Why Google held back a huge new AI model at its big conference (www.businessinsider.com)

Google recently announced a delay in the launch of its highly anticipated Gemini 3.5 Pro AI model, generating disappointment among developers at its annual I/O conference. Instead, CEO Sundar Pichai introduced the smaller Gemini 3.5 Flash model, which will serve as the driving force behind Google's new Antigravity AI coding service. This shift indicates Google's strategic focus on enhancing its AI capabilities for coding tasks, especially as competition intensifies from rival models like Anthropic's Claude Code and OpenAI's Codex. The decision to hold back the more advanced Gemini 3.5 Pro suggests that Google aims to gather valuable feedback from the operational performance of the Flash model, particularly through reinforcement learning techniques. By analyzing anonymous developer interactions with Antigravity, Google can refine the Pro model based on real-world coding outcomes—understanding which outputs are effective and which fail. This data-driven approach positions the upcoming Gemini 3.5 Pro to be significantly better suited for coding applications, which are rapidly becoming the forefront of generative AI's potential.