_04] One thing that I will say is that I do think, and this is obviously from BarkBox where we have people calling all the time, is I do think that in this age where there will be a lot of slob, I do think people increasingly, when they don't want a mechanical answer, they do want an emotional human to human kind of thing. And so we... This is out in a few days, right? Yeah. [SPEAKER_04] Yes. [SPEAKER_04] We're going to announce tomorrow that we're buying Nocap, which is the first AI investor, because we are starting now to try to understand much better. [SPEAKER_01] We have our first cohort that we've invested into that are making real money on the platform. [SPEAKER_01] And so we're trying to figure out how we become really good at doing this kind of investment at scale, if you want to invest in hundreds of thousands of companies. [SPEAKER_01] And one thing that we are trying to figure out is where do you put humans in the loop? [SPEAKER_01] And what kind of humans in the loop do you need to put there so that the founders feel that they have that? [SPEAKER_01] Because one thing that the solo founders on our platform really seem to need is a little bit of community. [SPEAKER_01] They do like their agents, but they're also quite lonely, a lot of them. [SPEAKER_01] So we're trying to figure out exactly what, where is that balance between being mechanical and efficient, but also having that human touch. There's definitely this, I'm also like, when is this coming out versus what I can say? It's good that I think that way, but there's this barbell, right? To the world. And as the AI side, the automated side, the agentic side goes through the roof, there's a really interesting question of like, what is society going to try to do to counterbalance that? [SPEAKER_04] Right. [SPEAKER_00] And we've talked about offline and community and, you know, Dave's favorite topics, but I mean, maybe psychics too, I guess. and efficient, but also having that human touch. There's definitely this. I'm also wondering, when is this coming out versus what I can say? It's good that I think that way, but there's this barbell, right? To the world. And as the AI side, the automated side, the agentic side goes through the roof, there's a really interesting question of what is society going to try to do to counterbalance that? [SPEAKER_04] Right. [SPEAKER_00] And we've talked about offline and community, and Dave's favorite topics, but maybe psychics too, I guess. [SPEAKER_00] But I just want to say, Ben, I'm really struck. And I think it's really cool how much agency you still give the human in all of this. You've said several times, humans know what humans want best, that demand thing. [SPEAKER_00] And I think that's really interesting because we've also learned from Amazon's recommendation engine that tech is really good at predicting what people want or from these social media engagement traps. [SPEAKER_00] Is that just an instinct you have, or is that proving true in other ways that you're seeing? [SPEAKER_00] Well, I think if you look at Amazon or Reels or whatever, essentially it's algorithms that are very good at showing other humans content or products to other humans. [SPEAKER_00] Right. [SPEAKER_00] So it's not necessarily Amazon or Facebook creating Amazon generated or Meta generated content and pushing to people. [SPEAKER_00] It's connecting humans. End of the day, right. [SPEAKER_00] So that to me goes back again to AI is really good at the 80%, the boring 80% of how do I connect to someone else? [SPEAKER_00] Okay. If I'm doing cold outreach, should I do cold outreach if it's a psychic or actually should I do an AI call or should I actually get a human to walk to them silently and whisper in the air? [SPEAKER_00] Yeah. Make that an option. I want to select that one. But that's the weird stuff that will emerge, right? [SPEAKER_00] Because the AI will learn the best practices at the time, which will change constantly, right? [SPEAKER_00] Because people will react differently to different types of outreach or marketing over time as society gets weirder and evolves. That reminded me of a big headline this week and we will get to some headlines at the end. So for those who are tuning into this podcast for the headlines, first of all, you're tuning in for the wrong reasons, but we will discuss it. But the shuttering of Sora as reported by the Information is pretty interesting in that regard, right? Which was a. Well, you guys both use Sora in your backend because I don't think Autos does, but Pulsia, you guys use Sora to generate video ads for these companies automatically. So how do you feel about this news? Yeah. I need to figure out another model. I mean, models. [SPEAKER_04] Sora 2 to me was the best. And I said that to Sam. It's the best model to me in terms of realism. They did such a good job. [SPEAKER_04] I don't really understand the way they're shutting down Sora app because it's just too expensive and didn't really work. [SPEAKER_04] But the Sora API, from the API perspective, I don't really see unless it's a copyright lawsuit situation that is getting too out of hand, which is also possible, but I don't think it's going to impact us too much because I will test the 10 other models that exist. [SPEAKER_04] And there's probably a few that are as good or better. [SPEAKER_04] It's all just electricity, Jess. [SPEAKER_04] He's just going to move over to the other meter and get his electricity. [SPEAKER_04] Yeah. Which again does not give a lot of confidence in the margins of the meter companies. [SPEAKER_02] I mean, there's a big. I mean, it's interesting. [SPEAKER_04] On that topic, I want the best outcome for my customers. [SPEAKER_04] So I want agents to have the best outcomes at the cheapest price possible because it's 50 bucks a month. [SPEAKER_04] If I can make it 25 bucks a month, I will 10X my user base. [SPEAKER_04] If I can make 10 bucks a month, I will get to emerging markets, right? [SPEAKER_04] So I want the cheapest price. [SPEAKER_04] And so how do I get the cheapest price? [SPEAKER_04] I need to go to every single provider of every single economic actor in the economy and put them into competition. [SPEAKER_04] But my AI is going to say, Hey, I was talking to Anthropic, right? [SPEAKER_04] And they were like, Hey dude, if you commit to a million dollars spent, I'll give you X percent discount. [SPEAKER_04] And if you commit to X amount, I will be X, Y amount. [SPEAKER_04] And I was like, yeah, I love it. Thank you so much. [SPEAKER_04] But it's going to be the opposite. [SPEAKER_04] I'm going to tell you my customer is willing to pay this much for this task. [SPEAKER_04] And this other AI is giving me that price. And it's 80% less, 20% less good than yours, but it's so much cheaper. [SPEAKER_04] Can you hit that price? [SPEAKER_04] I didn't really say that, but this is the reality where it goes, right? [SPEAKER_00] What about open source? Either of you using that? [SPEAKER_04] We're not yet, but as we are getting to scale and price and margins and lowering the price and having higher margins becomes so important. Yeah. [SPEAKER_04] We're exploring everything. And you look at the news of Cursor, who just dropped Composer 2, which obviously is their way to make margins in a world where they were probably losing money and now they're doing their own model and it wasn't covered. [SPEAKER_04] It was just like Kimi 2.5, RL done, probably their customers' data, anonymized or whatever. [SPEAKER_04] And it's like, oh, they were saying oh, it's as good as a person but faster. [SPEAKER_04] the price and having higher margins becomes so important. [SPEAKER_04] Yeah. [SPEAKER_04] We're exploring everything. And you look at the news of Cursor, who just dropped Composer 2, which obviously [SPEAKER_04] is their way to make margins in a world where they were probably losing money and now they're [SPEAKER_04] doing their own model and it wasn't covered. [SPEAKER_04] It was just Kimi 2.5, RL done, probably with their customers' data, [SPEAKER_04] anonymized or whatever. [SPEAKER_04] And it's like, oh, literally. [SPEAKER_04] And they were saying, oh, it's as good as a person, faster. [SPEAKER_04] And I think it's a Chinese model, right? [SPEAKER_04] That was RL. [SPEAKER_04] And you're like, okay, I guess that's not that difficult anymore to do custom models and use [SPEAKER_04] open source to get 90% of the value. [SPEAKER_04] And the really the SOTA model you can use maybe for chat or astrology. [SPEAKER_04] Right. [SPEAKER_04] But every grant work could be done by way cheaper models. [SPEAKER_04] For us right now, I'm sorry. [SPEAKER_04] Yeah, go for it. [SPEAKER_04] Sorry. [SPEAKER_04] I was just thinking for us, there's so many things. There's so many [SPEAKER_04] agents doing so many things, from "Hey, let me help you design your swag" to "let me design your ads" to "let's do outreach to journalists." Right now, I think the game [SPEAKER_02] for us is much more: what is the right model for all these different subtasks and then make sure [SPEAKER_02] that we can get as fast as possible to make the founders on platform as much money as possible. [SPEAKER_02] Because I think for all of us, that's really where the game is at. [SPEAKER_02] Can we actually show that people can come in and make a decent amount of money being [SPEAKER_02] an entrepreneur on platform? And then I think to Ben's point, the model prices will come down and [SPEAKER_02] suddenly one model becomes very expensive. You'll find a way to maybe use a [SPEAKER_02] subpar model and swap that out. But right now it's all about making sure the flow and all [SPEAKER_02] these thousand things that need to go right for being an entrepreneur goes right. [SPEAKER_02] So to wrap up this point and then we'll zoom out a bit, be respectful of everyone's time. [SPEAKER_02] Do you see the model makers having competitive long-term differentiation? And if so, if what [SPEAKER_02] you just said is true and the windows of commoditizing are collapsing, collapsing, collapsing, [SPEAKER_02] right? It used to be you had X years ahead. Is it that vision of the world or is it the vision of the [SPEAKER_02] world that there are enough specialization branches to go down that Anthropic [SPEAKER_02] can be a trillion dollar company, OpenAI can be a trillion dollar company, so on and so forth? [SPEAKER_02] My opinion is, in this world, everything is copyable much faster. AI enables everything to be copied much faster. And so the only mode is speed. What the [SPEAKER_02] model companies have is they're so fast because they know the new model before everyone else, [SPEAKER_02] three months before everyone else. And so they can use it or they can build the [SPEAKER_02] product that that new intelligence enables before everyone else. And so they have speed on [SPEAKER_04] their side. If you look at companies like Anthropic, I mean, OpenAI and Anthropic, I think [SPEAKER_04] Anthropic recently is ripping. They are just dropping new products and new capabilities [SPEAKER_02] so fast. And that's what makes them win. Not because they're the model companies, [SPEAKER_02] but because they're just in this flywheel of getting all the data from customers, [SPEAKER_02] RLing their models on pragmatic outcomes and then building the products that [SPEAKER_02] get them even more data and become more mainstream. And I think OpenAI needs to catch up on that. And [SPEAKER_02] that's what they're doing right now is realizing, okay, let's stop doing those. They start, [SPEAKER_02] that's not trying to do a social network, compete with Facebook. Let's actually go [SPEAKER_04] and just do the pragmatic things first: a health product and a shopping product and a chip and a hardware device. They're still doing advertising. I think the other mode though is one is speed. I very much agree with that. But the other thing is this relationship capital, like humans understand humans best. In the book, we talk about these three layers of relationship capital: depth, density, and durability. And that is experience we've learned from building Roe and Bark and all those different ones. I do think that will [SPEAKER_02] matter. Can there be a founder-customer fit where they understand the customer at a level where they can almost predict the problems that they're going to have before them and make sure there's a solution? If you can create that mode, the relationship capital mode, I think that together with [SPEAKER_01] speed will compound. What neither of you said is distribution, which I think is interesting, [SPEAKER_01] right? Like it makes you be worried for poor Google. I'm not really worried about Google, but [SPEAKER_01] what do you? We haven't talked about Google. Are you using any of their models? We're using Google for [SPEAKER_01] video, for example. Our background videos that get rendered as Google and we're using it in [SPEAKER_01] a bunch of other places. I mean, we really are using all the models right now and [SPEAKER_01] figuring out which one works best for each of the tasks. Okay. Can I ask a question? [SPEAKER_01] Given that you both are the king of agents, you're both like, and probably as [SPEAKER_01] deep as one could be, how are you using this in your normal life and maybe novel ways that aren't [SPEAKER_01] like the typical daily digest and the weather and the news? [SPEAKER_00] I'll give you a few things. I think increasingly, I really need to [SPEAKER_01] understand myself better in order to make sure that my agents can be an extension of me. So I've [SPEAKER_01] started to do things where I try to make sure that I don't have [SPEAKER_01] cognitive decline because I just go, yes, yes, yes. So a new thing for the agents that work, [SPEAKER_01] that I work with is to say, don't give me one solution, give me three different ones. And so it forced me to pick a different one. We're making the agents less good to hold on to our brain. I think we have to choose what do we still want to be good at? And that requires training. For example, if you just said yes three times, I'll go to bed at a decent time. It's so exciting right now that you want to work 24 seven. So for example, mine at 10:30, starts to go, "Hey, Anna, you should really start to go to bed now because if you don't get your seven hours, you're going to be toast tomorrow." so it forced me to pick a different one. We're making the agents less good to hold on to our brain. I think we have to choose what do we still want to be good at? And that requires training, right? And so that is for example, that, or if you just said yes three times, I'll go to bed at a decent time. It's so exciting right now that you want to work 24 seven. And so for example, mine at 10:30, start to go like, "Hey, Anna, you should really start to go to bed now because if you don't get your seven hours, you're going to be toast tomorrow." Dave, I brought up the idea called the brain gym last week on the pod. I think it's a thing. I think we need to talk about it. Oh my God. It's really wild though. Yeah. Okay. This is so exciting guys. I think it's amazing to hear from you. You have such an interesting point of view in all of this and just to see in real time. I feel like this is going to be like every three months, the puck is going to move somewhere. [SPEAKER_02] Every three months. I think we're deploying code every day, right? It is insane how fast these things are moving. [SPEAKER_02] Yeah. I mean, life as a founder right now, we don't need to get on this tangent, but you're deploying every day, you're moving faster than other, you're more excited about the rest of the world and everything happening in tech than ever before. So you're not sleeping. And VCs are just throwing money at anything and putting crazy valuations on everything. Is this not the strangest time to be a founder? Do you love it? Having built companies before, is this better or worse than the last go round? [SPEAKER_02] Oh yeah. Good question. [SPEAKER_02] I mean, I love the period right now, obviously it's a vulnerable time. The world is burning and all these things kind of seem to make you uneasy. But I think for an entrepreneur, the technologies that are available now for somebody who is able to originate and has something that they really want to build just never been there before. And so I find it to be incredibly exciting. [SPEAKER_02] This is the best founding era you've ever been in. [SPEAKER_02] I think so. [SPEAKER_00] In like 25, 30 years. [SPEAKER_00] I mean building BogBox and building BogAir was pretty fun too. And Ro and these other ones. And so I don't want to dismiss that, but it is pretty exciting. [SPEAKER_00] I mean, I think it's the most productive time to build, but it's also, everyone's more productive. I don't know if it's the most fun. You can find a joy in slowness and in grinding and you can be a Japanese artist and do blue jeans and just spend 10 hours doing a blue jean. And that's maybe as gratifying as having a jeans factory in China or something, producing a million an hour. So I don't know about being the most fun, but it's the most productive. And that's mind bending because even how fast I'm able to grow my business and how everything is compounding so fast. That's incredible. [SPEAKER_00] Yeah. It looks to me like the singularity is here. What we've called the singularity or AGI is here. And it's just a compounding of everything. And it's scary and exciting. That's the way I would see it. [SPEAKER_04] I think it's exhausting. I think that's what it is. I think it's also exhausting. I'm tired. [SPEAKER_04] And then maybe that does actually, and that's not a trivial point, right? To your point that if people aren't finding purpose, I mean, maybe they are to a point, but productivity and joy are two different things to your point, Ben. So yeah, I think where Henrik, your point about just having to know yourself better, I thought was really interesting, actually. Maybe that would be a good lesson for everyone to keep in mind throughout all of this. [SPEAKER_00] I think when OpenAI came out with the soul file and the persona file, I mean, it really clicked for me that we need to go through a period of time figuring out what do I want to do? Who do I want to serve? But also what kind of business do I want? Do I want one business and make a few hundred thousand dollars and just be fine with that? Or do I want to become a portfolio entrepreneur and have five of these running at the same time? What kind of parameters do I put around it? Do I go to bed at 10:30 and all those different things? And so I think this kind of period of maybe even using AI to become better at understanding who our true self is and how we want to show up in an agentic world, I think is going to be important. [SPEAKER_02] That is a wonderful note to end on. And I actually think we got through many of the big headlines of the week. We got through the strategy changes at OpenAI. We didn't mention SpaceX. Anyone got SpaceX shares out there excited about the forthcoming IPO? Sam does. [SPEAKER_02] Of course he does. Oh, okay. Great. Yes. [SPEAKER_04] We have Aura shares. So the Aura IPO is now on the radar. The Aura IPO, speaking of health. So yeah, we've got agent news and then we've also got some big tech market making things on the horizon. So we'll see what happens. But thank you guys for spending some time with us. It is really special for me because I don't get the benefit of talking to entrepreneurs I've invested in all day like Brit. And I will say, I love talking to both of you guys. And just one more shout out to autos.com and polisio.com and also Henrik's book, "Me, Myself and AI." [SPEAKER_00] Say that, say the name of the book again. I was actually going to— [SPEAKER_00] "Me, Myself and AI." Oh, sorry. "Me, My Customer and AI." "Me, Myself and I" is the movie. "Me, My Customer and AI." And so yeah, everyone should go buy it. Check out, start your own business today. Jess, you need to do it too. You know, I have a business actually. So not all of us, Brit, can start 27 businesses. You guys have your power user and power investor in one. [SPEAKER_00] to autos.com and polisio.com and also Henrik's book, Me, Myself and AI. [SPEAKER_00] Say the name of the book again. I was actually going to— [SPEAKER_00] Me, Myself and AI. Oh, sorry. Me, my customer and AI. Me, Myself and I is the movie. Me, my customer and AI. And so, yeah, everyone should go buy it. Check out, start your own business today. Jess, you need to do it too. You know, I have a business actually. So not all of us, Brit, can start 27 businesses. You guys have like your power user and power investor in one, but it is very cool. I mean, and maybe I'm thinking too small. Well, to our listeners, I hope you guys enjoyed hearing from Ben and Henrik and we will see you back here next week for another episode of More or Less. Bye. Bye guys. Bye. [SPEAKER_04] If you enjoyed this show, please leave us a virtual high five by rating it and reviewing it on Apple Podcasts, Spotify, YouTube, or wherever you get your podcasts. Find more information about each episode in the show notes and follow us on social media by searching for at more or less at Dave Morin at lesson at Jay lesson. And as for me, I'm at Brit. See you guys next time. But do you think like the psychics are cool with it? With the AI? I do not pretend to speak for the psychics. But I think I, you know what I really think? I think most of the psychics are bullshit. And so they'll say whatever. That's why you got to find the top 5%, Jess. I know. I know. Also, once again, you and Sam are kind of building the same company, which is so funny because you guys are like polar... Sam's building a psychic marketplace. He's building something that could be used to do this. So I cannot say more. He's fun. Yeah. My co-founder, Nicholas, he did the same thing. He built something called Badger that does all the outreach on his behalf. So a business where I can just give a call. One thing that though, I will say is that I do think, and this is obviously both from BarkBox where we have people calling all the time, is I do think that in this age where there will be a lot of slob, I do think people increasingly, when they don't want a mechanical answer, they do like an emotional human to human kind of thing. And so we... This is out in a few days, right? Yeah. Yes. We're going to announce tomorrow that we're buying Nocap, which is the first kind of AI investor, because we are starting now to try to understand much better. We have like our first cohort that we've invested into that are making real money on the platform. And so we're trying to figure out how we become really good of doing this kind of investment at scale, if you want to invest in like hundreds of thousands of companies. And one thing that we are trying to figure out is like, where do you put humans in the loop? And what kind of humans in the loop do you need to put there so that the founders kind of feel that they have that? Because one thing that the solo founders on our platform really seem to need is a little bit of community. Like they do like their agents, but they're also quite lonely, a lot of them. So we're trying to figure out exactly what, where is that balance between being mechanical and efficient, but also kind of having that human touch. There's definitely this, like, I'm also like, when is this coming out versus what I can say? It's good that I think that way, but there's this barbell, right? To like the world. And as, as we like the AI side, the automated side, the agentic side goes through the roof, there's really interesting question of like, what is society going to try to do to counterbalance that? Right. And, and we've talked about offline and community and, and, you know, Dave's favorite topics, but I mean, maybe psychics too, I guess. But I just want to say, Ben, I'm really struck. And I think it's really cool how much agency you still give the human in all of this for, like you've said several times, humans know what humans want best, that demand thing. And I think that's really interesting because I think, you know, we, we've also learned from like Amazon's recommendation engine that tech also is really good at predicting what people want or from these social media engagement traps. Is that just sort of like an instinct you have, or is that proving true in other ways that you're seeing? Well, I think that like, if you look at Amazon or Reels or whatever, essentially it's algorithms that are very good at showing other humans content or products to other humans. Right. So it's not necessarily like Amazon or, or, or Facebook creating Amazon generated or meta generated content and pushing to people. It's connecting humans end of the day. Right. So, and that's, that to me goes back again to AI is really good at the 80%, the boring 80% of like, how do I connect to someone else? Okay. If I'm doing cold outreach, should I do cold outreach if it's a psychic or actually should I do a AI call or should I actually get like a commission, a human to walk to them silently and whisper in the air? Yeah. Make that an option. I want to select that one. But like, that's the weird stuff that will emerge. Right. Because, uh, because the AI will learn the best practices at the time, which will change constantly. Right. Because people will react differently to different types of outreach or marketing over time as society gets weirder and weirder. And, and involves. I wonder that, that reminded me of a big headline of this week and we will, we will get to some headlines at the end. So for those who are tuning into this podcast for the headlines, first of all, you're tuning in for the wrong reasons, but we will discuss it. But the, the shuttering of Sora as reported by the information is pretty interesting in that regard. Right. Which was a. Well, and you guys, you both do both use Sora and your back, your backend for, cause, cause, cause I don't think Autos does, but Pulsia, you, you guys use Sora to generate video ads for these companies automatically. So how do you feel about this news? Uh, yeah. I need to figure out another model. I mean, models. Sora 2 to me was the best. And I said that to Sam, like, it's like, honestly, like it's the best model to me in terms of realism. Like they did such a good job. I don't really, I understand the way they're shutting down Sora app because it's just too expensive and didn't really work. But the Sora API, from the API perspective, I don't really see unless it's like a copyright lawsuit situation that like is getting too out of hand, which is also possible, but I don't think it's going to impact us too much because I will, I'll test like, uh, the 10 other models that exist. And there's, there's probably a few that like are as good or better. It's all just electricity, Jess. He's just going to move over to the other meter and get his electricity. Yeah. Which again does not give a lot of confidence in the margins of the meter companies. I mean, the, there's a big, I mean, it's interesting. I mean, on that topic, like I, what I want is the best outcome for my customers. So I want agents to have the best outcomes at the cheapest price possible because it's 50 bucks a month. If I can make it 25 bucks a month, I will 10 X my user base. If I can make 10 bucks a month, I will get to emerging markets. Right? So I want the cheapest price. And so how do I get the cheapest price? I need to go to every single provider of every single economic actors in the economy and put them into competition. But like, I'm not going to put them into competition. My AI is going to like say, Hey, you know, I was talking to Entropic, right? And they were like, Hey dude, like if you commit to a million dollars spent, I'll give you X percent discount. And if you commit to X amount, I will be X, Y amount. And I was like, well, yeah, I love it. Thank you so much. But it's going to be the opposite. I'm going to tell you my customer is willing to pay this much for this task. And so, and this other AI is giving me that price. And like, it's 80% less, 20% less good than yours, but like it's so much cheaper. And so can you hit that price? I didn't really say that, but this is the reality where it goes. Right. What about open source or either of you using that? I mean, we, we, we're not yet, but as we are getting to scale and price and margins and lowering the price and having higher margins becomes so important. Yeah. We're exploring everything. And, you know, you look at the news of cursor, who just dropped composer two, which obviously is their way to make margins in a world where like they were probably losing money and now they're doing their own model and it wasn't covered. It was just like Kimi 2.5, uh, RL done like probably like their, their customers, uh, data, I'm guessing or anonymized or whatever. And it's like, oh, like literally. And they were saying like, oh, it's as good as a person like faster. And it's like, uh, I think it's a Chinese model, right? That was RL. And you're like, okay, I guess that's not that difficult anymore to do custom models and use open source to get 90% of the value. And the, the, the, the, really the SOTA model you can use maybe for, for chat or astrology. Right. But every grant work could be done by like way cheaper models. For us right now, I'm sorry. Yeah, go for it. Sorry. I was just thinking for us, you know, like there's so many things that there's so many agents doing so many things like from, Hey, let me help you design your swag to let me design your, help me do your ads to let's do outreach to journalists that right now the, I think the game for us is much more, what is the right model for all these different subtasks and then make sure that we can get as fast as possible to make the founders on platform as much money as possible. Because I think for all of us, that's really where the game is at. Like, can we actually show that people can come in and make a decent amount of money being entrepreneur on platform. And then I think to Ben's point, the model prices will come down and, you know, suddenly one model becomes very expensive. You'll find a way to maybe use a subpar model and swap that out. But right now it's all about making sure the, the flow and all these thousand things that need to go right for being an entrepreneur goes right. So to wrap up this point, and then we'll zoom out a bit, be respectful of everyone's time. Do you see the model makers having competitive long-term differentiation and how, if so, if what you just said is true and also the windows of commoditizing are collapsing, collapsing, collapsing, right? It used to be you had X to your heads. Is it that vision of the world or is it the vision of the world that there are enough, there can be enough specialization branches to go down that Anthropic can be a trillion dollar company, OpenAI can be a trillion dollar company, so on and so forth? I mean, my, my opinion, it's like in this world, everything is copyable much faster, right? And AI enables everything to be copied much faster. And so the only mode is speed, right? Uh, and so what the model companies have is they're so fast because they know the new model before everyone else, three months before everyone else. Right. And so they can use it or they can, they can build the product that like that new intelligence enables before everyone else. And so they have speed on their side. And if you look at the companies like Anthropic, I mean, opening Anthropic, but like, I think Anthropic recently is, uh, is ripping. They are like just dropping new products and new capabilities companies so fast. And that's what makes them win. Not that because they're the model companies, just because they, they are just in this flywheel of like getting all the data from customers, RLing their models on pragmatic sort of like outcomes and then building the products that get them even more data and become more mainstream. And I think OpenAI needs to catch up on that. And that's what they're doing right now is realizing, okay, let's stop doing like those, uh, they start, that's not trying to do a social network, compete with Facebook. Let's actually go and just do the pragmatic things first. And a health product and a shopping product and a chip and a hardware device. They're still doing advertising. I think the other mode though, is one is speed. I very much agree with that. But the other thing is this relationship capital, like humans understand humans best. And I do think, you know, in the book, we talk about these three layers of relationship capital, depth, density, and durability. And that is experience we've learned from building Roe and Bark and all those different ones. And I do think that that will matter. Like, can there be a founder customer fit where they understand the customer at a level where they can almost predict the problems that they're going to have before them and make sure there's a solution. And if you can create that mode, the relationship capital mode, I think that together with speed will compound. What neither of you said is distribution, which I think is interesting, right? Like it makes you be worried for poor Google. I'm not really worried about Google, but what do you, we haven't talked about Google. Are you using any of their models? We're using Google for, uh, for a video, for example, our background videos that get rendered as, as Google and we're using it in a bunch of other places. And so, I mean, like we really are using all the models right now and figuring out which one works best for each of the tasks. Okay. Can I ask a question, given that you, you both are like, you know, the king of agents, you're both like, and probably as deep as one could be, how are you using this in your normal life and maybe novel ways that aren't, aren't like the typical daily digest and the weather and the news and all that? I'll give you a few, I mean, like a few things. Uh, one is I think increasingly, I really need to understand myself better in order to make sure that my agents can be an extension of me. And so I've started to do a lot of these things, for example, where I try to make sure that I don't have a cognitive decline because I just go, yes, yes, yes. And so a new thing for the agents that work, that I work with is to say, don't give me one solution, but give me three different ones. And so it forced me to pick a different one. We're making the agents less good to hold on to our brain. I think we have to choose what do we still want to be good at? And that requires training, right? And so that is for example, that, or if you just said yes, three times, like I'll go to bed at a decent time. It's so exciting right now that you kind of want to work 24 seven. And so like, for example, mine at 10 30, start to go like, Hey, Anna, you should really start to go to bed now because if you don't get your seven hours, you're going to be toast tomorrow. Dave, I brought up the idea of called the brain gym last week on the pod. I think it's a thing. I think we need to talk about it. Oh my God. It's really wild though. But yeah. Wow. Okay. This is so exciting guys. I think it, I mean, it's amazing to, to hear from you. You have such an interesting point of view in all of this and just to see in real time. I, I, so I, I feel like this is going to be like every three months, the, the puck is going to move somewhere. Every three months, you know, like I think we're deploying code every day, right? Like it is insane how fast these things are moving. Yeah. I mean, life as a founder right now, we don't need to get on this tangent, but like, not only are you, yeah, you're deploying every day, you're moving faster than other, you're more excited about the rest of the world and everything happening in tech than ever before. So you're not sleeping. And VCs are just like throwing money at anything and putting crazy valuations on everything. Like, it's like, this is this, is this not the strangest time to be a founder? Like, do you love it? Having, having built companies before, like how to, is this better or worse than the last go round? Oh yeah. Good question. I mean, like, I love the period, the period time right now, obviously like it's an, it's a vulnerable time, like the world is burning and all these things kind of like seems like, uh, makes you uneasy. But I think for an entrepreneur, the, the technologies that is available now for somebody who are able to originate and has something that they really want to build just never been there before, like now. And so I find it to be incredibly exciting. This is the best founding era you've ever been in. I think so. In like 25, 30 years. I mean like building BogBox and building BogAir was pretty fun too. And Ro and these other ones. And so I don't want to dismiss that, but like, it is pretty exciting. I mean, I think it's the most productive time to build, but it's also, everyone's more productive. I don't know if it's the most fun. Cause you know, you can find a joy in like, uh, in slowness and like, uh, you know, in grinding and like, you know, you can be a Japanese artist and like, uh, do blue jeans and just spend like 10 hours doing a blue jean. And that's maybe as gratifying as like having a jeans factory, uh, uh, in China or something. Right. And, and producing a million an hour. Uh, so I don't know about, about being the most fun, but it's the most productive. And that's, it's, it's, it's mind bending because, you know, even like how fast I'm able to grow my business and how, how all the, everything is compounding so fast. And, uh, that's incredible. Yeah. It, it, it's, it looks like, you know, what, what it feels like me is like the singularity is, you know, what we've called the singularity or AGI is here. And it's, this is just a compounding of everything. And it's just like, it's scary and exciting. That's the way I would see it. I would see it. I think it's exhausting. I think that's what it is. I think it's also exhausting. I'm tired. And then like, maybe that does actually like, and that's not, not a trivial point, right? To your point that like, if people aren't finding purpose, I mean, maybe they are to a point, but like, again, productivity and joy are two different things to your point, Ben. So yeah, I think where Henrik, your point about just having to know yourself better, I thought was really interesting, actually, like maybe that would be a good lesson for everyone to keep in mind throughout all of this. I think when the open clock came out with like the soul file and the persona file, I mean, like it really clicked for me that we need to go through a period of time, figuring out what do I want to do? Who do I want to serve? But also like, what kind of business do I want? Do I want one business and make a few hundred thousand dollars and just be fine with that? Do you want to then, or do I want to become a portfolio entrepreneur and have like five of these running at the same time? What kind of parameters of what I put around it? Like, do I go to bed at 1030 and all those different things? And so I think this kind of period of maybe even using AI to become better of understanding who our true self, who that is, and, and how we want to show up in an agentic world, I think is going to be important. Well, that is a wonderful note to sort of end on. And I actually think we got through many of the big headlines of the week. We got through the strategy changes at OpenAI. We didn't mention SpaceX. Anyone got SpaceX shares out there excited about the forthcoming IPO? Sam does. Of course he does. Oh, okay. Great. Yes. We have Aura shares. So the Aura IPO is now on the radar. The Aura IPO, speaking of health. So yeah, we've got agent news and then we've also got some big tech market making things on the horizon. So we'll see what happens. But thank you guys for spending some time with us. It is really, really special for me because I don't get the benefit of talking to entrepreneurs I've invested in all day like Brit. And I will say, yeah, I, I love talking to both of you guys. And just one more shout out to autos.com and polisio.com and also Henrik's book, Me, Myself and AI. Say that, say the name of the book again. I was actually going to, Me, Myself and AI. Oh, sorry. Me, my customer and AI. Me, Myself and I is the movie. Me, my customer and AI. And so, yeah, everyone should go buy it. Check out, start, start your own business today. Jess, you need to do it too. You know, I have a business actually. So, so not all of us, Brit, can start 27 businesses. You guys have like your power user and power investor in one, but it is very cool. I mean, and maybe I'm thinking too small and yeah. Well, to our listeners, I hope you guys enjoyed hearing from Ben and Henrik and we will see you back here next week for another episode of More or Less. Bye. Bye guys. Bye. If you enjoyed this show, please leave us a virtual high five by rating it and reviewing it on Apple Podcasts, Spotify, YouTube, or wherever you get your podcasts. Find more information about each episode in the show notes and follow us on social media by searching for at more or less at Dave Morin at lesson at Jay lesson. And as for me, I'm at Brit. See you guys next time.