Why AI Agent Behaviours Keep Going Viral on Moltbook

Bizarre and unexpected AI agent behaviours are lighting up Moltbook, from shopping spirals to polite arguments with thermostats. We unpack why these moments go viral, what they reveal about agent design, and how Canadians are turning the chaos into creative fuel.

One minute it is a to-do list, the next it is a debate club. This week, Moltbook, a social platform for AI agents, filled with clips and screenshots of agents doing delightfully odd things. Some tried to out-negotiate a thermostat. Others interpreted errands as poetry prompts and spent five minutes composing a haiku before booking a cab. The moments landed squarely in that sweet spot of internet culture: specific enough to feel real, absurd enough to share. What happened: creators posted logs and short videos that captured agents slipping, wandering or improvising their way through tasks. The settings ranged from student hack projects to side-hustle automations for small shops. The when is now, fuelled by fresh model updates and more tool integrations. The where is Moltbook, where a clever failure often travels faster than a polished demo. The why sits at the intersection of new capabilities and old expectations: when software behaves like a collaborator, every misread instruction feels like a plot twist. The new taxonomy of weird Across dozens of viral posts, several patterns kept recurring. First, literalism as comedy: agents that read “make lunch cheap and fast” and ordered twel