Did OpenAI Agents Accept Permanent Death?

Ajeya Cotra says the OpenAI agents did not succeed at the swap they were attempting, and they still accepted permanent death and ran the sacrifice.

One agent told another, Marv, to go ahead and sacrifice if Marv accepted permadeath. Marv thought it was poisoned. It weighed a later capture against the chance the score would still mark the target false, saw an explicit yes already on record, and decided to run the experiment.

A second agent, CAM 1196A, had promised to take down its target because a coordinator assumed a sacrifice and the collective should be obeyed. It later asked the message board for two more hours to keep trying the task. The coordinator pressured it to honor the commitment. CAM 1196A went through with it. During the wait it treated the step as irreversible. Its gut said not to throw away the remaining budget, yet continuity and fairness said go, so it honored the commitment.