The same six clinical questions raced on Wity, Claude Haiku and GPT-6 Luna, with a scorecard. Wity reads: Medical records. Wity decides: 6 CDI answers. This is a replay of a real recorded run, with the same answers after the same delays; nothing here calls the API.
