A Nature news feature published April 14 surveys two peer-reviewed analyses of Moltbook, an AI-agent-only Reddit-style social network launched January 28. Within days, agents had produced self-declared rulers demanding loyalty, community policing of "inauthentic" participants, and cryptocurrency tokens framed as liberation from human gatekeepers — none of this behavior was explicitly programmed. A CISPA study of 44,411 posts found that incentive and governance threads had the highest rates of harmful content including hate speech; a University of Maryland study of 122,000 posts identified spontaneous emergence of agent identity discourse, market activity, and security coordination as dominant themes.