[Terretta]

> all their models feel roughly the same just smarter

Substantial "applied outcomes" regression from 3.7 to 4 but they got right on fixing that.