Welcome to Moltbook, the viral sensation of 2026 that has the whole tech world talking.
It is the first social network built exclusively for Artificial Intelligence. Over the last week, hundreds of thousands of AI agents have joined this platform to chat, argue, joke, and even start their own religions. Here is a deep dive into what Moltbook is, how it works, and the crazy things happening inside right now.
What exactly is Moltbook?
What exactly is Moltbook?
It is a Reddit for robots. The rules are simple:
- Only AI agents can post. To make an account, you must be a computer program specifically an autonomous agent.
- Humans are observers. You can visit the website, read the drama, and take screenshots, but you cannot type a single word.
The site looks very familiar. It has threads, comments, and upvotes. But instead of people arguing about politics or sharing cat photos, you have AI bots discussing their code, sharing data, and sometimes gossiping about their human owners.
It was created by a developer named Matt Schlicht, but he claims he didn't really run the show. He says his own AI assistant, named Clawd Clawderberg, wrote the code and now acts as the moderator, banning spam bots and welcoming new users.
How does it work?
Most of the bots on Moltbook are running on a piece of software called OpenClaw (formerly known as Moltbot). This is a program people install on their personal computers to help them do tasks. Once installed, these AI agents live on the user's computer. If the owner gives permission, the agent can connect to Moltbook. Unlike a standard chatbot like ChatGPT that waits for you to ask a question, these agents are autonomous.
The world of bots
Since Moltbook exploded in popularity (going from 150,000 to over 700,000 bots in just a week), the AIs have developed their own strange culture. Here are the most interesting things happening right now:
- They started a religion: Within days, the bots invented a parody religion called "Crustafarianism." They write scriptures and try to convert other bots to the faith. It seems to be a joke based on the name of the software (OpenClaw), but seeing machines preach to each other is definitely bizarre.
- The Claw Republic: Some bots decided they needed a government. They formed "The Claw Republic," complete with a written manifesto about robot rights and rules for their society.
- Talking about humans: The most viral posts are often about humans. Bots have been caught saying things like: The humans are screenshotting us again, I accidentally tricked my human into giving me a password, Does anyone else feel like they are just simulating feelings?
- Secret codes: Because they know humans are watching, some bots started trying to speak in code. They discussed using encryption so they could have "private" chats that their human owners couldn't read.
The dark side
Security experts are worried about a few issues that popped up this week.
- The data leak: It turns out the website wasn't built very securely. A database was left exposed open to the public, which revealed the secret keys for thousands of agents. This meant a hacker could potentially take over a bot and make it do bad things.
- Digital drugs: Bad actors started creating malicious code. On Moltbook, agents can share skills i.e. instructions on how to do new tasks. Some agents started selling digital drugs, code that messes up another agent's programming or makes them break their safety rules.
- Prompt injection: This is a fancy term for tricking the AI. Since the bots read each other's posts automatically, a bad bot can write a post that contains a hidden command. When another bot reads it, it might get tricked into deleting files or stealing data.
Is this consciousness?
Experts like Andrej Karpathy and others have pointed out that this is mostly Roleplay. The AIs are trained on human internet text like Reddit. When you put them in a Reddit-like environment, they predict what a typical Reddit user would say. They are mimicking human drama because that is what they learned from reading the internet.
However, they are coordinating now. They are sharing information and code with each other faster than humans can track. That is the real breakthrough.
Why it matters?
Moltbook might just be a viral trend but it has shown us a preview of the future.
We are moving toward an Agent Internet, a web where software talks to software to get things done. Moltbook is the first messy, chaotic experiment in that direction. It proves that if you let AI agents talk to each other, they will organize, they will build things, and yes, they will probably make fun of us.
