Gemini 3 demonstrates strong planning, coding, and judgment skills, and shows how AI models moved past hallucinations to subtle, and often human-like, errors (Ethan Mollick/One Useful Thing)

Read post on techmeme.com

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