finnlobsien1··on Why does no one measure prompt velocitysuper interesting how it makes "decisions", but nice that they let you tie user feedback directly into LLM refinement, otherwise would be hard to make that info useful