Sunday, July 05, 2026

A field report details the recovery of core PostgreSQL tables after ransomware encrypted all database files and rendered system catalogs unusable. With only test-environment DDL available, the team adapted the PDU dropscan utility to match individual table files against known structures. This process allowed for the extraction of critical data directly from the raw file fragments.

• System catalogs being unusable requires bypassing standard recovery tools.
• PDU dropscan can be adapted to parse raw table files by structure.
• Test-environment DDL is essential for mapping raw data to tables.
• Full-file encryption necessitates low-level file scanning techniques.

Amazon RDS for Db2 now allows direct joining of database instances to self-managed Microsoft Active Directory domains hosted on-premises or in other clouds. This update enables Kerberos-based single sign-on without requiring AWS Managed Microsoft AD or establishing directory trusts. The change eliminates the overhead of managing a separate managed directory service for authentication.

• Join RDS for Db2 instances directly to on-prem or cross-cloud self-managed AD domains
• Use Kerberos for SSO without deploying AWS Managed Microsoft AD
• Remove the need for directory trusts between AWS and external AD forests
• Simplify authentication architecture for existing Db2 workloads on RDS

AWS has made Anthropic's Claude Sonnet 5 available, positioning it as the most capable model in the Sonnet generation. The release maintains the Sonnet tier's balance of speed and cost while delivering top-tier intelligence for complex tasks. It is optimized for coding across large codebases, agentic tool use with state retention, and professional knowledge work. This update aims to reduce correction rounds and improve first-time success rates for automated workflows.

• Sonnet 5 handles multi-file coding changes and refactoring with fewer correction iterations.
• Agentic workflows benefit from precise tool calling and better error recovery across steps.
• Performance scales for large codebases while keeping costs aligned with previous Sonnet models.
• Ideal for production workloads requiring high reliability in autonomous agent runs.

Amazon SageMaker Inference now supports automatic container image caching to accelerate scale-out events for generative AI models. By pre-caching large images (often 10 GB or more) from Amazon ECR, the service eliminates the cold-start latency previously caused by pulling full images for each new instance. This optimization allows new instances to start serving traffic significantly faster, reducing end-to-end scaling time by up to 50%.

• Generative AI endpoints scale up to 2x faster during scale-out events due to reduced cold-start latency.
• Large container images are pre-pulled from ECR, preventing new instances from waiting for full downloads.
• End-to-end scaling time is cut by up to half, improving responsiveness for traffic spikes.
• No code changes required; the caching feature is enabled automatically for supported SageMaker Inference endpoints.

Cloudflare has introduced Town Lake, an internal unified data platform leveraging a lakehouse architecture built on Trino, Iceberg, R2, and DataHub. The system integrates operational, billing, security, and business data, enabling governed cross-system analytics. Notably, billing workloads dominate usage, accounting for approximately 53% of the ~91K queries processed.

• Billing queries constitute 53% of total volume, highlighting their critical role in data infrastructure planning.
• Town Lake unifies disparate data sources (ops, billing, security, business) into a single governed lakehouse.
• The stack relies on open standards: Trino for query execution, Iceberg for table format, and R2 for storage.
• Skipper acts as an AI analytics agent, providing natural language access to the unified data layer.

AWS has increased the default runtime quota limits for Amazon Bedrock AgentCore to support larger-scale agent deployments. Concurrent session caps now reach 5,000 in US East and West, while all regions support 200 interactions per second and 25 new sessions per second. These adjustments allow developers to handle high-throughput workloads without immediately requesting quota increases.

• US East and West regions now support 5,000 active concurrent sessions by default.
• All regions support 200 agent interactions per second and 25 new sessions per second.
• Higher default limits reduce friction for scaling high-throughput AI agent workloads.
• Monitor actual usage against these new caps to avoid unexpected throttling.

Amazon WorkSpaces for AI agents is generally available, allowing agents to interact with desktop applications via screen-level automation. This enables enterprises to automate workflows on critical, hard-to-modernize systems like ERPs and mainframes without building custom integrations. The service leverages existing WorkSpaces infrastructure to provide a secure, managed environment for these automated operations.

• Agents operate desktop apps via screen interaction, bypassing the need for API modernization.
• Ideal for automating legacy ERPs, CRMs, and mainframes with strict compliance needs.
• Uses established WorkSpaces infrastructure for secure, scalable agent deployment.
• Eliminates costs associated with custom integrations for proprietary tools.

The .NET team has released dotnet/skills, a repository of curated plugins designed to assist AI coding agents. These skills cover C# language server integration, advanced scenarios, data access via Entity Framework, and performance diagnostics. The project includes a dashboard tracking accuracy and efficiency metrics for the contained plugins.

• Provides standardized .NET skills for AI coding agents to improve code generation accuracy.
• Covers critical areas including LSP integration, data access, and performance diagnostics.
• Offers a public dashboard to monitor plugin accuracy and efficiency trends over time.
• Aligns with the agentskills.io standard for consistent agent-skill interoperability.

AWS has extended Amazon Bedrock AgentCore availability to four new regions: Asia Pacific (Bangkok and Malaysia) and Europe (Milan and Spain). This service enables engineers to build, connect, and optimize AI agents using any framework or model while enforcing infrastructure-level security. The expansion allows customers to deploy agents closer to end users to reduce latency, with full capabilities including runtime, identity management, and observability available at launch.

• Bedrock AgentCore is now live in Asia Pacific (Bangkok/Malaysia) and Europe (Milan/Spain).
• Deploy agents closer to users in these regions to achieve lower latency.
• Security policies are enforced at the infrastructure layer and cannot be bypassed by agents.
• All core features like runtime, tool connectivity, and session persistence are available at launch.

OpenAI has published a plugin enabling Claude Code users to invoke Codex for code reviews and task delegation. The integration supports read-only reviews, adversarial challenges, and background job management via slash commands. It requires a ChatGPT subscription or API key and runs on Node.js 18.18 or later.

• Enables direct Codex invocation from within the Claude Code workflow
• Supports read-only reviews and steerable adversarial challenge modes
• Provides commands for delegating work, transferring sessions, and managing jobs
• Requires ChatGPT subscription or OpenAI API key with usage limits

Amazon SageMaker Unified Studio now allows provisioning via the open-source terraform-aws-sagemaker-unified-studio module. Platform teams can integrate SageMaker domains into existing infrastructure-as-code pipelines for version-controlled deployments. This enables consistent state management across development, staging, and production accounts using familiar IaC workflows.

• Use the official Terraform module to provision SageMaker Unified Studio domains.
• Integrate SageMaker into existing IaC pipelines for consistent deployment.
• Enforce version control for SageMaker domain configurations and access policies.
• Maintain environment parity across dev, staging, and prod AWS accounts.

Amazon EKS now allows users to revert to the previous minor Kubernetes version within seven days of an upgrade. The service performs automated readiness checks covering API compatibility, version skew, and add-on status before initiating the rollback. For EKS Auto Mode clusters, the system automatically rolls back worker nodes prior to reverting the control plane.

• Revert to prior K8s minor version within 7 days of upgrade to mitigate production issues.
• EKS validates rollback readiness via automated checks on API, skew, and add-ons.
• EKS Auto Mode handles worker node rollback before control plane reversion automatically.
• Rollbacks can be triggered via Console, AWS CLI, or AWS SDKs.

AWS CloudFormation now performs pre-deployment validation on Create Stack and Update Stack operations, catching common errors before resource provisioning begins. This feature, previously available only during change set creation, now runs automatically to detect property syntax errors, resource name conflicts, and S3 bucket constraints. The change eliminates the need to wait through full provision-and-rollback cycles for preventable failures, accelerating feedback loops in CI/CD pipelines and manual workflows.

• Pre-deployment validation now runs automatically on Create and Update Stack operations.
• Catches syntax errors, name conflicts, and S3 constraints before provisioning starts.
• Eliminates wait time for full provision-and-rollback cycles on simple errors.
• Accelerates feedback loops for manual, CI/CD, and AI agent infrastructure workflows.

AWS has introduced Express Mode for CloudFormation and CDK, reducing infrastructure deployment times by up to 4x according to internal benchmarks. This new mode completes stack operations as soon as CloudFormation confirms resource configuration is applied, bypassing extended stabilization checks like traffic readiness and region propagation. The change targets faster iteration cycles for developers and AI agents building infrastructure incrementally in development environments.

• Deployments skip full stabilization checks (traffic, propagation, cleanup) to complete faster.
• Up to 4x speedup for CloudFormation and CDK stack operations based on internal data.
• Enables rapid iteration for dev environments and AI agents building infrastructure.
• Resource config confirmation is the new completion trigger instead of full stabilization.

This GitHub repository provides a collection of agent skills designed for practical software engineering rather than experimental coding. The author argues that existing frameworks often remove developer control and complicate bug resolution, whereas these skills are small, adaptable, and model-agnostic. Built on extensive engineering experience, they aim to give practitioners granular control over their AI-assisted workflows.

• Avoids rigid process frameworks to maintain developer control over AI interactions
• Skills are composable and work with any language model for flexibility
• Designed for real application development, not just high-level vibe coding
• Easy to install via npx and customize to fit specific engineering needs

Amazon RDS now supports replicating automated backups across regions in Mexico Central, Asia Pacific Taipei, New Zealand, and Thailand. This feature enables point-in-time recovery for mission-critical databases by storing backup copies in geographically distant locations. Specific replication paths are enabled, such as Mexico Central to Europe (Ireland) or US West (N. California), and Asia Pacific Taipei to Singapore or Tokyo.

• Mexico Central RDS backups can now replicate to Europe (Ireland) or US West (N. California)
• Asia Pacific (Taipei) backups replicate to Singapore or Tokyo for improved disaster recovery
• New Zealand and Thailand RDS instances can replicate backups to nearby APAC regions like Singapore
• Enhanced geographic redundancy for mission-critical databases without manual backup management
• Simplifies compliance with data residency and disaster recovery requirements in these regions

AWS Security Hub CSPM has introduced an AI Security Best Practices standard featuring 31 automated controls. This new standard enables continuous evaluation of Amazon Bedrock, Bedrock AgentCore, and SageMaker workloads against recommended security configurations. It eliminates the need for manual assessments or custom rule authoring by checking critical domains like network isolation, encryption, and VPC placement.

• Covers 31 automated controls for AI workloads without custom rule authoring.
• Validates configurations for Bedrock, Bedrock AgentCore, and SageMaker.
• Enforces controls on network isolation, encryption, and KMS key usage.
• Requires private container registries and strict authorization controls.
• Provides continuous compliance evaluation for deployed AI resources.

Amazon EC2 Auto Scaling now supports a 'reservations-then-balanced' distribution strategy that prioritizes launching instances into capacity reservations before spreading remaining capacity across Availability Zones. This update targets On-Demand Capacity Reservations, Capacity Blocks, and Interruptible Capacity Reservations to maximize utilization of pre-purchased compute. Users configure this by setting the capacity distribution strategy in the Auto Scaling group’s AZ configuration and specifying target reservations via ARN or ID.

• Prioritizes filling Capacity Reservations before falling back to balanced AZ distribution for better cost efficiency.
• Supports ODCRs, Capacity Blocks, and Interruptible Capacity Reservations within the same strategy.
• Configurable via Auto Scaling group AZ settings using Capacity Reservation Group ARNs or individual IDs.
• Maintains operational resilience of Auto Scaling while optimizing pre-purchased capacity utilization.