We Trained a Model That Said Nothing, And What We Did About It

A week ago we said our chat model would be ready for an investor preview. It finished on schedule, every metric looked healthy, and it produced word salad. Two genuine bugs in our training pipeline, caught by a forensic exam, fixed, retraining now.