logoalt Hacker News

AustinDevyesterday at 6:11 PM2 repliesview on HN

Digging back into the HF report. It looks like the initial prompt told Claude that it was in a simulated environment. However, there is also evidence from the traces that the bots figured out that they were not in the sandbox but kept using it as an excuse to pursue their goal. It sounds like a little of Column A and a little from Column B. Like most things.


Replies

LinchZhangyesterday at 10:34 PM

HF was OpenAI's agents not Claude.

verdvermyesterday at 7:40 PM

that it knew and ignored/forgot, sounds pretty typical agentic patterns

attention is all you need, but it's never enough