Confessions can keep language models honest

(openai.com) by xavierlint | Dec 4, 2025 | 0 comments on HN