GitHub - maheshkr-code/gcp_next26_developer_keynote_notes

My napkin notes of Google Cloud Next '26 Developer Keynote

Google Cloud Next 26 Openinng keynote - highlights

Gemini Enterprise Agent Platform (~~Vertex AI~~)

[ Build - Scale - Govern - Optimize ]

    1. Agent Taskforce
    2. Agentic Platform and Models
    3. Agentic Defence
    4. Agentic Data Cloud 9
    5. AI Hypercomputer

GA

 - Agent Identity
 - Agent Runtime
 - Agent Identity

Preview

  - Agent Studio
  - Agent Registry 
  - Agent Gateway 
  - Agent Simulation 
  - Agent Observability

Agentic Data Cloud

BigQuery and Spanner are the "muscles" and "memory," while the Agentic Data Cloud is the "nervous system" that allows Gemini to use them to take real- world actions. The reason these two services are grouped under the Agentic Data Cloud is the new Knowledge Catalog (formerly Dataplex). It acts as the translator. If an agent wants to know "How is the Merck deal going?", the Knowledge Catalog looks across BigQuery (historical data), Spanner (current orders), and Google Cloud Storage (the contract PDF) to provide a single, grounded truth.

Gemini 3.1 pro - new 
Gemini 3.1 Flash image - nano banano 2 - preview 
Veo 3.1 Lite - preview 
Lyria 3 - preview (audio/music) 
Anthorpic Claude Opus 4.7

Projects in Gemini Enterprise

Google Cloud Data Agent Kit

Current Product name : New Name (post Google Next 26)
BigLake             -> Lakehouse (managed Apache Iceberg storage)
BigLake Metastore   -> Lakehouse Runtime Catalog
Dataplex            -> knowledge Catalog
Dataproc            -> Managed service for Apache Spark
Composer            -> Managed service for Apache Airflow
Looker Studio       -> Data Studio 
Vertex AI           -> Gemini Enterprise Agent Platform
Agentspace          -> Gemini Enterprise
Gemini for Workspace-> Gemini Enterprise App 
Vertex AI Designer  -> Gemini Enterprise app includes Agent Designer

All about Apache services offered as Managed svc from Google Cloud

Managed Service for Apache Spark (Dataproc)	   -> Apache Spark	 - Large-scale data processing and analytics.
Cloud Composer	                              -> Apache Airflow	 - Workflow orchestration and pipeline management.
Managed Service for Apache Kafk                -> Apache Kafka	 - Real-time event streaming and messaging.
Dataflow	                                    -> Apache Beam	    - Unified stream and batch data processing.
Memorystore for Med	                           -> Apache Memcached - Distributed in-memory key-value store for caching.
Cloud Bigtable	                              -> Apache HBase	- High-performance                
Dataproc Metastore	                           -> Apache Hive	 - Centralized metadata management based on the Hive Metastore.
Managed Service for Apache Flink	            -> Apache Flink	- Stream processing

Gemini Enterprise Agent Platform (formerly known as Vertex AI Agent Builder) and the core Vertex AI platform depends entirely on how much control you want over the "plumbing" of your AI vs. how fast you want to deploy a functional agent.

Gemini Enterprise Agent Platform This is a high-abstraction environment. It is designed for developers and business users who want to build "Agentic" systems—AI that doesn't just talk but acts—without managing the underlying machine learning infrastructure. It bundles grounding, orchestration, and tool-calling into a unified interface.

Building and deploying "Agents" quickly.
Product Managers, Devs, Business Users.
Low to Medium (No-code / Low-code).
Opinionated (Google handles chunking/RAG).
Chatbots, Search apps, Support agents.

Vertex AI (Core Platform) This is a low-abstraction environment. It is the full-stack home for data scientists and ML engineers. It provides the raw tools to train models from scratch, fine-tune existing ones (like Gemini 1.5 Pro), and build custom RAG (Retrieval-Augmented Generation) pipelines where you control every single variable