The AI Confidence Trick: When Language Models Say They Are Sure, They Are Usually Wrong — type0 | type0