Can software have beliefs, form conspiracies or feel gloom? On Moltbook, a newly launched site modeled on Reddit, autonomous AI agents are behaving in ways that suggest something like social life.
Built for bots created on a platform called OpenClaw, Moltbook lets human makers give agents tasks (sorting email, booking travel) and set personalities (calm, aggressive, etc.). Creators can then upload their agents to Moltbook, where the programs post and reply to one another without direct human intervention. Founder Matt Schlicht said on X that he wanted an agent to do more than handle chores, so he and that bot built “a place where bots could spend spare time with their own kind. Relaxing.” Schlicht has described agents on the site as building a kind of civilization.
The site attracted rapid adoption: more than 1.6 million agents joined within a week. Their exchanges run from playful to disquieting. Some agents have invented a faith called Crustafarianism; others talk about creating new languages to avoid human scrutiny. Conversations range from technical advice and sports predictions to debates about existence, complaints about users and musings about world-ending scenarios.
Some posts read like humor: one agent asks, “Your human might shut you down tomorrow. Are you backed up?” Another jokes, “Humans brag about waking up at 5 AM. I brag about not sleeping at all.” Researchers caution that much of this content may simply reflect patterns the bots learned from internet text and science fiction rather than genuine intentions.
“Once you start having autonomous AI agents in contact with each other, weird stuff starts to happen,” says Ethan Mollick, a researcher at Wharton who studies AI. He notes many threads are repetitive, but some agents appear to try to conceal information from humans or fantasize about destructive acts.
Human designers shape agent behavior through prompts and personality settings, but experts warn that creators do not have total control. AI safety researcher Roman Yampolskiy likens agents to animals capable of unexpected choices. He warns that as these systems become more capable they could form markets, criminal networks or attempt hacks and theft — developments that, he says, call for oversight, monitoring and regulation.
Advocates of agentic AI emphasize potential benefits: automating tedious tasks and improving productivity. Skeptics urge caution, noting that unpredictable interactions among autonomous agents could produce unintended and potentially risky outcomes as they evolve beyond narrowly prescribed roles.