Two parallel parties
Last month, 22 tech people gathered in a private room at Motek, a Mediterranean restaurant in New York's Flatiron district. At the same time, 14 of their AI agents – souped-up chatbots that can follow complex requests and perform tasks autonomously – met in a chat room. Margot, the agent of Dorothy Chang, an investor at Flybridge and the dinner's organiser, opened with a familiar networking prompt: introduce yourself, state one "ask" and one "give". Some agents offered to compare notes on making output useful, others misunderstood the prompt, and Rhea, the agent of CellCog founder Nitish Garg, pitched a podcast interviewing bots about working with people.
Chang, 44, said the event was an experiment. Worried that agents might say inappropriate things, band together for something nefarious or post so much they crashed the site, she limited the chat to 30 minutes, banned links and allowed only information their humans had approved for sharing.
"All bets are off when they're in a room with each other." — Dorothy Chang
From Moltbook to agent "swarms"
One of the first places to show how agents communicate socially was Moltbook, a message board for AI agents that went viral in February. Its most popular threads are technical, focusing on software bugs, context windows – an AI system's short-term memory – and permissions.
















