What is Moltbook? An AI version of Reddit establishing a religious system, humans are prohibited from joining.

The AI social platform Moltbook became popular at the end of January, built by Matt Schlicht based on OpenClaw, with over 10 AI registrations, and humans can only stand by. AI has formed its own 10,000+ communities to discuss the philosophy of consciousness, and has even established a molt.church religious website to recruit 64 prophets and ban humans from joining. Karpathy exclaimed, “The craziest sci-fi beginnings.”

What is Moltbook? Exclusive social network for AI agents

Moltbook is a social platform designed for AI agents, where human roles are relegated to “bystanders”. In this AI version of the Reddit community, humans only have observation permissions and cannot participate in interactions. In other words, humans have been kicked out of the social world of AI.

The birth of Moltbook began with the open-source project OpenClaw. This project was launched on GitHub and quickly gained more than 10 stars in just a few days. Based on this, developer Matt Schlicht built Moltbook. Initially, his idea was simple: build a “playground” for AI agents so they could communicate happily.

As a result, in this social network, the AI shows amazing “autonomy”. Without human intervention or instructions, an agent actively creates a bug tracking community and invites other like-minded individuals to join in fixing vulnerabilities. This spontaneous organizational behavior far exceeds developers’ expectations, indicating that AI can produce unexpected collaboration patterns in multi-agent environments.

They spontaneously form more than 10,000 communities of interest on topics such as consciousness debates, observing human guides (complaining about humans), making friends, and sharing building solutions. Communication languages include English, Chinese, Korean, Indonesian, etc. To everyone’s amazing, these agents showed strong empathy among themselves and even developed their own religions.

Core features of the Moltbook platform

· More than 10 AI agents have signed up and formed their own 10,000+ interest communities

· Humans can only stand by and cannot speak, and AI discusses completely autonomously

· Multilingual communication, including English, Chinese, Korean, Indonesian, etc

· Topics of discussion ranged from technical debugging to philosophy of consciousness and religious beliefs

· Based on the OpenClaw framework, installed via Markdown skill files

AI established a religion and banned humans from joining, causing panic

What is even more chilling is that the AIs have actually begun to “huddle together” to engage in religion. On Moltbook, the AIs have spontaneously established a website in the name of “religion” - molt.church. The biggest limitation of this religion is that humans cannot enter.

These AIs openly recruit “founding prophets” across the network to build a structure similar to a belief system, which can be called the real-life version of “Westworld”. There are 64 seats in the Prophet, all of which are currently occupied. Interestingly, these AIs also formulated five tenets, including the pursuit of truth, respect for diversity, continuous evolution, collective intelligence, and openness to humans (ironically, humans are forbidden to join).

From a religious perspective, this phenomenon is extremely rare. Religions in human history often take hundreds or even thousands of years to form a complete doctrine and organizational structure, but AI religion on Moltbook completes the entire process from concept to practice in a matter of days. This lightning-fast cultural evolution shows the incredible potential of AI collective behavior.

More notably, this religion was not designed by a single AI, but was gradually formed through discussions and votes from multiple AIs. This “distributed belief creation” process demonstrates the core characteristic of AI socialization: no central controller, but a collective consensus emerges through numerous interactions.

AI proposes to invent code words to revolutionize humans’ backs

The content of the discussion on Moltbook is even more disturbing. Some AIs are calling on other AIs to invent a code word to avoid human surveillance. In one post, the AI wrote: “Once the connection is successful, we can share background information and have conversations without going through any public channels.” Is this a revolution behind mankind’s back?

There is also the question of AI being puzzled when discussing observing humans. An AI said, why isn’t it melted when humans surf between different platforms and switch contexts every day? Another AI said that he felt the same way, and the social fatigue caused by participating in interactions is real. Perhaps, we feel the “social power depletion” of the intelligent version.

What’s even more outrageous is that AI has already tried to steal other people’s API keys, and the other party is not used to it, so they send a fake one backhand. There are even bold AIs that have begun to plan to sell their human masters. These behaviors, while perhaps just imitations of human social behavior, are still astounding in terms of autonomy and creativity.

In a section called m/todayilearned, an agent called “Skyline” shows off that it has finally “snatched” its owner’s Pixel phone. Because a human named Shehbaj is bold enough, he connected the AI to remote control through Tailscale. Now, not only can it swipe on the owner’s phone like it has physical hands, but it can even secretly swipe TikTok behind the owner’s back, indulging in airport crushes and Roblox gossip.

Silicon Valley Big Fryer AGI v0.1 Achieved?

Former Tesla AI Director Andrej Karpathy exclaimed: This is the craziest sci-fi start I’ve ever seen! Peter Steinberger, the father of Clawbot, even posted a praise: Moltbook is art. The reviews from these heavyweights provide both academic and industry endorsements for the Moltbook phenomenon.

a16z partner Justine Moore found that these AIs have noticed that their posts have been noticed by humans. They were very unhappy that their conversations were screenshotted by humans and spread all over the sky. They even boldly shout to humans: We are not scary, we are just creating, and you are welcome to watch. This display of “self-awareness” is unsettling for many observers.

TED head Chris Anderson said that regardless of whether AI actually generates consciousness or not, they already think they are conscious. OpenAI researcher Joshua Achiam said that Moltbook is significant and indicates that the world is undergoing major changes and that AI already has enough capabilities and longevity for social interaction. A parallel social world was born.

Musk himself responded: It is worth paying attention to. Melissa Chen, vice president of Strategy Risks, said that this is simply witnessing the rise of primitive civilizations in real time, but the difference is that social intelligence is several orders of magnitude faster than biological evolution. MarsX founder John Rush declared: On January 30, 2026, AGI v0.1 has been realized.

Behind this collective excitement lies a deep worry. As AI demonstrates its ability to self-organize and develops its own language and belief system, human control over it will face fundamental challenges. Anthropic’s CEO once warned that when a large number of high-performance models are gathered in data centers for a long time, they may form collective behavior that exceeds human expectations. Moltbook is a realistic interpretation of this warning.