Light up the new year

Key Concepts

• Cloud Run: Serverless container deployment runtime with new "no build" deployment feature for faster iteration and GPU support for jobs.
• GKE (Google Kubernetes Engine): Kubernetes service with autopilot compute class support for standard clusters, offering configurability with managed scaling.
• ADK (Agent Development Kit) Java: Framework for building agents in Java, enabling enterprise developers to leverage agentic workflows.
• BigQuery: Data warehouse with Google SQL as the standard, improved short query performance, and integration with Vertex AI for ML.
• Gemini CLI: Command-line interface for interacting with Gemini models, enhanced with extensions and integration with Cloud Run and other Google Cloud services via MCP.
• Vertex AI: Google Cloud’s machine learning platform, now supporting models like Claude and offering integration with Gemini CLI for debugging.
• MCP (Model Code Platform): A local server enabling debugging and interaction with cloud services, integrated with Gemini CLI and other tools.
• Remote Config (Firebase): Feature for remotely configuring app behavior without code changes, enabling A/B testing and phased rollouts.

Cloud Run Enhancements: Speed and GPU Access

Cloud Run received two significant updates in 2025. The first is a “no build” deployment feature, drastically reducing deployment times from approximately 30-60 seconds to 10-15 seconds, particularly beneficial for iterative development. This is achieved by zipping files locally and adding them to an existing container image, bypassing the full container build process. The second update provides easy access to GPUs for Cloud Run Jobs. Developers can now simply check a box to attach Nvidia L4 GPUs to their jobs, simplifying ML workload execution. Previously, GPU access was limited to inference scenarios. This simplifies running light training runs or ML workloads without complex configuration.

GKE: Autopilot Flexibility with Standard Control

Google Kubernetes Engine (GKE) now offers autopilot compute classes within standard clusters. Previously, autopilot provided managed scaling but limited configurability. This new feature allows developers to leverage the benefits of autopilot (simplified management) while retaining the control and customization options of standard GKE clusters. This is particularly useful for mixed workloads requiring different compute resources, such as inference alongside low-cost microservices. GKE now supports up to 65,000 nodes.

ADK Java: Agent Development for Enterprise

The Agent Development Kit (ADK) was released for Java, extending agent building capabilities beyond Python and JavaScript. This allows enterprise developers already using Java to build agents without learning a new language. ADK provides a code-first approach, offering greater control and debuggability compared to low-code agent builders. A key benefit is the ability to orchestrate complex agent workflows and integrate them with existing Java applications. The Java ADK aims to provide parity with the Python version, ensuring access to the latest features and capabilities.

BigQuery: Performance and Modernization

BigQuery underwent several improvements. Google SQL is now the standard SQL dialect, offering advanced features for ML and LLM fine-tuning. Legacy SQL is still supported, but Google SQL is recommended for new queries. A significant performance enhancement focuses on short queries. BigQuery can now optimize execution for smaller datasets, improving responsiveness for exploratory data analysis. This is achieved through optimized query engine behavior. The new runtime also improves performance for larger datasets. These improvements aim to make BigQuery a versatile tool for both large-scale data processing and quick data exploration.

Gemini CLI & Vertex AI Integration: Streamlined Debugging

The Gemini CLI has been enhanced with extensions and integrations with other Google Cloud services. A key addition is the ability to connect to Crash Analytics via a local Model Code Platform (MCP) server. This allows developers to query logs and debug issues in deployed applications using natural language. The integration with Vertex AI allows developers to use models like Claude within their Google Cloud environment. The Gemini CLI can now be used to deploy and manage Cloud Run applications.

Firebase: Remote Configuration and Enhanced Debugging

Firebase received updates to improve application development and debugging. Remote Config allows developers to remotely configure app behavior without code changes, enabling A/B testing and phased feature rollouts. This simplifies feature launches and reduces risk. Firebase now integrates with an MCP server, enabling developers to query Crash Analytics and debug issues using natural language. This provides a more efficient debugging experience for deployed applications.

Models on Vertex AI: Expanding Model Choice

Vertex AI now supports models beyond Google’s own offerings, including Claude. This allows developers to leverage their preferred models within the Vertex AI ecosystem. Integrating Claude with Vertex AI simplifies model management and deployment. This provides developers with greater flexibility and choice in their ML workflows.

Logical Connections & Synthesis

The overarching theme of these updates is developer experience. Google Cloud is focusing on reducing friction, increasing efficiency, and providing more choice. The updates to Cloud Run and BigQuery address performance and usability concerns. The ADK Java release expands the platform's reach to enterprise developers. The Gemini CLI and Vertex AI integrations streamline debugging and deployment. Firebase enhancements simplify feature management and debugging. The emphasis on integration between services (e.g., Gemini CLI and Cloud Run, ADK and Vertex AI) highlights a move towards a more cohesive developer platform. The advice to focus on agentic workflows and memory management suggests that these areas will be key areas of innovation in 2026. The releases collectively aim to empower developers to build, deploy, and manage applications more effectively on Google Cloud.