Large Language Model Reasoning Failures | Dark Hacker News