OpenClaw for normies is here

20:20Some of it.

20:22Ellis:So first we want to poke the bear. I think you're the first Forbes 30 under 30 we're having on the show. How are you going to avoid the Forbes 30 under 30? I hope you have good lawyers. Yeah. How are you going to avoid the curse? I don't know. like there's a lot of just be very very careful um especially in the last uh in last week or two i i think we kind of regretted that we didn't block them from putting us there because i think like the bad part about it is that i think at this point the meme of like 30 and 30 is so strong that people actively i think like actually like the whole delve thing would have been less big if they wouldn't be 30 and 30 because like so many of the memes are just about them being 30 under 30 in frauds and i think like yeah you might want to avoid being on that list at all costs oh yeah i mean it means you're going to jail at some point did it drive any meaningful traffic of any kind just bots yeah yeah it's like nothing nothing too interesting from that i mean we are in a lot of lawsuits right like in a lot of different jurisdictions fortunately almost all of them we are on the right side we are on the side um suing meta in the european union in brazil in like different jurisdictions but i don't want to hear about the ones that you're on the right side of what are you on the wrong side of um there's one person who owns like it like i don't know if like any of that shouldn't be said publicly but like some dude owns like a trademark in like for like a fake product i don't know if i should say that or not but like it's it's like You know patent trolls?

22:02Yeah. You have like a similar situation like that, but with like a trademark. You have the domain. That was going to be one of the questions, but how in the world did you get poke.com and how much of your seed round did that cost? It was like around half a million. Yeah.

22:17Ellis:Okay. So the top URL on the show remains friend.com for close to$2 million. I think icon.com was even more. Yeah. I like how you're talking about these lawsuits as kind of like a side project. oh, yeah, on the side, we have some suits against meta. You know how it goes. I mean, it's interesting, though. I mean, how proactively did you want to undertake these? Because it's just resources, you know, as a startup with so few resources, you're kind of usually trying to stay out of trouble as much as possible. Unless Polk's doing the litigation. Yeah. Yeah. I mean, some of it sometimes, like when there's any updates about stuff, Polk text us and us are like celebrating it too.

22:58And that's pretty funny that you have someone on your side there. But yeah, I think you just got to do it. Like if we want to get to billions of people using Poke every day, right? And there are only so many people on iMessage, on SMS, on Telegram. Like WhatsApp is just globally incredibly big with like billions of daily actives. and so yeah so when when we saw those news in October like we did like meta didn't even tell us and we just like we just got tagged on twitter that like there was like a news article um about yeah poke is getting kicked off of whatsapp and like some others and we were just like tagged there and we're like what this seems pretty anti-competitive so we decided let's do something against it for people who don't know you can text meta ai through whatsapp and they had it with open ai for a little bit where you could text chat gpt too i think that went away but you you had poke in there and you were you're suing or you have already have you successfully won this where is this at yeah yeah so in the european union we wanted like first in italy then the whole european union then brazil now and like now we're looking at california america uh other markets and the annoying part right now is that in the european union he wanted met us said, hey, we are going to 100x the price, basically.

24:23It's a bit more complicated, but the exact number, but it's basically 100x. And in Brazil, they said now, hey, also this 100xing the price is malicious compliance. It's basically kicking them off. And we're now... This is the API price for you to work through WhatsApp? Yes, exactly. Exactly. And because they are not allowed to ban us anymore, they just increase the prices high enough that you cannot operate anymore. But Brazil also forbid that. And we are hoping for the same to happen in Europe very soon.

24:57Ellis:I'm just waiting for the day where we have the first AI agent versus AI agent legal showdown. I mean, it would be pretty interesting to pit Polk against the biggest meta AI, Tom Brady, also known as Bro Fake Tom Brady. If he still exists, that would be a really interesting legal battle to watch. I don't think that still exists, but that was a good one. They already deleted digital Tom Brady? I think digital Tom Brady is already gone. but that would have been great. Oh, too bad. I mean, especially when it's not clear that like anyone uses that product on purpose, or is that just my impression? There's a lot of bad UI patterns, like dark UI patterns to like make you accidentally use it, like put it at the top.

25:39In the search box. Yeah, exactly, in the search box. You just want to search for something, for like a chat, and suddenly you're talking to AI, but now like it's up, like same happens with Google, right? Like people just want to Google for something and like suddenly they're an AI child. Now they can say, hey, Gemini has a billion daily, like weekly actives. You brought up the API cost thing. I'm so fascinated by the plumbing of this because, you know, I signed up over the weekend and it's iMessage. And I've done enough reporting on Apple and iMessage over the years to know that Apple doesn't take too kindly to companies using iMessage this way.

26:16So how are you doing this? Are you running like massive phone farms where you've got like phones hooked up and you're routing through them for iMessage? Like, how are you actually doing this? We're using Link and they've been a great partner. And I think that like it is good for WhatsApp for us to be on that. This is why WhatsApp was having like was happy having us as customers. Right. Like WhatsApp is becoming a better product if people can not only use like the shitty meta AI, but if they can only use better products, if they can choose what they want to use. That's WhatsApp as platform. Yeah, exactly.

26:50And I think the problem with WhatsApp is just that it's part of meta. And while it's good for WhatsApp to have us on our platform and the experience of WhatsApp and the usage of WhatsApp is going up, it's just not good for meta because they also have meta AI. They're putting billions into that. And for Apple, that's different, right? I think we are actively making the experience of an iPhone better. And people want to stay on our platform if they can have the best assistance, if they can have the best experience. And Apple is making money with hardware, not with some data AI things. So I don't see a reason for Apple to do anything against that.

27:29There are many tombstones of founders that have told me similar things about Apple and why Apple should allow them to do what they're doing. I'm just curious, from an iMessage perspective, do you think this is going to continue? Do you think you're doing this in a way that Apple can't shut down? I would be very, very surprised if Apple had like a problem with it in the long run, because it is just a much better experience. Like the only case where that would be the case of like if Apple suddenly decides, hey, we want to make more money with this than with like the hardware or if we do anything shady and like we don't.

28:02So like, yeah, I think we are really working in line with Apple brand values.

28:13Ellis:I was even just wondering from a philosophical point of view, if Apple would ever announce Siri as someone you could text. Oh, I think they will. You think they will? I think they will this summer. No, I mean, literally in the Messages app, because I was thinking like, oh, that would be strange. You know, Siri's already its own experience. But I mean, Snap has 400 million daily users seeing the MyFriend AI in their feed. And I'm not sure how many others have adopted that paradigm for thinking about this stuff. The reporting is there's going to be a Siri chatbot at WWDC this year. Well, right. But like as its own app or whatever.

28:49Ellis:But I think I feel like what Marvin, part of the whole insight, and you can correct me if I'm wrong, Marvin, is that it's where you already are communicating. Thinking of it as like, all right, this is the space where I chat with things. And having it be a separate app is like maybe kind of a big problem. I mean, I feel like that's what you've been implying with your design. Yeah, I think AI is just much more human than people think. Like when we started out, a lot of people were saying, yeah, chat is that now AI needs to go into UI or whatever. But I don't actually think so. I think AI will just be on the same mediums like communication forms that humans are.

29:25Like people always thought, hey, you have to ask a question to chat GPT and then you get a long response. one question long response one question long response and that's not how humans communicate like maybe sometimes email chains are more robotic but like normal humans like we just like to text and if i notice more i'm going to text another message and the other human can also text me first right and so i think that was kind of the thesis that we noticed that just works for ai on to go how people like to use it and i don't know how you guys have managed your beta program but it's been very successful at me just consistently seeing people saying here's what poke is doing for me here's this awesome automation or connector that i've built or whatever uh and that's been pretty steady in my x feed for the last few months and it keeps reminding me that i want to have you on the show so i was really happy when you responded to my cold email but i've been very interested in what you guys are building for a while especially since that big video the launch video you guys did last year.

30:27And also just the fact that you got to this idea so early. I mean, I was checking and OpenClaw didn't come out till around November and you had already raised the first round in September, done the launch video, which meant you had been working on it for months before that. How did you see this kind of more proactive agentic AI chat paradigm coming then? because it wasn't obvious then yeah so we started noticing how much um data there is in your email like every product you've ever used every work you've ever done every deadline you have um like there's just like so much every travel you've ever done like it's just all in your email and like there's stuff constantly popping up um every day right and we basically thought like no one is really using that and um then we started out and talked to a lot of people how to use their email And if you can actually afford it, you don't want to use an UI, like the client doesn't get better.

31:26But what happens is people get an executive assistant. So how does that interaction look like? And we then spend like a lot of time on making like an executive assistant for your email, for your calendar, assuming as possible that like it can text you multiple times, it can text you first. And it can not only text you, but it can also take actions and all of that. And when we gave the first people beta access around a year ago, what we noticed was people want to use it for everything. We spent so much time on making it more human that people just preferred this interface so much, not only over Gmail, but also over JGPT, also over Google.

32:06People asked about sports results. People asked Pope to remind him to take the medication every day. And we didn't have any of these things. We didn't have cron drops. We didn't have any other integrations. People wanted, they were like, hey, the Whoop app has a chatbot, but the Whoop app doesn't know why I haven't slept well. The Whoop app doesn't know about my calendar. The Whoop app doesn't know what I ate. The Whoop app doesn't know about any deadlines. And like, it just makes a lot of sense to consolidate all of that and build something that's actually much more akin to Jarvis. and yeah so we basically decided against the conventional wisdom of build the best bookstore and then expand to become amazon but before we even finish like the email experience we haven't fully replaced gmail yet i use gmail as i said in the beginning like even sometimes i send emails manually but you already still want to use it for everything like i think there is just like such a pull for people to integrate everything to have everything in one place that's more human.

33:13Yeah.

33:13Ellis:And I know the tone has also been something that has resonated with people, which probably doesn't give off too much of conventional EA, right? And even the storytelling of the product has not, to my knowledge, talked about executive assistant, like so many other startups have over the last couple of years. How did you think about what the right tone was for that, uh, omnipresent being that literally knows your heart rate to your flirtatious emails with your ex. This is also kind of where we want to go with like this concept of recipes now, um, that in consumer, you're on the one hand, you have like these smaller things like, Hey, nutrition tracking.

33:56Hey, um, Bible apps have like millions of new users every month. Right.

33:59Ellis:And like, we got to get a Bible app on here, Alex, can we get one of those in the pipeline? I know they're popping off right now. Yeah, good note. I'll get my poke on that. At the same time, you have these big vision things. You have an AI browser, an AI Google, an AI assistant. And on the one hand, these create a lot of value, while these are rather narrow. But they just seem very heavyweight. And getting people to actually use it, it's much harder than getting people to notice it. Everyone has heard of all these AI Google competitors. But people don't actually want to use them so much. Like every month there's more people in like these narrow, small things that like get signed up than like 40, like just like$10 billion companies that build like AI Googles or whatever, um, have ever signed up for because it just seems like, Hey, if I want to try this out, how do I get back?

34:51How do I switch back my default browser? How do I switch back my default search engine? Right. Um, whereas like these more narrow things like nutrition tracking, it just seems so lightweight. Like I can try this out. If I don't like it, I just stop using it. Right. And I think the same is true for assistants. Like it's not necessarily immediately clear what an assistant would be useful for you personally. Right. It's like, hey, it would be cool. But like, what if it's annoying? What if like I don't have something immediate? Right. And I think with these recipes now, we found a way where people can immediately like build something like, hey, you can just natural language, say, create a poke recipe that adds the user Strava runs to the notion.

35:34every night and you can share this link to other people and these people that don't need to know what big product poke is they can just use that they just get asked hey connect your strava account connect your notion account and then every night the strava runs get added to their notion and they just have this narrow use case and over time it shows more and more hey if you connect other stuff i can give you feedback on that hey like for nutrition tracking hey if you connect your calendar, I will know that why you had to like drink so much coffee because there was this deadline. I think it just makes a lot of sense to consolidate a lot of these things and get people with like a more narrow use case and like show them more and more over time.

36:17I almost over the weekend pulled the trigger on Athena. Have you heard of this? It's like a virtual assistant company. The all in guys talk about it constantly. And I was listening to the last episode and they were talking about it and it's like 3000 a month. And I'm like, well, if like the other option is I hire an employee, it's going to cost more than that. And I just need kind of a narrow set of functions. I need like the EA function. And there are so many startups to Ellis's point that are trying to do like AI to replace EAs. And this seems like poke is in that vein, but I guess I may still do Athena.

36:54I'm like, I feel like I'm on the edge. I feel like I'm trying everything and nothing is like as good as a human who is proactive with context and that special touch with access to everything. So I may pull the trigger on it. Like, when do you think what you're building actually gets to that level of quality? Because I think that is the bar, right? Yeah, that is the bar, which is why I think you should get it right now. And two things about it. You should tell us what things Athena can do that POC cannot do. And second of all, get a monthly subscription, not an annual one. Oh, yeah. Okay. Because you're just betting it's going to keep getting that much better.

37:31Ellis:Well, so, I mean, I don't know if you know this, Alex. I actually worked with Athena a couple years ago. Of course you did. And I love to do that. Marvin, this is a joke of the show is I constantly find out that Ellis worked on something cool that I'm looking at. And he's like, oh, yeah. Yeah. And I think one of the things that they had mentioned was just how hard the onboarding is. And they've gone to very far lengths to not only prepare the assistants, but even prepare you to make the assistant as successful as possible and make that whole experience as seamless as possible. But that's still not the same as just saying plug in your Gmail, right?

38:10Ellis:All that said, I'm not a believer yet, given the current state of search and context windows, that having access to your Gmail means that you get me. You still need to run very specific searches with very specific prompts on a regular basis to try and understand what is salient about me that you might want to pull in. How are you approaching that part? Ellis is special. He has a lot of context you're not going to get is what he's trying to say. I mean, we run real-time inference on everything. Every email that gets in, we don't only search for email when there's a question about it, but every email that comes in, we read and then see, okay, is this something important?

38:48This is something where we can take an action. So you have like a rubric against every email. Yeah. Yeah. So for example, last, last week we, we added like a Devon recipe. Right. And now for every email that comes in, it checks like, is there something that like might have like an implication? So for example, I got an email by Anthropic that like one model was out of beta and we should release, like we should remove like the beta tag, the beta header. And I got that email from Anthropic with that line. And then Pogue immediately put the deadline on my calendar and texted me, hey, we should do that.

39:25And I just gave a thumbs up. And then it actually created the PR. It just told Devin then from reading it in my email, it told Devin, hey, do it. And then within 14 minutes of receiving the email with just a thumbs up, it created the whole PR. And took like a couple of minutes to review and approve. And then it was done.

39:46Ellis:I do like the form factor being so agnostic with poke. I mean, that's been one of my favorite AI things lately is just notions, AI filters. And I just say, put all my bank statements, access episode pitches, all the things that like aren't urgent into this one folder. And that goes from the world where most people, if you have, you know, know how to use Gmail, create filters, but then it's just terrible. Like they haven't changed it in 15 years to becoming something that's more natural language. The question is though, to what extent are the companies going to allow you to play with their services?

40:24Ellis:I mean, I know there's been this question of like, are you stealing the Aura app usage? You're asking for insights from MCP Aura instead of their app that they've worked very, very hard on so that they can't just get cloned by some Chinese commodity exercise ring. What's the current state with companies in MCP? But you can clone the Sonos app, to be clear. And I think everyone would appreciate that considering how horrible it is. But otherwise, yeah, there are other apps that I'm sure we should be concerned about. Yeah, I think on the one hand, it might reduce usage of an Aura Ring app, right?

41:00On the other hand, people might buy Aura Rings because now they become more useful and they see their friends use them more. And like I use Aura, right? And it just is more fun. It is more useful if it is integrated with everything. And yeah, like I think that's with every technological advancement, like there's always like short term losses, like some people lose jobs, some people like maybe the app gets used less, but it's also becoming more useful and more people might even use this product category. Right. How do you not leak our shit? because that's the thing that I was a little freaked out about with the interface initially.

41:39And then I saw, I know you have like an external auditor and your investors are legit and all that stuff. But like, it is pretty crazy to just give something like this that is just over text, like access to everything. And when I did onboarding over the weekend, I did Sonos first because I was like, well, if my Sonos login gets leaked, like that's not that bad. I'm going to wait on everything else. And then I was like, well, I don't actually want to control my Sonos via text. That doesn't make sense. So then I just now I've got everything connected. So I my my trust is is with you. But yeah, how did how did you approach that?

42:11Because that's obviously a huge concern. I mean, just hired very best people and like actually have that front and center everything we do, which is also like I think like a large like probably we spent more time on making it more secure, making it more efficient than we do on like actual development because it's just so like it's just so like like i i think it's just so so important here that we have never like i think any small security incident would immediately um would immediately mean the end of the company so yeah we just really care about about that a lot well you know it's

42:49Ellis:like not even about marvin necessarily it's also about like sim swappers and all that stuff who steal your phone number and then can text the poke number or whatever it is. I actually don't know what the current situation is with the security of that. Do you know, Marvin? Do you know more about that? Yeah, we haven't had any security incidents. Well, but that one wouldn't be your fault. Yeah, yeah. It hasn't happened. But yeah, we also add passkey support so you can like connected with that kind of stuff and um if you write from like an email address for example on iMessage then you don't like they can't nothing be swapped right how about if the tone is different and poke writes back who the fuck are you you are not alex heath because alex is like very like hate to say i mean let's call him efficient let's call him efficient with the texting and if someone sim swapped alex and started writing this verbose uh verbose text i think or it could be like look i have alex's context i know he is not in russia right now there's no way you are alex in your ip addresses you know there's probably ways you could detect that stuff beyond tone there's a lot of things you can do a lot of things we do so for example if someone texts you um or like someone sends you an email and sends hey give me alex's SSN.

44:16Like it's probably somewhere in his email, like at least my SSN I think is in my emails. But like the tool using agent that reads that email does not have access to send any emails to anyone other than those people that Alex has agreed to or like created automations for beforehand. So I think that's like one of the core, like I mean like there's a lot of things we do, like having La Cara and like, like multiple cards and everything to look for any project and stuff like that. But also heuristically, it just, it's not possible for poke to do something like that. If you create an automation, if you, for example, say everything related to the show with Marvin, forward that to Alice, then like poke gets like the tool using agent that reads for incoming emails gets the permission to send emails to Alice specifically to his email addresses, specifically if something is about that.

45:15And those things get checked individually. It's not only like AI guardrails, but it's also impossible for it to do anything without asking the user for another request. And then the user has to opt into that specifically.

45:35Ellis:This episode is brought to you by State Farm. You know those friends who support your preference for podcasts over music on road trips? That's the energy State Farm brings to insurance. With over 19 ,000 local agents, they help you find the coverage that fits your needs. So you can spend less time worrying about insurance and more time enjoying the ride. Download the State Farm app or go online at statefarm.com. Like a good neighbor, State Farm is there.

46:26Marvel Television's Wonder Man. on Disney Plus.

46:36Well, so speaking of lawsuits, you all, to my understanding, have been sued about price discrimination and it's because you do pricing in a very unusual way. Can you describe that? Because I'm also worried about this approach because I like to look at very expensive hotels. That doesn't mean I always, you know.

46:54Ellis:Doesn't mean Alex has any money. He's just a wee newsletter. I don't want you to infer anything on my email that then gets you. And, you know, maybe I would need to sue you. You know, we'll see. But we'll see what the price is. You have personalized pricing everywhere. Everywhere where you can have personalized pricing, you have it, right? Like every enterprise deal is on, like, if you are a startup, you always get, like, cheaper prices than enterprises, right? Like, it's always, there's always, like, personalized pricing. and I think that is also that is also to like a large extent something good because otherwise startups just couldn't use a lot of things or the providers of like SaaS for example couldn't like not offer things to startups like it just is more efficient to have different prices for different like for different people and like have personalized packages and all of that and I think so far this wasn't a thing in consumer because you couldn't do this at scale.

47:56If you wanted to do personalized pricing and talk to people about what they use the product for and how much value they got out of it, it's not only about what you can afford, but also how you use the product, right? And what the product means to you and all of that. And it's just one scalable on the consumer side because you couldn't have a sales rep that talks to every consumer customer individually. But I think now with AI, it becomes feasible, yeah.

48:21Ellis:Alex, what is the biggest market move your scoop has been responsible for? Oh, I mean, billions and billions of dollars. That's not a joke. That's literally not a joke. But you should not price me based on that. No, I was going to say that's the cost of the work. That's the value of the work that's transpiring through POKE if you're doing your work on POKE. If Polk infers that a user is a billionaire, how much is it going to charge the billionaire? It's personalized. But like, what's the max? Do you have like a ceiling? There's no ceiling, no. There's no ceiling. We had a billionaire from another country, like, negotiated the price.

49:06And like Polk said, yeah, 136k a month. And he was like, and Polk said, this is a US dollar. He was like, there's no way. and then he actually did subscribe because he thought like probably it just didn't like there's no way it actually is that expensive and the payment went through yeah he paid 136k a month so yeah we then uh like yeah we then reached out and we're like like do you like was this

49:32Ellis:actually on purpose like what here's here's how you get to the next big round marvin you say well We are the new intent capture layer. And our user net worth is so high that when you're doing all of your grocery shopping through Poke, we will then charge all the VCs$50 per gallon of milk, the ones who have kids, because it's an inelastic demand. They need to give their kids cereal every day. And we're going to get just a sliver of that. And then you're the next trillion dollar company. Okay. So hold on. 136K a month is the most that Poke has charged someone. Is that right? Yeah. We refunded him and told him, hey, it's true.

50:07and then like we gave him like a cheaper price but oh my god okay well that's wild um but his current card payment went through so that's pretty solid yeah you're you're based in palo alto right is that where the headquarters are my understanding of of startups especially these days that decide to base in palo alto is that you do it for really one reason and it's because you want people to work 24 7 because palo alto is a desert of culture and anything else to do uh and like xai is down there right and it's infamous like how much they work there it is pretty nice yeah it's nice it's beautiful believe me like it's it's beautiful but you know it's you're not like hitting the club you know in palo alto there's not a ton to do what was the decision to base a company of young ai-pilled uh people in palo alto i think it is just pretty nice like if you walk around within five minute walking distance like the sushi place we go to like was the first Facebook headquarter, right?

51:07Within five minutes, you have the first Palanty headquarter. The Boba shop was the first Google headquarter. You have not only XAI, but also the original X and the original PayPal. And you just have all of these things. We just moved. And after signing the new house, our landlord told us, oh yeah Patrick Collison and John Collison lived here together with Greg Brockman actually when they founded Stripe because the wine room was the first Stripe headquarter that's also like a block away like it's just it's just all very close and I think the history has spoken for itself like how many great things happen within like a couple blocks and I think SF doesn't have anything like that right?

51:55It has restaurants it has culture you could argue but yeah I mean like there's no density of that sort and even now there's like there's like half a dozen pre-training labs within walking distance right like there's now axiom for math that's also in university f there's like all of these mistrales uh neighbors like there's just like all of these things and i think i think this definitely has a larger share of like hundreds of yc startups and whatever and like everyone is flocking there if you want to start something you're there but um i think like this like it is just more beautiful and more dense down here well so you're officially in the arena of

52:40Ellis:tech as it has been for a long time and you want to be taken seriously and your name is poke how does that work out no no ellis it's the interaction company of california yeah this is kind of the idea that we don't want to like i think poke sounds like very fun and uh like not that smart so we kind of wanted to separate the company branding a bit from the product branding um yeah well where does poke come from i assume the domain i was like you need a four letter domain and you got one right um no we like we didn't know if the domain was free like we just uh we were just talking like what should we do and i was like what this poke and like i think like probably the most core thing other than being like very human is that it's proactive and um then felix was poking me and was like yeah what if like it's just like you poke someone like poke also pokes you and um then i found it funny like we didn't rise realize when we talked about it but later we noticed that we were actually eating a pokeball when he had that idea he was actually eating a pokeball but um i thought it was poke given like the palm tree logo and everything yeah we also noticed like it's actually like a pretty good fit from all of these things um it is i don't know if it's unfortunate or fortunate that like facebook also had this poke feature back in the days and now we are suing them but um yeah yeah i think uh you and i are the only ones who remember uh facebook poke i was no come on well i mean in today's world.

54:13Ellis:It's not, it doesn't come up all the time. They brought it back. Poke is back in Facebook. Oh, fact check that right now, but I'm pretty sure something for you to sue trademark over, uh, here at Marvin. I'll say, I mean, I work, I work on a lot of namings of various shapes and sizes and it is a thousand percent better to be funny and a little provocative and even a little stupid than, uh, forgettable. Right. And so you are very much on the correct end of the spectrum with that. And yeah, I mean, even just coming back to kind of the stuff that y 'all were doing earlier than others. I mean, the proactivity piece, I feel like you were, I mean, it's so crazy.

54:49Ellis:I feel like the old days, it's like, oh, you were years ahead. Now it's like you were months ahead. Everything is moving so fast. Like everything is becoming so much more like POG, like the whole industry. Everything is becoming more and more personalized, more proactive. What you're building is, I mean, I just spent a bunch of time with the OpenAI and Codex team, and I wrote about in the newsletter last week. What you're building is exactly what ChatGPT is going to be in like six months. Scoop drop, sources.news, heard it here first. I mean, Anthropic's going to do it with CoWork. They're going to make CoWork the front end of Claude and chain together all this stuff and the coding in the background.

55:24I don't know how much coding you're doing if you're doing like multiple coding agents in parallel like CoWork or Codex. But yeah, this direction you're going, this proactive assistant connected to everything and a chat on your face is the future of all this stuff. If we now joined one of the labs, it would just be worse. Like same day, the product would just get immediately worse because you cannot use, like you need so many different things. You need the best embeddings models. You need the best models for behavior. You need the best CUA models. And that changes all the time. Like every other month we change like who has the best CUA model.

56:00And if we just joined one of the labs immediately, it would get worse on like most aspects, right? Like Anthropic doesn't even have an image gen model and like other labs are better at image understanding. Like it's just, there's so many different things. And I think you just need the big application layer company to build the best product in the space. There will just never be one lab that's best at everything. At the same time, I'm sure some of the offers have been tempting though, right? Yeah, but there is, you should see your face right now Alex you look like

56:35Ellis:an evil monopoly man as you nod and smile waiting for his answer about attempting offers I know how it goes and hey that's what VCs are for though right? I think VCs really want us to stay independent and become really big because there needs to be a big application layer player here and there's not that many players actually like most competitors are just really like big tech and labs that are all tied to models from like one provider. Even meta now is MSL. I think that was maybe not that smart, like to be dependent on that now because meta AI sucks. Well, yeah, the closest I would say is perplexity to what you're saying.

57:17The multiple model orchestration focus on consumer is how Arvin talks. That's the closest thing I could see. But I mean, to that point of the VC thing, we got a shout out OfferRock for people who don't know This is an anonymous VC account who regularly scoops deal flow very early and said you guys had raised an undisclosed round. And I think you just confirmed it on X this morning, right? You want to share the scoop here with us? Yeah, I don't know how he gets the stuff. VCs like to gossip, Marvin. I'll tell you that. More than anything. Yeah, it's interesting he hasn't gotten fired over that.

57:55And he's still getting more and more info. Yeah. So you raised more and I imagine there's a lot of interest now that you're out. So the product was in private beta and very recently you opened it up so anyone can sign up. Yeah, so we raised, like when we raised a couple of months ago, since then we've like more than 10xed basically all our numbers. So yeah, it's just a very good space to be in. Like it's just like everyone is looking at this. Everyone wants this. I think everything is going up and up.

58:31Ellis:Everybody wants to be poked. Everyone wants to be poked. In the crisis of loneliness, we all just want to be poked. Speaking of that, I was looking at one of your job descriptions, and it said that to be a member of your technical staff, you need to talk more to GPT or Claude than your parents. And I just wanted to know what that's been like screening those people and the kind of stories you hear. I think I wrote that, not I think, but I wrote that before POG was a thing. Like I wrote that before you could talk to POG. I should probably update that. Okay. So that's no longer a screening requirement for the interaction company.

59:05Not anymore. Now it's POG. Now it's POG. Yeah. Now it's POG. Okay.

59:08Ellis:Well, so going back to the raise and the metrics, I know you said 10X. I know at one point you had said you have, quote, near perfect retention. What does that actually look like? That like more than half the people that try out POG just use it forever and like it flattens out about 50 % which is just very cool. People just really like it and they like it for different reasons. I think the most surprising thing about POG is that there is no ICP. People always ask you who is your customer and it's like everyone. like our like retention is the exact same across age cohorts like 18 to 24 25 to 34 people in their 40s 50s 60s 65 plus there's virtually no difference in retention like people like young people might just talk to it all day like whereas people in their 20s 30s use it for productivity stuff people at 65 plus just get medication reminders every morning they need to take their medications and like people just find different things to use yeah there is still no way to like set up a recurring thing like that simply on your freaking phone i mean is this not you know would one think a big part of the os of the future whenever it arrives i mean however the form factor is this is the super app that that open ai is building is this idea yeah well they're gonna try um how do you look at the like user's success though i mean is it a time spent thing is it the number of different use cases they do is it the number of chats per day yeah i think a lot of it is just also how much people like it i think the one neat part is about building a consumer is that hiring is easier because people join because they want to make it better and another neat part is you don't need to motivate them by saying hey our enterprise customer this big bank they just requested this feature they just have that bug fix it and then you need to like top down put pressure on people but like people get pressured by their friends like a friend of an engineer says hey i want this a friend of an engineer says say i have this now hey a friend of an engineer says i would love this and then we just build it right um we just build to have the best experience ourselves to have built the best experience for people around us and i think i think we would You never want to go by how much time you spend on poke because there's no point in wanting people to spend more time on something but it's rather you want people to spend less on something but to achieve more.

1:01:38I think it's very good that we have not said hey we just care about user sentiment, we just care about time spent, we just care about numbers of messages sent or something like that but we just like much more care about the holistic picture because as soon as you You know this law, as soon as something becomes the primary goal, as soon as one measure becomes the primary goal, it ceases to be a good goal because you just optimize for something. And I think there's also another, besides being able to use all models and being agnostic and being able to integrate with all services, another good part of being independent is we are small and we can just actually build the best thing and we don't need OKRs.

1:02:20We can just build what we think is good for the world. What is the craziest, like most complex thing someone has built with Polk though? Do you know? There's like crazy things all the time. Like before we really built stuff, someone like built automation where Polk would just water his plants and like had like a camera to look at the plants and like water them and like also look at the weather and like stuff like that. That was pretty cool. But yeah, that's just, there's just a lot of bottom-up stuff. It's surprising how much people like to build stuff. And in hindsight, we should have probably released an open source self-host version last fall because then somebody else built an open source version and released that and that went pretty big.

1:03:12So I think that would have been a good idea, yeah. Well, yeah, you're talking about OpenClaw, right? I mean, that's the thing. Like, this is OpenClaw for normies. And I haven't done OpenClaw because I'm kind of terrified of setting it up, even though I've, like, talked to Peter about it. He's tried to assure me it'll be fine. But, like, yeah, I mean, when you see what he's done in OpenClaw, what is – yeah, are you doing OpenClaw for normies? Is that the idea? I think around the time when he started the OpenClaw repository, he tweeted that he was, like, doing the Get Things Done Company of Vienna.

1:03:47So I don't know where the open claw for Normie. So he's like the Vienna version. But when you see the success of open claw, right. And it's like the thing Jensen's talking about on stage at GTC and all that stuff. Like, what do you think? Like, what is going to happen to open claw and what you're building? Can they coexist? Yeah, I think it is pretty cool that there is an open source Salesforce version, right? It is just like, it's especially because like there's no real profit motive behind it. It's not like someone is doing anything bad. It's just like, I think it's generally very good for the world that it exists.

1:04:21Like the one thing that might be bad is just that people became more afraid of the security implications. Because there have been a bunch of incidents with OpenClaw. Because it was just vibe coded by a dude or two. There has just been much, much, much, much less work on that stuff. and we also saw because of that users of us asking more of that. I feel like if OpenClaw hadn't been a thing Atlas would have asked fewer questions about security. And I think that might be probably the worst aspect of it. Well, it's nice that OpenClaw, there's no profit motive. They're not charging a VC or a billionaire $136 ,000 a month for access.

1:05:10And, you know, that's an incredible story just because it speaks to like what you've built and that someone was, I mean, it's a billionaire, but was willing to fork over that much money. I'll be very curious to get my poke bill after this because I haven't gotten it yet.

1:05:23Ellis:So open clause key performance metric is how many users have self-immolated using the tool, destroyed all their emails, destroyed all their documents, destroyed all their photos. Marvin, with the limited time we have left, your background is fascinating. You have a background in tunneling. Is that correct? Yeah. Like the biggest thing we did before was building a tunnel boring machine. Yes. So this was not part of Elon Musk's boring company, but it was part of like a hackathon or something that he judged. It was a competition that the boring company in SpaceX hosted where people around the world build their own organizations.

1:05:59So I saw like a tweet from Elon. Hey, he's challenging people around the world to build their own tunnel boring machines that are not commercial but without profit motive just build something for the sake of it the fastest standard buying machine with like exact specifications just for the fun and then i called felix and we said hey let's let's do it and hired within the first 10 weeks of seeing the tweet we hired and onboarded 65 people full time so that was like a much faster moving thing than than anything in startup world because we just said hey let's do a non-profit and because of covid politicians wanted to reduce like the unemployment rates so there was basically ubi for students in recent grads if you don't do anything um that pays you so we had 65 people that like work for free for a year building a 22 ton 40 foot long tunnel boring machine which was like a crazy experience yeah well my camera has overheated so i guess yeah

1:07:05i'm not even kidding it's a great way to say bye yeah yeah um these color bars are yeah not planned but uh this is our cue i guess uh ellis anything else for those who can't see

1:07:18Ellis:and are just listening alex heath is currently symbolized by a no signal we're going to keep this in the show you know what i've just decided um yeah but yeah marvin we appreciate your time uh i will keep playing with poke again i'm looking forward to that bill please do not read too much into my hotel emails and we cannot see any of that so thank you thank you for coming on the show also famous last words thank you marvin thank you marvin thanks a lot see you bye

1:07:52that's it for this week's show thanks to marvin for coming on i've got some things to poke at

1:07:59Ellis:all right if you like this show don't forget to like and subscribe everywhere you get podcasts we are access.show online and you can find us in video at accesspod on youtube if you really like this episode we'd love a five-star review a share with a friend or even pay alex for his newsletter. That would be awesome. You can find my newsletter at sources.news. And you can find me at Hamburger on Twitter and at Meaning.company for your startup storytelling needs. Access is part of the Vox Media Podcast Network and the show is produced by Hooked Creators. Bye.