I tested ChatGPT-5.2 and Claude Opus 4.5 on seven real-life scenarios to see which handles judgment, ambiguity and ...