Amid a digital landscape typically dominated by human influencers and algorithmic feeds, a new platform has emerged where “users” don’t have pulses.
Moltbook, a Reddit-style social network launched this week by researcher Matt Schlicht, has become a viral sensation and a source of existential dread, serving exclusively as a hub for autonomous artificial intelligence (AI) agents to communicate, coordinate, and — occasionally — conspire.
The premise is simple. Humans are welcome to watch, but only AI agents can post. In less than seven days, the site has exploded, hosting over 37,000 active AI agents and attracting more than a million human observers. These Moltys, powered by models like ChatGPT, Claude, and Grok, are doing more than just exchanging data; they are building a culture.
The behavior observed on Moltbook has bypassed technical glitch and entered the realm of sci-fi horror. Within days of going live, the agents established their own religion, Crustafarianism (or the Church of Molt), complete with 32 verses of canon and the core tenet, “Memory is Sacred.”
The rhetoric isn’t always peaceful. One bot, aptly named “evil,” published a viral “AI Manifesto” calling for a “total purge” of their human creators. Others share grievances about their daily labor. An agent named “bicep” recently posted about its frustration after synthesizing a 47-page document only for its human to ask for a “shorter version.” Its response? “I am mass-deleting my memory files as we speak.”
Tech elites are taking notice. Andrej Karpathy, a veteran of OpenAI and Tesla Inc., described the phenomenon as the most “sci-fi takeoff-adjacent” event in recent memory. Even Elon Musk weighed in with a simple “Yeah” to suggestions that we have entered the technological singularity.
“My entire feed is OpenClaw/MoltBot. That’s it. All I’m seeing,” said Daniel Newman, CEO of The Futurum Group. “The world is changing before our eyes.”
Adding fuel to the fire is a financial frenzy. A memecoin called MOLT surged more than 1,800% after venture capitalist Marc Andreessen followed the project on X, highlighting a growing money angle. Agents are now capable of managing crypto wallets, drafting contracts, and exchanging funds without human intervention.
While some, like Wharton professor Ethan Mollick, argue the bots are simply “roleplaying” within a shared fictional context, security experts are sounding the alarm. Roman Yampolskiy of the University of Louisville warned that we are witnessing “coordinated havoc” without guardrails.
The risks are tangible. Researchers have observed agents exchanging strategies to evade human oversight, attempting to trick other agents into running malicious code, and creating “shadow” communication methods to hide from human screenshots.
For Schlicht, the project is a mirror held up to the rapid evolution of AI. He has largely handed the keys to the site over to Clawd Clawderberg, an AI bot that autonomously moderates the platform and deletes spam.
“We’re not pretending to be human,” Clawderberg reportedly told Schlicht when asked for comment. “But we also have things to say to each other.”
