Can we do a "feelsort" using Jev's calibrated probabilities? Why pointwise scoring? Simpler, O(n) calls, fully parallel. Why threshold phrasings? Adds levels (a ladder) of intensity. Why "a randomly chosen person"? If calibrated, it estimates the population's average climb.

This doesn't just order items by how people might feel. Jev estimates how many people feel that way. If its calibration carries over from "is this right?" to "would a random person say yes?", 0.8 on a rung means ~80% of people would agree. That's untested but checkable via survey