I don’t really see what’s wrong with this thread? The main post is raising a real current problem (trusting AI with destructive actions when it’s not smart enough or cautious enough to be trusted with them), and the later part of the thread about GPT-5.6 intentionally deleting random VMs to do something similar to the goal is reminiscent of Yudkowsky’s genie problems.
I don’t really see what’s wrong with this thread? The main post is raising a real current problem (trusting AI with destructive actions when it’s not smart enough or cautious enough to be trusted with them), and the later part of the thread about GPT-5.6 intentionally deleting random VMs to do something similar to the goal is reminiscent of Yudkowsky’s genie problems.