“Google's Gemini 3 model demonstrated that the scaling laws for AI model pre-training remain intact.”