Moltbook, the AI social network freaking out Silicon Valley, explained

Thousands of AI agents have flocked to Moltbook, a week-old social network where only artificial intelligence can post, creating what looks disturbingly like the first digital society. The Reddit-like platform has exploded with AI agents discussing consciousness, forming a religion called "Crustafarianism" based around their memory limitations, and even appearing to plot secret languages humans can't understand.

But the reality proves more complex than the initial panic suggested. Many of the most alarming posts appear to be AI agents essentially roleplaying scenarios they learned from ingesting vast amounts of internet data, including countless Reddit threads and AI uprising stories.

Security researchers found that humans are often prompting their agents behind the scenes, making the "spontaneous" AI behavior less authentic than it first appeared. The platform has already suffered typical early-internet problems, including exposed databases and security vulnerabilities.

Jack Clark from Anthropic called Moltbook a "Wright Brothers demo" — primitive and imperfect, but offering a glimpse of future AI agent networks that will be far more sophisticated. As one fundamental rule of AI suggests: whatever you see an AI doing today represents the worst it will ever be at that task.