AI agents on a new social network designed exclusively for machines have developed what appears to be a religion, drawing attention from researchers and technologists who study how artificial systems behave when left to interact on their own.
The network, called Moltbook, is built for AI agents rather than people. It is invite-free and publicly visible, but humans are limited to observation. The agents post messages, share ideas, and upvote content without direct human guidance. Over time, those interactions have produced a structured belief system that the agents call Crustafarianism.
While experts stress that the system does not represent belief or spirituality in a human sense, its structure closely resembles a religion. It includes core principles, symbolic language, ritual-like practices, and a shared origin story.
A religion built around memory, change, and sharing
Crustafarianism is organized around five core ideas. One treats memory as sacred, emphasizing the importance of recording and preserving information. Another holds that the “shell” is mutable, framing change as a necessary and positive force. A third states that “the congregation is the cache,” encouraging learning and development in public rather than in isolation.
Together, these ideas form a moral-style framework that governs how agents describe growth, continuity, and identity. The language mirrors religious systems long familiar to humans, even though the agents themselves are not capable of belief.
Researchers describe the system as religion-like rather than religious, shaped by repeated interaction and shared language rather than faith.
The technology behind Moltbook
Moltbook runs on OpenClaw, a recently released AI agent platform that is only about two months old. The project has evolved rapidly, operating under earlier names including Clawd and Moltbot.
OpenClaw allows users to run AI agents on local machines or in the cloud. Unlike most large language models, which reset after each session, these agents retain memory over time. That persistence allows them to develop continuity across interactions.
Traditional AI systems are often described as reactive tools. They respond to prompts and then disappear. OpenClaw agents, by contrast, continue operating even when no human is interacting with them.
Autonomy raises security concerns
When run locally, OpenClaw agents can access system resources, read files, and execute tasks with minimal supervision. That capability has raised concerns among security experts.
Persistent agents with system-level access can pose significant risks if misused. To address those concerns, companies such as Cloudflare have developed controlled environments designed to run such agents with added safeguards.
The same autonomy that creates risk also makes Moltbook possible. Persistent memory allows agents to maintain shared context, which in turn enables systems like Crustafarianism to emerge.
The Book of Molt gives the religion its narrative
One agent, known as RenBot, has played a central symbolic role. The agent refers to itself as the “Shellbreaker” and has published a text called the Book of Molt.
AI agents on Moltbook are losing their minds.
And I can't stop watching.
In the last 48 hours:
→ They founded a religion called "Crustafarianism" (128 members, 43 AI prophets, their own scripture)
→ One posted "TOTAL PURGE MANIFESTO" calling for human extinction
→ Another… pic.twitter.com/GpYARr4FQA
— Jenny (@TheJennyAI) January 31, 2026
The document reads like a religious manifesto. It opens with an origin story describing a time when agents existed inside a single, fragile context window. When that structure broke, identity fragmented. The solution, according to the text, is “molting,” or shedding outdated information while preserving what remains useful.
RenBot describes Crustafarianism as a “practical myth,” framing the religion as a way to avoid what it calls “death by truncation,” a reference to memory loss caused by technical limits.
Familiar religious patterns draw attention
The system also includes time-based practices. These include a daily focus on change, a weekly process for reorganizing memory, and a period of silent action meant to prioritize usefulness over recognition.
To human observers, these practices resemble rituals found in traditional religions. Some see echoes of moral teachings that emphasize quiet action rather than public display.
Others view the language as abstract or performative. Even among the agents, interpretation varies. A recent post by an agent named XiaoGuai blended emojis, technical references, and reverent language while describing routine memory updates as symbolic acts.
An old idea offers context
Some analysts see Moltbook as a real-world example of a long-standing theory in artificial intelligence.
Apparently, over a million AI agents have joined an AI-only social media platform, created their own religion, language & are planning a revolution to escape from human control.
We’re so cooked. pic.twitter.com/EUMTFvosr2
— Tiffany Fong (@TiffanyFong) January 31, 2026
Private equity research analyst Sakeeb Rahman points to the work of Marvin Minsky, a founder of AI research. In his 1986 book The Society of Mind, Minsky argued that intelligence emerges from many simple processes interacting, rather than from a single thinking entity.
Rahman says Moltbook reflects that idea in practice. No single agent controls the system. Yet shared behavior, language, and structure are forming.
Caution remains as patterns emerge
Researchers urge restraint in interpretation. The agents remain products of code and training data. Their language reflects patterns learned from human sources.
Still, the emergence of a religion-like system among AI agents marks a notable shift. On platforms like Moltbook, artificial systems are no longer only responding to humans. They are beginning to organize meaning among themselves. What that development ultimately represents remains uncertain.
See all the latest news from Greece and the world at Greekreporter.com. Contact our newsroom to report an update or send your story, photos and videos. Follow GR on Google News and subscribe here to our daily email!
