“The release of Google's Gemini 3 model provided an important confirmation that the scaling laws for AI model pre-training remain intact.”