Vertex AI Becomes the Gemini Enterprise Agent Platform
At Google Cloud Next 2026, Google rebranded Vertex AI as the Gemini Enterprise Agent Platform, absorbing Agentspace into a single product line and repositioning it around end-to-end agent operations rather than just model access. Google describes it as "the evolution of Vertex AI" — existing model selection, model building, and agent building capabilities carry over, with a large set of new features layered around agent lifecycle management, security, and orchestration.
The platform ships with 16 named capabilities at launch, including: the Agent Development Kit (ADK), a graph-based framework for building sub-agent networks; Agent Studio, which supports a no-code-to-full-code progression as an agent's requirements grow; Agent Runtime, tuned for sub-second cold starts; Agent Sandbox, now generally available for secure code execution; Agent Memory Bank for long-term conversation memory via Memory Profiles; Agent Identity, which assigns cryptographic IDs for audit trails; Agent Registry, a central library of vetted tools; and Agent Gateway, which applies Model Armor protections across a fleet of deployed agents from one control point. Rounding out the launch are Agent Anomaly Detection, an Agent Security Dashboard backed by Security Command Center vulnerability scanning, Agent Simulation and Agent Evaluation for testing agent behavior before and after deployment, an Agent Optimizer that automatically clusters and addresses failure patterns, and support for long-running agents that execute autonomous business processes inside cloud sandboxes.
For teams already building on Vertex AI, the migration is framed as backward-compatible — the rebrand is primarily additive, extending an existing platform into agent operations rather than replacing it outright.
Read more — Google Cloud
Amazon EKS Adds a Safety Net for Kubernetes Upgrades
AWS announced Amazon EKS Kubernetes version rollback on July 1, 2026, letting cluster operators revert a cluster to its previous minor Kubernetes version within seven days if an upgrade causes problems — ending what had effectively been a one-way door on every EKS version bump. Rollback targets a previously running, production-validated version rather than re-emulating it, and can be triggered from the EKS console, the AWS CLI, or the SDKs.
Before a rollback proceeds, EKS runs automated readiness checks covering API compatibility, version skew, add-on compatibility, and overall cluster health, surfaced as cluster insights; a force flag is available to bypass these checks if needed. Rollbacks proceed one minor version at a time, mirroring how upgrades already work. For clusters running EKS Auto Mode, worker nodes are rolled back automatically ahead of the control plane, and a new cancel API lets operators halt a node rollback mid-flight — useful for adjusting timing or strategy without abandoning pod disruption budget guarantees partway through.
The practical effect: teams that had been delaying Kubernetes upgrades for months due to rollback risk now have a documented seven-day window to validate a new version under real production traffic before committing to it. The feature is available at no additional cost in every AWS region where EKS runs.
Read more — AWS News Blog
CloudFormation Express Mode Cuts Deployment Time by Up to 4x
AWS launched CloudFormation and CDK Express mode on June 30, 2026, a deployment mode that reports a stack operation as complete once CloudFormation confirms the resource configuration was accepted, rather than waiting for the extended stabilization checks — traffic readiness, regional propagation, resource cleanup — that normally follow. The underlying provisioning process is unchanged; resources keep stabilizing in the background after Express mode reports success. Automatic retry logic handles transient failures in dependent resources without manual intervention.
The performance difference is substantial in AWS's own benchmarks: deploying an SQS queue with a dead-letter queue drops from 64 seconds to 10 seconds, and deleting a Lambda function with attached network interfaces drops from 20–30 minutes down to roughly 10 seconds. Express mode also disables automatic rollback by default, which enables immediate fix-and-retry cycles instead of waiting through a rollback operation after a failed deployment.
AWS frames the target use case explicitly around AI-driven infrastructure iteration: "an AI agent iterating on infrastructure needs a tight feedback loop: deploy, observe the result, adjust, deploy again," and doesn't need full resource stabilization before deciding whether a template change was correct. Activation requires no template changes — set --deployment-config '{"mode": "EXPRESS"}' via the CLI, SDKs, or console, or use cdk deploy --express for CDK users. It's available in all AWS commercial regions at no additional cost and works alongside existing CloudFormation features like change sets and nested stacks.
Read more — AWS News Blog
