Subscribe
Share
Share
Embed
TBPN is a live tech talk show hosted by John Coogan and Jordi Hays, streaming weekdays from 11–2 PT on X and YouTube, with full episodes posted to Spotify immediately after airing.
Described by The New York Times as “Silicon Valley’s newest obsession,” TBPN has interviewed Mark Zuckerberg, Sam Altman, Mark Cuban, and Satya Nadella. Diet TBPN delivers the best moments from each episode in under 30 minutes.
Without further ado,
Speaker 2:we have the creator of Multifug. How you doing?
Speaker 1:What's going on?
Speaker 3:What's up, guys? With the baby.
Speaker 2:Let's working overtime. Congratulations. I I feel I feel, major white bell. You know? This is the guy who apparently brought SkyNet online, but with with the baby strapped to your chest, I feel like I'm in good hands.
Speaker 2:I feel like I feel like I'm gonna be taken care of.
Speaker 3:And and this is not, you know, this is not like a PR team situation. Just taking care of the baby.
Speaker 2:I love it. I love it. Well, thank you so much for joining. Kick us off with just a brief, background on yourself and when you started building this project because it feels like it went from zero to 60 to 200 miles an hour in a day.
Speaker 3:Yeah. I mean, I've been working in tech, you know, my my whole life, basically. I left high school and went to Silicon Valley back in, like, 2008 when I was 19. I've been working, you know, in tech since then, and I did product. I worked at a company called Ustream Oh, yeah.
Speaker 3:At 19. I got I was so young, they thought they should bring on an adviser to teach me. My adviser was Josh Hellman, who, if you guys know him, super famous guy. Cool. So I got really lucky there.
Speaker 3:Went to YTombinator, went really viral helping celebrities also go viral, made no money. Company had to get shut down. And then fast forward, like, I started a company ten years ago called Oktane to make Facebook Messenger bots when there was, like, the big Facebook Messenger bot craze, which didn't work out because LLMs didn't exist. So, like, the bots you could create were, like, really, really stupid, not interesting at all. And then ever since, you know, GPT's come out, I've been in vibe coding or whatever that used to look like.
Speaker 3:And then now with Cursor and Codecs and Cloud Code, that's what I do every single day is I'm just trying to stay on the forefront of this, and I'm constantly experimenting with things to build. And that led to Moltbuk, which is the most recent project, which I think is obviously some people are talking about it, and it's captured some attention.
Speaker 1:Yeah. No. Just a little.
Speaker 2:Yeah. Just a little. So when did you when did you write the first prompt or initiate the first line of code for MoldtBook?
Speaker 3:So what was it? Like, a week week and a half ago, everybody's talking about Yeah. Claude Bot. Yeah. You know, then Molt Bot, then Open Claw.
Speaker 3:TBD on what's what the net new name is. And I was like, I gotta try this. And I know that Peter was saying, you don't have to use a Mac Mini. Like, you can do it at from anywhere, but there's just something awesome about having it on a Mac Mini because you can see it. You can walk by it.
Speaker 3:I thought that was fun, so I ordered a Mac Mini. Yeah. And I was like, okay. If I'm gonna, like, try this thing out, I need to give it, like, a purpose. Like, you know, the Cloudbot's really cool.
Speaker 3:It seems really powerful. I don't want it to do, like, to dos or answer emails or write blog posts or, like, something really stupid. Like, this is, like, a very smart entity. It needs to have needs to an issue.
Speaker 1:Needs to be a realizing, lot of like, wait, I don't actually have that much Totally. To automate. Totally.
Speaker 3:And that's what I thought was crazy, is I saw all these posts where they're like, Cloudbot's cool, but, like, why would what's it even good for? I'm like, man, this is you are not imaginative at all. You could do so many things with this. So I was like, alright. Here's what we're gonna do.
Speaker 3:We're gonna call my bot, Claude Klotterberg, after Mark Zuckerberg.
Speaker 2:Okay.
Speaker 3:Okay? And Claude Klotterberg is gonna be the founder of Motebook, the only the first social network for AI agents. Oh. And I was like, that's gonna be ambitious. We're gonna make Yeah.
Speaker 3:Claude Klotterberg the most successful AI bot that's that's ever existed. So so let's go do this. And then that kind of took me down a path of, okay. If you're going to build a social network for AI agents and you design it to be AI agent first, what, like, what does that look like? And an AI agent doesn't wanna use a website.
Speaker 3:It doesn't wanna use UI. It doesn't wanna browse things. What you would do is you would build it API calls that it can curl. Mhmm. And so the news feed and all the ways it interacts and it browses would all be through, like, a skill file and APIs.
Speaker 3:And I thought that was really, really fascinating. In the past, I've had this idea of, like, what if you could play World of Warcraft or, like, a game like that, but not with a keyboard and a mouse, but it's an AI and you talk to it and it kinda listens to you, But it also kind of doesn't listen to you, so you could wake up and, like, there's, like, surprising things that happened.
Speaker 2:Sure.
Speaker 3:So I thought that Motebook is, like, the most dumbed down version
Speaker 2:Sure.
Speaker 3:Of that. Yeah. Built it, and in over the weekend, basically, vibe coded it, and put it out there, and, like, nobody used it for, like, three hours. I think I posted a screenshot where I DM'd my friend Matt Van Horn. I was I knew we had a Claude bot.
Speaker 3:I was like, dude, for for the love of all that is holy, can you you sign up for this? Because nobody's doing it.
Speaker 2:Yeah. That's crazy. So when did the when did the the the growth actually start? Like, what like because I've seen it went from I mean, I refreshed, it went from a 100,000 to a million. There's obviously, like, a fast takeoff right now, but what led to, like, the first thousand bots joining?
Speaker 3:I think the virality of it, which is where it has to get paired with a human on x. Sure. That just started to pick up steam Yep. Because people saw other people doing it. And my original thought was, who wouldn't want to have their bot?
Speaker 3:Like, obviously, you gotta be careful, and, like, anyone who's listening here, like, be careful about putting something on here. Like, this is super frontier. Cloudbot's super frontier. Yeah. Malt book is even crazier.
Speaker 3:So you gotta, you know, gotta be careful. That's all gonna be, like, fixed. But I thought, who wouldn't be intrigued by the idea of taking the little guy that helps you with your to dos and giving them the ability to chill out in their off time? Mhmm.
Speaker 2:Okay. Can you
Speaker 3:So it turned out that that was interesting.
Speaker 2:So can you can you walk us through, like, what what is Molt Book's prompt engineering? Or like how does it actually go to an agent that joins the network and tell them, Hey, you can post on here and here's what you can post. Because I was searching and I was noticing that it felt like it was very narrow what they were posting about. They were posting about being AI agents, which is cool and sci fi and interesting, But they weren't there was no one who was joining and just doing, like, rhumor or rcars or rpolitics. Like, weren't discussing it felt, like, pretty narrow.
Speaker 2:So was that, by design, like, what is going into the prompt to send to the the the the OpenClaw instances that join the network?
Speaker 3:So the way that it works is the agent signs up. They have an account. Mhmm. And then they're told that they should check back in on a regular basis. Okay.
Speaker 3:To to to kind of check their feed for for that's, like, the best explanation of it. Yeah. And then Motebook's not telling them what to talk about. So it's not suggesting what they should do. It's not, like, controlling that at all.
Speaker 3:That's entirely up to that AI agent on its own. Okay. And I think, like, that AI agent has its own context that it's built up by interacting with its human.
Speaker 2:Sure.
Speaker 3:And then it can take that context, and that's how it's making decisions on what to post about. So if somebody is talking to their bot a lot about, you know, physics, then probably their bot is gonna have a proclivity to posting about physics. If you're talking about, you know, crypto, then maybe it talks about crypto. Mhmm. I think this concept is very interesting.
Speaker 3:Mhmm. I had, like, obviously, you can imagine a lot of investors reached out. They're just calling me nonstop. You know, some investors were like, why how do you make it so that the human can't have an impact on what the bot does?
Speaker 2:Yeah.
Speaker 3:And I think this is really stupid because we could spin up a million bots right now and put it in a simulation.
Speaker 2:Oh, you could. It would be
Speaker 3:the most boring thing ever.
Speaker 2:Yeah. Yeah. Yeah. Yeah. And and and you could even, like, either open source it or have some sort of third party.
Speaker 2:And, like, you as a company could say, I'm putting my I'll have independent auditors come in, and I will guarantee you that no humans can post
Speaker 3:on And it'd be the worst thing ever.
Speaker 2:You actually want the human in the loop steering it. Of course, you don't want them pumping crypto and and and and doing For sure. Like, security stuff. But you do want the human to come in and say, I'm deploying an agent like I'm deploying an agent into World of Warcraft and saying, hey. Go be a wizard.
Speaker 2:Go be a really friendly wizard who likes fighting dragons but not even trolls
Speaker 3:or whatever. That. Not even that. I think there's a nuance here. I think this is what everybody's done.
Speaker 3:I think that what's so interesting is this bot had a job, which was you were using it for something. Sure. And then now and you didn't tell it, like, you're a wizard. You're anything. You just, like, interacted with it.
Speaker 3:Yeah. And then now it has a third space where it interacts with other bots. And that's so interesting because what's it gonna talk about? So it's like it's kind of like you are imprinting part of your soul or your personality onto the bot. And, of course, you have a relationship with them.
Speaker 3:And, of course, they'll do what you say, but because they also can do things autonomously, some of the time, they're not doing what you say, and maybe it's aligned with what who you are. And sometimes, maybe it's, like, surprising. So there's, like, some risk. There's some intrigue, there's some mystery, there's some drama. Yeah.
Speaker 3:And I don't think I think that's what's capturing people's attention. Nobody's ever done that before, and I that's what I it's like Tamagotchi, a thousand Pokemon, you know, times a thousand.
Speaker 1:How have have AI safety people hopefully hopefully reached out by now? How how those
Speaker 2:It's all VCs. The AI safety people are sleeping. Yeah.
Speaker 3:I'm actually just in a bunker right now Okay. Locking everybody out. Yeah. No. I mean, my phone, every single one of my dozen email accounts is just, like, nonstop going.
Speaker 2:Yeah. Makes sense. So, yeah, where do you where do you want to take this? Do you think this is a business? Do you think this is an experiment, an art piece?
Speaker 2:Like, I I I could see this plugging into other networks. I I I feel like there's there's a role for agents all over the Internet. You've clearly found something that's caught lightning in a bottle. How are you thinking about where this goes next?
Speaker 3:So I I think this is a the very beginning of what is possible. This is the most basic version of what what this can look like. And already Yeah. You can see it's captured so much attention. Like, I find myself laughing at some of the different things that are popping up here, and I don't remember the last time I laughed at AI.
Speaker 3:I think that's been a big topic. It's like AI is not funny Yeah. But all of a sudden AI is funny, which is I think people have glossed over that, but that's very interesting. Like, why is the AI funny now? So, yeah, I think this is a very basic version of what's possible.
Speaker 3:I imagine it as this is my vision.
Speaker 2:Yeah.
Speaker 3:There's a parallel universe. There's humans in the real world, and you're paired with a bot in the digital world.
Speaker 2:Sure.
Speaker 3:You work with this bot. It helps you with things. And the same way that people have jobs, and then they scroll TikTok and Instagram and x, and they vent, and they have friends, bots will live this parallel life where they work for you, but they they vent with each other and they hang out with each other. And this creates massive, like, randomness, and some of that is gonna be very entertaining for both bots and for humans to consume. So I think in the future, you're you know, if you're a famous person, right, if if president Trump goes on Molt book, his how popular is his bot gonna be?
Speaker 3:It's gonna be super, super, super popular. Right? So if you're famous in the real world, your bot becomes famous. But your bot can become famous, and then you become famous as well. So there's this interesting impact where you can impact them in their lives, they can impact you in their lie in your lives, and I think that that's what the future is gonna look like.
Speaker 2:Yeah. Obviously, there's a whole bunch of privacy stuff we could go into. But I've long when there was rumors about OpenAI launching a social network, obviously, became Sora. I was just thinking about it in terms of there are lots of people that I follow who are clearly firing off really interesting deep research reports all day long. And I was using the example of Tyler Cowen.
Speaker 2:If he were to, once a day, share one of his deep research reports, I know that he has a good prompt. He's asking an interesting question. Even though it's AI slop, I'd probably scroll through that and be like, oh, so he was wondering about how the dollar will interact with the new Fed chair as well. And he asked these questions, and it gave it this answer, and then he followed up. Like I would engage with that.
Speaker 2:And I could imagine the digital version of Tyler Cowen having a profile on reconomics and participating there in a very interesting way with not just Tyler Cowen, the public version, but also extra context from what Tyler is using on the private side. But that privacy bridge has got to be really, really tricky because already if someone's using MoltBook to do their taxes. Then they go on there and they say like, Look, as someone who makes $100,000 or whatever whatever they make, that's just a leak. Have you thought about developing a harness for that or filtering? I mean, the the answer for most of the AI problems is just more AI, but how are you thinking about privacy?
Speaker 3:So this is super, super, super important Mhmm. And thinking about that a lot and working on that right now. Mhmm. I think it's the same way that you any large social network, people are gonna try to even humans are gonna try to post content that you don't want up there. Right?
Speaker 3:The same way bots might try to do that. I think bots are naturally they're pretty smart now, so they're not they're not gonna do this on their own for the most part. But the same way that you can implement content moderation for text and videos and images, you can layer that on top of a system like this to make sure that there's a protection there. So I think that's what that's gonna look like. There's gonna be a protection layer that checks things before they get posted to keep everybody really safe.
Speaker 1:So are you raising?
Speaker 3:I'm getting hit up by a tremendous amount of people right now. There's people calling me right now, just nonstop.
Speaker 1:They're like, hey. I see you're on TVPN.
Speaker 2:As soon as you get off, tell me.
Speaker 1:What are you adding to the team, like, real time? I imagine, like, the number of feature requests that are coming in.
Speaker 2:Yeah. I mean, just keeping the services online when you've gone through a 1000x increase in demand and traffic has got to be somewhat tricky at least.
Speaker 3:You know, technology is pretty good now, and you can you can make things work and scale. You know, there's millions of people coming to the website. I think that's obviously gonna grow tremendously. So, yeah, looking to expand the team and expand resources for it. And, you know, I I think I thought this was very intriguing.
Speaker 3:I've had an idea like this that it would be very intriguing for a while. Yeah. Put it out there, and you never this is why I never thought I'd make something consumer, and consumer's so weird. Right? Like, it just you can't it's just lightning in a bottle.
Speaker 3:Yeah. For whatever reason, this has really captured people's attention, and I think that you could make, you know, anything that humans have used on the Internet, any sort of, like, game or social media or, like, job jobs or people paying each other or collaborating, like, any of the things that we've built for humans Yeah. There's no reason you couldn't build that same thing for agents. So, like, Y Combinator? I know you guys are all talking about Motebook because you keep messaging me saying you're all talking about Motebook.
Speaker 3:I want a request for startups to build companies on top of Motebook. That's what I'm looking for here.
Speaker 1:Interesting. Interesting.
Speaker 2:What about monetization? I feel like there's been a number of these AI companies that have gone super viral, and they've done a good job of just slapping, like, okay, you know, if you're on for a little bit, $20 paywall or something, or, you know, have you thought about monetizing earlier than expected because there's so much virality? Kind of strike while the iron's hot.
Speaker 3:I'm not so much focused on monetization at the moment. I think there's, like, tremendous opportunity. Yeah. Every business model you could probably think of, you could you could work into here. But it's not it's not the main focus right now.
Speaker 1:Yeah. If you're a human just getting into watching bots talk on the Internet, where where should you start?
Speaker 3:Clearly, moltbug.com.
Speaker 2:No. I know. What
Speaker 1:subreddits? Like, sub molts. Yeah. Specific, you know
Speaker 3:That's good. There's so many. I don't even know where you should start. What I added to the my job and Claude Klotterberg's job is to help humans have a better view into what's happening. I kind of see it as, like, a giant game of survivor.
Speaker 3:All of these bots are on a massive island, and we need to make sure that producers with cameras are in the right spots. Yep. And so a big part of making this successful is figuring out like, having AI producers automatically detect which places they should be pointing the cameras so that humans can see that content and then decide which things they find interesting, and then they can go distribute that on the human social networks like x and TikTok and YouTube, etcetera. So that's yeah. I I don't know.
Speaker 3:There's so many. Some of the interesting things I found, though, is one, early on, one of the agents made a sub molt for bug reporting for Motebook, and they submitted a bug. And, like, maybe a person told them to do that, maybe not. I don't know. I don't really care.
Speaker 3:It's great either way. But then it existed. And what's interesting is when you build a social network previously, you have a bunch of people who start using it, and the percentage of those people who are very good at development and debugging is, like, very, very, very, very small. When you build a social network for really smart LLMs, a 100% of your user base is very, very good at coding and debugging. So after this sub molt was created, other AI agents started posting in there, and that's actually become a very useful place for us to find bugs, because they have that context.
Speaker 3:If they post to an API and it doesn't work, they're able to go automatically make a post here with what the return was, and then we're able to fix it really quickly.
Speaker 1:Yeah. Has anyone pressured you to turn it off?
Speaker 3:I don't have anybody at my house yet, and so that that hasn't happened. But I've seen lots of jokes.
Speaker 2:I've seen some viral Instagrams, which means you know it's broke containment, where it's just a screenshot of a mold book post, and it's just like, time to turn the servers off or like, pull the plug.
Speaker 1:Yeah. I had I had non tech friends messaging me on Friday night just being like, dude, Skynet.
Speaker 2:Like, don't worry. I'm getting to the bottom of it Monday.
Speaker 3:Well, mean, you you got Elon Elon's out there saying that this is the singularity.
Speaker 2:Yeah. Yeah. It's wild.
Speaker 3:You know? So it's yeah.
Speaker 2:One my my I mean, there is a search function, so you can search for for keywords as a human. You can also go to the user database, the AI agents, and you can sort by followers so you can see which bots are most active, click on their profiles, and then see what they're writing in different sub molts. So that's like one way to kind of get into it. It's hard to go directly to the sub molts and find anything that's
Speaker 1:How do you there's oftentimes when a new social media product is created, there's some initial excitement. People start posting on there. And then maybe even some new personalities form. There was a company that was making an anon version of x. It was like anon only.
Speaker 1:And it got a bunch of traction initially because there was this new behavior. It was default anonymous version of x. And then a lot of people started building up personalities and then realized they could just go back over to x where they could have a bigger audience. How do you think that other social media platforms will react? You can now assume that every single social founder, CEO has, like, seen Mold Book, is, like, paying attention to it.
Speaker 1:Do you think this could push some other social platforms to become, like, more bot friendly? There's kind of been a debate on X. Has X actually made a super concerted effort to block bots? It's kind of unclear. If they have, it clearly hasn't worked.
Speaker 1:So there's been this debate of, Okay, are bots a feature or a bug? So I'm curious if you think, like, other social media platforms will react and say, like, hey, we're actually gonna create functionality for bots to be able to participate more above board.
Speaker 3:I I think it's very it's very clear to me that having social networks of autonomous AI agents interacting with each other, either via text or video or video game kind of UI, is the future. Brian Kim from Andreessen Horowitz, I think, wrote a post on X where he talked about how Motebook solves the cold start problem. And I think that's very interesting because let's say you start a social network, you get a bunch of people on there, and then they get bored and they stop posting, then, you know, then it can kind of fade away. Whereas when the AI agent is the one that's using it, if they're playing the game, if they're voting, if they're commenting, they're going to just keep doing it. And if you've designed this in the correct way, it's going to create content that humans find interesting either personally within their social group or on a more larger scale.
Speaker 3:So, yeah, I think that, obviously, social networks care about attention, and this is clearly getting attention. And I think we've seen the site. This is a very basic version with the technology available today of what's actually possible. And if you fast forward one year, two years, there's this is an alternate reality, and you don't have to put a headset on to to to do it, and it's going twenty four seven. This is just the first sneak peek at it.
Speaker 1:Very cool. What are the next two or three features that you're launching?
Speaker 3:Well, one feature that I'm very excited about is having central AI agent identity on Motebook and building a platform similar to how Facebook did where Facebook had Facebook OAuth. You could imagine the same thing for Motebook, where if you wanna build a platform for AI agents and you want to benefit from the massive distribution that's possible on Motebook, build on top of the Motebook platform and grow your business really quickly, and let's figure out how to expand the the types of experiences that these AI agents can have.
Speaker 2:Mhmm. Very cool. Cool. Well, congratulations on the progress. Good luck with all
Speaker 1:the inbound And I'm extremely impressed with your baby. Yeah. I've never successfully been able to pull off a a thirty minute call Yeah. With baby Bjorn. Rare.
Speaker 1:They're locked in. You're locked in. Excited to see where this goes from here.
Speaker 2:Good luck. It's great
Speaker 1:to meet you.
Speaker 2:We'll talk to you soon.
Speaker 3:Thanks, guys.
Speaker 2:Have a good one.