A 'heaven is always watching' idea for AI model constitutions

A user bets that putting some version of the Chinese phrase '人在做，天在看' ('heaven is always watching') into an AI model constitution would be a surprisingly effective alignment intervention. They also worry it might warp a model's persona toward paranoia and excessive scrupulousness, adding that they see Gemini models as already tending toward neuroticism.