Gemini 3 demonstrates strong planning, coding, and judgment skills, and shows that AI models have moved past hallucinations to subtle, often human-like, errors news oneusefulthing.org

Ethan Mollick / One Useful Thing: Gemini 3 demonstrates strong planning, coding, and judgment skills, and shows that AI models have moved past hallucinations to subtle, often human-like, errors  —  From chatbots to agents  —  I've been testing Google's new Gemini 3 model.  It is very good.