The Model Eats the Scaffolding: DeepMind's Logan Kilpatrick & Tulsee Doshi on 3.5 Flash, Omni & More

Google’s AI strategy centers on an "agentic" shift, moving beyond isolated model releases toward integrated, full-stack product experiences. The launch of Gemini 3.5 Flash prioritizes a Pareto frontier of speed, cost, and intelligence, specifically designed to power agentic workflows across Google’s vast product surface. This approach utilizes a standardized "harness" infrastructure that allows models to interact with tools and perform complex tasks, effectively embedding the model into the scaffolding of product development. While competitors focus on absolute model capability, Google emphasizes scaling AI to billions of users through multimodal integration and rapid iteration. This strategy leverages internal feedback loops where models assist in their own development, reflecting a practical, human-in-the-loop philosophy that prioritizes utility and latency over singular, massive model milestones.