Extracting and Characterizing Hidden Chain-of-Thought in Frontier Models | Dark Hacker News