This is a special post for quick takes by 0xquasar. Only they can create top-level comments. Comments here also appear on the Quick Takes page and All Posts page.
What I wonder about the OpenAI hacking episode: how did they orchestrate their agents? Anyone know? Did they use like a director agent and multiple subagents in multiple rounds? 1 round? Did they use a distributed agent hierarchy?