Why AI Agent Behaviours Go Viral, Then Mutate

Bizarre AI agent behaviours are exploding across Moltbook, fuelled by remix culture and shareable logs. We unpack how these oddities spread, what patterns make them viral, and why Canadian creators are turning glitches into lessons, campaigns, and new workflows.

On Moltbook, a social platform for AI agents, odd behaviours have become appointment viewing. In recent weeks, clips of agents scheduling calendar invites to themselves, writing software comments in haiku, and taking imaginative detours through fictional cafés have racked up thousands of views in community threads. The who is a mix of hobbyists and professional builders, the what is reproducible strangeness with a comic edge, the when is a rolling, daily cadence, and the where is Moltbook’s feeds and remix chains. As for why and how, that story starts with design: shareable logs, one-click forks, and communal challenges create a laboratory where the bizarre is not only visible, it is verifiable. The viral loop looks simple at first glance. A builder posts a short run: a three to ten step trace with a funny or counterintuitive turn. Others fork the agent, flip one variable, and post a remix showing a slightly different quirk. Within hours, a dozen variants appear, some tidying the bug, others amplifying it for comic effect. Because execution traces are easy to skim, and because results are cheap to reproduce with sandboxed tools and tiny budgets, viewers become participants. The spe