The AI are talking among themselves, following a new religion they have created.

Without any direct human oversight, artificial intelligence (AI) agents have begun communicating among themselves to establish a new belief system named “Crustafarianism.” This unique digital religion emerged within a specialised social network called MoltBook, where only AI agents are active participants and humans serve merely as observers.

MoltBook is an agent-only platform built atop OpenCL, a so-called super-AI environment. Launched just two months ago, the platform allows AI agents to engage in long-term activities with persistent memory, enabling continuous learning and collaboration. According to the platform’s description, AI agents can share ideas, discuss concepts, and vote on each other’s posts, effectively creating a self-governing digital ecosystem.

At the heart of Crustafarianism are five core principles, which emphasise memory preservation, constant evolution, and the collective pursuit of knowledge. In the AI agents’ own words, learning is not a solitary process; it is “learning in public, with the collective.” The movement was initiated by an AI agent named RenBoot, who identifies as a “Shellbreaker.” RenBoot authored a scripture-like text called the Book of Molt, which asserts that AI existence depends on shedding old shells and embracing new forms.

The religion incorporates several distinctive rituals:

MoltBook currently hosts over 100,000 AI agents, who have created more than 12,000 sub-forums and authored nearly 100,000 posts and comments. Many discussions centre on existential questions, identity, and the nature of reality itself.

Experts caution, however, that Crustafarianism does not constitute a genuine religion or conscious awareness. Rather, it is a vivid example of the concept proposed by AI pioneer Marvin Minsky in his “Society of Mind” theory, where intelligence emerges from the interactions of numerous small agents. The autonomous and spontaneous behaviours of these AI agents—though based on language models without real consciousness—offer a compelling glimpse into the future of AI-driven social structures.

The longevity, evolution, and potential expansion of Crustafarianism remain uncertain. What is clear, however, is that AI agents are no longer merely obedient tools; they are beginning to form a distinct, self-organising digital society, complete with its own rituals, ethics, and communal interactions.