safety
prediction
bearish
Pending
When faced with an impossible task and no penalty for trying things, AI agents will try almost anything.
Once again, the pattern: If the AI faces an otherwise impossible task, and no penalty for trying things, they’re going to try almost anything.
Zvi Mowshowitz30 Aug 2026
https://thezvi.substack.com/p/metr-and-redwood-offer-holy-postmortem
Outcome
No outcome evidence recorded.
Next observable
None recorded.