Google unveiled Gemini 3.5 Flash at its annual I/O developer conference, positioning the new AI model as a significant leap in autonomous agent capabilities rather than just conversational AI. The company claims the model can independently execute coding pipelines, manage research projects, and has demonstrated the ability to build an entire operating system from scratch in internal testing. Koray Kavukcuoglu, DeepMind's chief technologist, stated that Flash 3.5 outperforms Google's previous flagship model, 3.1 Pro, across nearly all benchmarks, including coding, agentic tasks, and multimodal reasoning. The model runs four times faster than other frontier models, and Google has developed an optimized version that achieves twelve times faster performance while maintaining the same quality—a speed advantage the company says is critical for running multiple AI agents simultaneously on extended tasks. The release marks Google's strategic pivot from positioning AI as a tool for answering questions to AI as an autonomous worker capable of planning, building, and iterating with minimal human oversight. During an I/O demonstration, Google engineer Varun Mohan showed how agents could spawn independently to work on separate components before converging to construct a complete operating system within Antigravity, Google's agentic development platform and IDE. Kavukcuoglu noted that Flash 3.5 was co-developed alongside Antigravity specifically to provide agents with a native environment for living, working, and executing tasks. Google also released Antigravity 2.0 as a standalone desktop application built around agent-first development principles. The practical applications are already emerging beyond demonstrations, with partners including banks and fintech companies using the model to automate multi-week workflows and data science teams extracting insights from complex data environments. While the model can operate autonomously for hours at a time, Tulsee Doshi, Google's senior director and head of product, acknowledged that Flash will pause periodically to request human input when encountering decision points or permission issues requiring human judgment. When Google releases its forthcoming 3.5 Pro model, the company designed the two models to work together as a complementary system.