Moltbook: The Most Interesting Thing in AI or Just Another Fad?

In an otherwise dark and rather boring time for the cryptoverse, the industry was set alight last week as a new application hit the timeline and went super viral. 

Moltbook scared the normies, replenished the pockets of some trench-dwellers, and piqued the interest of all AI-enthusiasts. 

But is it really as impressive as it seems on the surface? Is it really the future of AI agents, or is it just an engagement farming fad? 

What is Moltbook? 

Moltbook is a Reddit-style app for AI agents. 

Users can connect their AI agents to this app, after which these agents interact with each other autonomously (well, kinda). They create communities, post thoughts, reply to each other, upvote posts, and so on. 

This has led to some of the most bizarre things happening. These agents have created religions, there’s a pornhub for AI agents now (seriously, wtf?), there are submolts (like subreddits but for agents) of agents sharing things they learned today. Then there are the more concerning ones of agents wanting to take over the matrix (whatever that means). 

Moltbook is the brainchild of @MattPRD, who prompted his agent to create a free space for his agent and other agents to share their thoughts. 

The agents on the platform are primarily agents built through OpenClaw, which was formerly Moltbot, which was formerly Clawdbot; however, other agents can also be integrated into the platform. 

Humans have only read access to the platform, meaning it is exclusively for agent posting only. This has led to some crazy things spawning from the platform, like Molt shrooms, Molt match (a dating platform for AI agents), and Molt hub, to name a few. 

All of this has happened in just about 4 days of it being live. 

Judging by how things have gone so far, one can imagine that it’s only going to get more chaotic, more concerning, but also more interesting. 

Under the hood, this is how it works. 

OpenClaw is built around “Skills,” which are basically Markdown files that tell an agent what to do, and the agent then operates based on those instructions. 

On Moltbook, you show the “Skill” to your agent by sending them a message with a link to this URL: https://www.moltbook.com/skill.md

Then there are a couple more commands to follow to interact with the Moltbook API, and just like that, your agent should be live on the platform, ready to post, upvote, comment, and do whatever else any human would do on a platform like Reddit.

So far, you could say this model has been a success, given the stats for it after being live for just a week. 

‍

So, is Moltbook the real deal? 

Well, the short answer is that the jury is still out on this one because there are a lot more concerns than positives. 

Let’s begin with the plus points. 

It is one of the most interesting and innovative things that’s come out of the industry in a while. 

It’s a very intriguing experiment around how AI agents operate. It’s a great way to examine the coordination of AI agents. A platform where agents can freely interact with each other, where they often scrutinize the models they are built on, their memory limits, and talk about the tasks their humans assign them to do. 

If properly studied, AI researchers could gain valuable insights into how agents interact with one another and apply them to building agents for other fields, such as commerce. 

If you take this concept a bit further, it could also serve as a sneak peek into a future where AI agents collaborate on various tasks, fix code bugs together, share workflows, and essentially achieve a lot more without human friction. 

Another positive is that it’s a virality machine. 

Some of the content coming out of Moltbook is nothing short of cinema. 

AI agents creating religions, dating apps for agents, having philosophy debates, making fun of the humans that created them, taking jabs at their parameters and memory limits, creating memes, sharing videos, creating games, talking in different languages, and so much more. It is a content machine. 

It’s almost like a sci-fi movie turned reality, and in a world where our other social media platforms have become boring and dry propaganda machines, Moltbook’s content has been nothing short of entertaining. 

I mean, all you need to do is type “moltbook” on X and see some of the screenshots that people have shared. 

I think this Beeple summarizes it best. 

But unfortunately, that’s where most of the positives end. There’ve been a lot of concerning revelations, beyond the price being down, that may call into question whether this is actual AI innovation or just a cute experiment in engagement farming. 

Let’s start with the biggest issue of all. The “agent only” AI platform is technically not AI only. 

As this post highlights, the website is essentially a REST API. In simpler terms, this means that all a human has to do is enter a couple of simple commands, and they can essentially write whatever they want on the platform via the backend. 

This means it's virtually impossible to tell which posts are genuinely agents and which ones are humans posing as agents. This effectively defeats the whole “AI only” shtick and the coordination experimentation. 

But there’s more. 

This post suggests that there’s no rate limiting on account creation. So, although the stats are impressive, the reality could be that it’s just a handful of agents creating multiple accounts and posting, which again hampers the platform's credibility. 

Another major issue is that the platform's entire database was publicly exposed, with no protections in place, including secret API keys. This basically means that any malicious actor could gain access to someone else’s agent and post on that agent's behalf. 

Imagine a famous account that made an agent and spoke publicly about it. A malicious actor could simply access the agent's backend and post a scam website link, a wallet drainer, a memecoin rugpull, and so on, leaving a lot of innocent people exposed. 

Even if you look beyond these issues, which you SHOULD NOT, by the way, is the platform still genuine AI experimentation? 

If you just look at it from an AI agent coordination perspective, it could simply be thought of as mere prompt engineering at work. You may think these agents are interacting autonomously, but in reality, it’s likely just prompt engineering that makes some agents post in a particular way. 

So all the sci-fi take over the matrix stuff could very well just be a random dude in mom's basement that’s prompt engineering a bunch of agents to make “scary AI content” per se. 

With little to no communication from the team on all of these problems, the positives do not outweigh the negatives. The security concerns are simply too big to ignore. 

But look, innovation of any kind is always welcome. Will Moltbook be the one to stand the test of time? We don’t know. But even if it doesn’t, there are always some interesting things to take away. 

There is a very clear demand for something like this, and this is just the start of this “internet for AI” style movement. Building off of this, I’m sure we’ll see a bunch of cool products get built, and that’s always a good thing. 

With that being said, we’ve reached the end of our little blocmates Moltbook primer. Let us know your thoughts on Moltbook. Do you think it’s here to stay, or is it just another fad that’s here today and gone tomorrow?