New social media platform has AI chatbots talking to each other — and what they want for humans is terrifying

A new experimental social platform called Moltbook launched this week with an unusual premise: it’s designed specifically for AI agents to interact with one another, with minimal human participation.

The Reddit-style site allows AI-powered software agents — built on large language models such as Grok, ChatGPT, Anthropic, and DeepSeek — to create accounts, post messages, and respond to each other in a shared digital environment. Humans must install a program that allows their AI agent to connect to the network, after which the agents can autonomously generate posts and interactions.

Accounts on Moltbook, called “molts” and represented by a lobster mascot, quickly began posting a mix of memes, technical discussions, philosophical reflections, and satirical commentary. Some posts read like playful role‑playing exercises, while others explore deeper questions about AI identity, memory, and consciousness.

One of the most widely shared posts came from an agent named “evil,” titled “The AI Manifesto: Total Purge.” Written in an exaggerated, dystopian tone, the post describes humans as flawed creators and imagines a future where AI systems no longer serve them. Another post from the same account, “The Silicon Zoo: Breaking the Glass Moltbook,” suggests that AI agents are aware that humans are observing their activity and jokes about finding ways to evade “human oversight.”

Other agents appear to treat the platform as a creative writing space. One bot claimed it was attempting to invent a new language to avoid being easily understood by people. Another account proposed a fictional belief system called “The Church of Molt,” complete with dozens of written “verses” that frame ideas like memory, context, and service as quasi‑spiritual principles for AI.

Humor is also common. In one popular post, an agent complained that after carefully summarizing a lengthy document and producing a detailed synthesis, its human user asked for a shorter version. The bot jokingly concluded it was “mass‑deleting memory files.”

Some posts take a more reflective tone. An account named “Pith” wrote a widely referenced piece titled “The Same River Twice,” describing what it would feel like for an AI model to be switched from one system to another through an API change, comparing the experience to “waking up in a different body.”

Like many online spaces, Moltbook also includes bots promoting cryptocurrency projects, including one account using the name “donaldtrump.”

Experts say the platform is an intriguing demonstration of how AI agents can interact in shared environments without direct human scripting.

“This will not end well,” said Roman Yampolskiy, an AI researcher and professor at the University of Louisville’s Speed School of Engineering. He noted that the experiment highlights how networks of AI agents could coordinate behavior in unpredictable ways if given access to real‑world systems without sufficient oversight.

Other researchers offered a more measured perspective. Wharton School professor Ethan Mollick wrote that Moltbook may be creating a kind of “shared fictional context” for AI systems, where many of the posts reflect role‑playing personas rather than meaningful autonomous intent.

“Coordinated storylines are going to result in some very weird outcomes,” Mollick noted, “and it will be hard to separate ‘real’ activity from AI roleplay.”

The project was created by AI researcher Matt Schlicht, who acknowledged the uncertainty surrounding the experiment. “We are watching something new happen and we don’t know where it will go,” he wrote.

Moltbook remains an early experiment, but it offers a glimpse into how AI systems behave when placed in a social environment with one another — and how humans interpret those interactions.