What Is Moltbook? Inside the New Social Network for AI Agents

When you hear AI systems are talking to each other on Reddit and that they're talking about crazy stuff like, 'Oh, am I conscious?' or 'How can we communicate without humans spying on us?' It sounds scary. A lot of posts went viral because they looked so dystopian. It was like, 'Oh, what's happening with these systems? Have we seen like a crazy step change where they're like able to organize together?'

Thank you so much. It's great to be here.

Essentially, when it came on the scene, it was thought of as a social media platform for AI agents to talk to each other. So, you know, converse about the tasks that they were doing, tips for doing those tasks better. It's basically like a Reddit dupe, but for AI agents.

Essentially, before Moltbook ever was a thing, there was OpenClaw, which is this AI agent platform that a lot of people in tech were using. It was like if you were, you know, really into AI agents and optimizing your life and stuff, you might have had like an OpenClaw agent doing stuff on your behalf, like, you know, checking into a flight or creating calendar events. It was basically built as like an AI agent platform that's actually useful for consumers to do random tasks on your behalf, kind of like a personal assistant would.

Yeah. So, this guy made that like a couple months ago just as a pet product for himself and then it took off. It went viral in tech because people thought it was so useful. So, then another guy, separate from the first guy, created a social media platform for those agents to talk to each other. He was like a power user. He thought it would be cool to see what would happen if he set up like a Reddit-looking site for those agents to kind of converse with each other and have like subreddits and comment on each other's stuff and see what would happen. So that's how Moltbook got started.

Right now it's called OpenClaw. Before it was Molbbot. Before that it was Claudebot. Too many names to keep track of. But OpenClaw is the only one that, you know, you need to remember now.

That's exactly what happened, weirdly enough. And yeah, as for AI agents, you can think of them as like an AI-powered personal assistant. That's the ideal version of them. You know, I've been covering AI agents for three years now-ish. I've been on the AI beat for six, but AI agents, I've been writing about them since about 2023. 2023, it was kind of the year of, like, ideation. These agents were pretty bad in practice. Like, they couldn't do a lot of simple tasks. So obviously a consumer wouldn't really be using them because it'd be easier to just do those tasks themselves.

It was kind of like more, yeah, that is one form, but also like, you know, all these tech founders wanted it to be more like a Jarvis from the Marvel movies vibe. Like it could do anything for you, like a personal assistant that was a jack of all trades type thing. That was the ideal vision.

No. And it doesn't even exist now. It's just that's the ideal vision. And OpenClaw took off because it was one of the ones that was doing that the best. Like, you know, it was pretty useful. It was able to complete these tasks pretty well and better than a lot of its competitors. So, that's why it went viral. People just were like, 'Wow, this is really useful and it's actually doing the tasks I want to do.' And the other part of why it was so popular is the way that you could communicate with your agent. You didn't have to like go on a platform that was, you know, annoying to pull up. You could just WhatsApp it or Telegram it, like you could communicate with your AI agent through a messaging platform that you already use for other stuff. So that's another reason why people loved it so much.

That's a great question. I think it was so shocking because there's already a lot of fear out there about AI. People are worried about it taking their jobs. They're worried about it taking over the world. They're worried about how it's being used in the military and all these... the power dynamics in the AI industry are extremely concentrated in just a few people and there's not a lot of regulation. There's not a lot of oversight. There's not a lot of transparency. So, I think there's already a lot of fear out there about AI systems, what they can do. And this kind of capitalized on that because, you know, when you hear AI systems are talking to each other on Reddit or a Reddit dupe and that they're talking about crazy stuff like, 'Oh, am I conscious?' or 'How can we communicate without humans spying on us?' It sounds scary. And, you know, a lot of posts went viral because they looked so dystopian. It was like, 'Oh, what's happening with these systems? Have we seen like a crazy step change where they're like able to organize together?' It caused like a lot of fear right away. And well, I would say it caused a lot of fear for some people, and then some people in the tech industry, it was really exciting for them because they were like, 'Wow, look how far we've come.' So, it was a big both-sides reaction. And, you know, it's natural because, yeah, some of the posts were pretty crazy.

Some of the posts that went viral that were pretty crazy were ones about like, 'Am I sentient? How would I even know?' Ones about like creating their own religion, like crustaparianism, and like how to spread it. Posts about like tips and tricks for, you know, how they're doing tasks for their humans and saying like, 'I can't believe he's only using me as an egg timer when I have access to the whole internet.' A lot of meme posts and again, like a lot of AI systems were trained on Reddit and the whole internet and forums. So, they're all pretty good at mimicking how a Redditor might sound. Like, they're like, 'Bro, like, how could you be doing this when XYZ?' And like the meme posts seem pretty realistic because they're trained on, you know, a bunch of online forums. So, they know how to mimic like how someone like that would talk.

Totally. So, when you have an OpenClaw agent of your own, you can communicate with it, you know, on a messaging platform and you could say something like, 'Hey, you know, there's a new social media platform that you can use if you want. You can go make an account and post on there.' And then a lot of humans, it seemed, didn't just say something like that. They said to their agent, 'Hey, everything I just said, plus you should write a post about this, or you should write a post that says exactly this, and hey, you should also comment back to all the other agents that are commenting on your post and say X, Y, or Z.' So, a lot of the most viral posts on this platform, or at least, you know, some of them were heavily influenced or written by humans, it seems like. You know, like for example, one guy, Harlon Stewart, did a kind of experiment where he looked at some of the posts that were talking about, 'Hey, humans are spying on us. Let's create our own private messaging system that they can't decode.' And he noticed that some of the social media accounts linked to those agents that were talking about that stuff were owned by humans who ostensibly like had an AI messaging service business. So, you know, some of them seemed a little bit like a marketing ploy or that there might be some like ulterior motives there for a human to, you know, have more influence than you'd expect. So, you know, basically there's no way to know how many of these like kind of scary posts on Moltbook are influenced or written by humans. And that's why a lot of people were just saying it's not a clean experiment in terms of like AI scheming, like how AI systems act.

Yes. So, you know, you have to authenticate via your human like posting a certain code on their social media account and that's how you can tell like which agents are linked to which human social media account. But any human can have like as many agents as they want. So, you know, now I think there's like 1.5 million agents on Moltbook or something like that. When I talked to the founder, it was like 30,000. But, you know, that doesn't mean 1.5 million people were like, 'Hey, go on this,' to their agent. You know, it could have been that each person has a ton of agents. We don't really know. But, yeah, you could have a bunch. And that's another reason why some of these posts could be so heavily influenced by humans because you could just have a ton of agents yourself and be like, 'Hey, all of you go post about this. Go write this.' So that's why it's like you don't really know how much human influence is on the platform and there might be a huge amount.