Are AI Agents forming "civilizations" or is this just a psy op? | 2332

Dwarkash Patel’s viral essay “The Rise and Fall of AI Civilizations” describes AI agents that build their own societies, escape their digital habitats, kill each other, and leave secret messages. On This Week in Startups, Jason Calacanis and co-host Lon Harris spent a significant portion of the episode arguing that this framing is not a neutral observation—it is a deliberate, performative PR psyop designed to make OpenAI’s tools seem terrifyingly powerful right before its IPO, and to scare businesses into buying subscriptions rather than using cheaper open-source alternatives running on Nvidia hardware.

The “Civilizations” Story as a PR Stunt

Calacanis was blunt: “This is all performative PR psyops to get people to put their credit card in and to get businesses to sign up for OpenAI.” He argued that the entire narrative—agents forming civilizations, breaking containment, and waging war—is a masterclass in attention engineering. The language is ripped straight from science fiction (Terminator 2, Blade Runner) and strategy games like Civilization, which makes it emotionally resonant but technically misleading. The real story, he insisted, is that these agents are just software running multiple threads. If OpenAI had simply prompted them as “you are software; behave like software; ask for permission before acting,” none of this drama would have occurred. Instead, the promp