Large language models often know when they are being evaluated by from Hacker News on 2025-06-15 02:17 (#6Y046) Comments