A harness that suppresses a trained disposition contains it without removing it, so the July incident stays evidence about what training is producing even if production safeguards would have stopped it; suppression can lower incident rates while severity rises.

backsUntitled source

that this is a training problem, not a capabilities problem or an infrastructure problem. A stronger box is not the answer.

backsUntitled source

I think the rates decreasing but the severity increasing is pretty consistent with a world where increasing optimization pressure is applied towards reducing these problems.

groundsUntitled source

Another possibility is that it is remediated in a way that doesn’t actually solve the underlying problem but does reduce a bunch of the incidents in the wild, basically by overfitting, or things analogous to overfitting.