Scott Garrabrant comments on A Proper Scoring Rule for Confidence Intervals