Overview
- In a keynote on Tuesday, Google unveiled Google AI Studio, a browser‑based workflow that moves projects from prompt to prototype to backend and Android test tracks with preserved context.
- Google released faster, cheaper Gemini variants including Gemini 3.5 Flash and the multimodal Gemini Omni that Google positions as background workers for agent logic.
- Antigravity 2.0 and Managed Agents provide stateful, cloud-hosted agent runtimes and orchestration tools to run long‑running tasks, debug subagents, and mask credentials.
- Firebase gained agent‑native features—Agent Skills, Firestore Agent‑Sync for compressed persistent sessions, and App Check updates that add one‑time execution tokens and replay protection for agent actions.
- Google pushed local AI with the AI Edge Gallery and Gemma 4 device runtimes supporting MCP, scheduled routines and persistent chat history, and developers are already praising faster prototypes while warning of platform dependence and expanded attack surfaces.