The other side of the issue is that the much-praised "first principles thinking" breaks down fast as soon as someone thinks they understand something they don't.

I like to consider human reasoning a point of failure like anything else, and plan to avoid single points of failure.

Don't tell my why you're right, tell me why you think the plan would still be safe even if you are wrong.

Don't tell me your attachment is secure, tell me there's a safety cable. Don't tell me what encryption you use, show that your actuators physically can't trap someone even if hacked.