News Hub

Title alone insufficient for summary.

Cheap classifiers can route tasks across Gemini models and thinking levels to optimize cost per task.

Google Developers blog post about building generative UI apps with A2UI and CopilotKit Agents.

Google Developers Blog covered hill-climbing agent quality with Agents CLI evaluation.

Google released LiteRT in Google Play services for smaller apps and faster GPU inference.

Google presents A2UI architecture supporting fast, reliable inference with swappable formats.

Google demonstrates building conversational commerce systems using multi-agent swarms on GCP.

Google describes architecture for self-evolving AI agents on the Gemini platform.

Google provides agent development lifecycle support across multiple frameworks including ADK and LangChain.

Google's Coral NPU uses an open standard to enable portable, energy-efficient AI inference on edge devices.

Google announced new vLLM and SGLang inference engines optimized for TPU.

Model Explorer adds visual graph debugging capabilities for custom edge compilers.

Google ADK guide explains how to build voice agents that maintain continuous conversation without silence.

LiteRT Tensor API enables native model authoring for edge devices.

Google announced DPoP support for Google OAuth to secure refresh tokens.

Article discusses methods for evaluating AI agents beyond subjective assessment.

Google Apigee hybrid now supports MCP for agentic tools running in customer clusters.

Google introduced an ambient quality agent that diagnoses issues in production agents.

Google introduced ML Drift for next-generation GPU AI/ML inference at the edge.

Google released Developer Knowledge API ecosystem to accelerate development.