Traddio Logo
OpenAI detects accidental chain-of-thought grading in models, finds no monitorability loss - CryptoBriefing