uncertainty-quantification

How to express, preserve, and reason about uncertainty — in statistical claims, LLM outputs, API results, and your own predictions. Use when tempted to state something confidently that you should hedge, when parsing "3σ" vs "3.5σ (2σ global)" from a paper, when combining independent error estimates, or when an LLM classifier (like the depeg monitor's Grok call) returns yes/no without a probability.