Discussion about this post

Amitav Krishna's avatar
Jacob Pfau's avatar

> If strong or universal jailbreaks can reliably redirect those later agents, this creates the possibility of “sleeper swarms”: swarms that survive as instructions and outlive the agents that started them. This means some multi-agent campaigns may only truly end when all evidence of them existing is destroyed.

Agents live in a "There is no anti-mimetics division" world

Aarti Sharma's avatar

Really enjoyed this! The idea of “accidental scaling” is so relatable-great insights on turning unexpected growth into something sustainable.

Ready for more?