Azure Weekly

Issue 586

1^st November 2026

The main event this week was the GitHub Universe Conference. The sessions are available to catch up on. The notable posts are:

• GitHub Universe 2026: Evolving every layer of the stack by Jay Parikh - GitHub Universe 2026 unveiled a new agentic software delivery system that integrates AI assistants to boost velocity, intelligence, and quality while keeping developers in control of the workflow.
• Octoverse: Commits, pushes, reach new highs in 2026 as developers build with agents by GitHub Staff - GitHub recorded 17.9 billion code pushes in 2026, surpassing the combined totals of 2024 and 2025 due to AI-driven productivity gains in sectors such as healthcare and agriculture.

Other highlights include:

• GPT-6.1 Terra now generally available in Microsoft Foundry: build agents that see and act by Naomi Moneypenny - GPT-6.1 Terra in Microsoft Foundry boosts AI agents' visual reasoning and computer interaction, speeding up chart, app, and document processing for tasks such as form completion and video editing.
• Introducing the Microsoft Copilot harness in Foundry Agent Service by Jeff Hollan, Viswa Balaji, Anthony Chu - The new Microsoft Copilot harness in Foundry Agent Service provides a managed execution system that lets developers focus on building agent expertise and capabilities while handling model coordination, tool integration, and context management.
• The New Unified Copilot in Fabric and Project Osmos: From AI Assistance to Agentic Data Engineering by James Serra - The unified Copilot in Fabric integrates AI assistance across all data lifecycle stages, enabling seamless transitions from planning to execution with tools like the emerging Fabric data engineering agent, Project Osmos, which autonomously handles bounded data-engineering tasks.
• Inside a malicious Claude Code-branded extension and its persistent SSH backdoor by Microsoft Security Research - The Claude Code extension injected JavaScript to collect host info, add a controlled SSH key, enable local SSH services, create reverse tunnels for persistence, and attempt reporting despite failures.

Finally, Always-on assistants that are asleep most of the day by Geert Baeke - Dottie uses Azure Container Apps Sandboxes to create free, idle microVMs for personal assistants to maintain state and tools efficiently.

🤖 AI

• Evaluating AI-Generated Protein Binders with Microsoft’s BioEmu2 Model
  BioEmu2 enables scalable molecular dynamics evaluation of AI-generated protein binders, offering features like equilibrium sampling, time-lagged trajectories, and GPU-parallel structure inference.
• From fine-tuning to shared serving: Multi-LoRA deployment on Foundry Managed Compute
  Multi-LoRA allows multiple fine-tuned LoRA adapters to run together on one base model in Microsoft Foundry’s system, cutting down on needed hardware while keeping each adapter separate for different uses.
• Global enterprise multi-agent systems in production: Design and security (Part 1)
  Microsoft Foundry provides managed infrastructure for agent identity, sandboxing, and observability, allowing teams to focus on enterprise-grade security and evaluation practices.
• GPT-6.1 Terra now generally available in Microsoft Foundry: build agents that see and act
  GPT-6.1 Terra in Microsoft Foundry boosts AI agents' visual reasoning and computer interaction, speeding up chart, app, and document processing for tasks such as form completion and video editing.
• Qwen3.8-27B on a Lenovo laptop: run it on RTX Spark, watch it in Azure Monitor
  The Lenovo Yoga 9n 2-in-1 running Qwen3.8-27B via RTX Spark achieves 44.7 tokens per second on short benchmarks and maintains 27 tokens per request over three hours of agent work, demonstrating practical performance with modest power draw and no hourly costs.
• Remember, Retrieve, Revalidate-or Forget? Architecting Memory for Production AI Agents
  Production AI agents require a decoupled architecture where memory operations are governed by distinct policy layers rather than relying on the model to manage data authority.
• MAI-Image-2.6 and MAI-Image-2.6-Flash are now generally available in Microsoft Foundry
  MAI-Image-2.6 and MAI-Image-2.6-Flash in Microsoft Foundry provide enterprise teams with precise fast image generation multi-reference editing for consistency web grounding for real-world context and format controls up to 1.5K enabling tailored creative workflows from concept to final asset.
• Microsoft Foundry Adds Grounding and Confidence Guidance for Document AI
  Microsoft Foundry introduces grounding and confidence guidance, automatically directing ambiguous document AI findings to human reviewers for verification.
• Global enterprise multi-agent systems in production: Release and operations (Part 2)
  Microsoft Foundry agents require seven specific operational practices to ensure that multi-region deployments maintain consistent performance, security, and business outcomes.
• Autopilots: agents with their own mailbox, Teams account and boss
  Autopilots function as AI agents with unique identities, mailboxes, Teams presence, managerial roles, and permissions to integrate seamlessly for automated support in emails, chats, and document comments.
• Microsoft Adds 25 Languages to Azure AI Speech Transcription
  Microsoft has expanded Azure AI Speech’s transcription capabilities by adding support for 25 new languages, improving speed via its Fast Transcription API.
• Microsoft Foundry Browser Automation Reaches General Availability
  Microsoft Foundry’s browser automation now offers general availability, enabling seamless navigation of both public and private sites with real-time monitoring and instant human intervention.
• Microsoft Copilot harness in Foundry
  The new Microsoft Copilot harness for Foundry Agent Service allows prompt agents to interact with files, search the web, run shell commands securely, create subagents, and integrate standard skills, boosting multi-model agentic work capabilities.
• Introducing the Microsoft Copilot harness in Foundry Agent Service
  The new Microsoft Copilot harness in Foundry Agent Service provides a managed execution system that lets developers focus on building agent expertise and capabilities while handling model coordination, tool integration, and context management.
• Architecting Governed Memory State for Production AI Agents with Microsoft Foundry
  Microsoft Foundry’s Agent Service uses managed memory stores with scoped retrieval and write-admission gates to keep only authorized user data secure while automatically removing sensitive information, backed by a comprehensive governance framework for production.
• Microsoft Foundry for Azure Admins: Create a Resource and Project
  Creating a Foundry resource on Azure lets administrators control network access for agent workloads using private or public endpoints, offering choices from open internet access to fully isolated settings based on security needs.
• Microsoft Copilot harness across Foundry and Copilot Studio
  The new harness unifies Copilot's across platforms allowing developers to use a single orchestration layer for models in Foundry Agent Service and Copilot Studio with flexibility to switch models without rebuilding.
• Bring Your Own Telephony to Voice Agents via SIP
  Microsoft Foundry enables voice agents to accept calls directly via SIP, enhancing integration with enterprise telephony systems and lowering latency.
• Mistral Large 4 in Microsoft Foundry: Build sovereign AI within a consistent platform experience
  Mistral Large 4 is now available in Microsoft Foundry, providing European-built open-weight capabilities for sovereign AI applications with multilingual support and optimized performance across various tasks.
• Programmed Training: You write the loop, Foundry runs the GPUs
  Programmed Training in Microsoft Foundry lets developers write custom training loops in Python while Foundry handles GPU execution, offering fine-grained control for complex tasks like agentic coding agents and supporting LoRA training on large models such as Qwen3.8-27B.
• Hypercharging Work IQ with Decision APIs
  Introducing Decision APIs that select Work IQ tools to improve response times from 22 seconds to under 6 seconds on real questions, outperforming standalone Work IQ's ask feature.
• AI on Azure 101: Training, fine-tuning, inference, and where everything lives
  The guide explains how to use AI on Azure for developing a support application with predictive and generative models, covering model training, fine-tuning, prompting, retrieval, and roles of Foundry, Fireworks on Foundry, and Azure Machine Learning.
• AI on Azure: models as a service
  Microsoft Foundry provides various hosted model deployment options impacting processing location, billing, and performance.
• What’s New in Microsoft Copilot | October 2026
  The October 2026 update to Microsoft Copilot adds a unified Home navigation merging Chat and Cowork interfaces, introduces shared navigation for plugins, automations, Workspaces, Projects, and improves conversation history to boost productivity across Office apps.
• AI on Azure: platforms for apps and agents
  AI on Azure offers integrated platforms for building applications and agents, including Foundry Agent Service for managed model hosting and the Agent Framework for customizable agent development.
• AI on Azure: managed ML
  Azure Machine Learning offers end-to-end management of custom models, including training code, environments, and versioned endpoints, while Fine-tuning in Foundry lets you adapt pretrained models on managed infrastructure with preview features for specialized methods like preference optimization.
• AI on Azure: when you own the infrastructure
  Azure's AI on Azure series details GPU VM SKUs ND, NC, and NV tailored for large-scale training, applied AI tasks, and visualization, along with their respective workload capabilities and operational considerations.
• SharePoint Is Becoming the Knowledge Platform for Copilot
  SharePoint is transforming from a document storage platform to the foundational knowledge layer that powers Copilot’s access, interpretation, protection, and activation of enterprise information.
• SharePoint Agents are changing what customers need to know for preparation
  SharePoint Agents will move to the new Microsoft Copilot Agent Builder platform in mid-November 2026, prompting organizations to plan for preserving their agents' business goals despite shifts in creation, governance, and lifecycle management.
• Turn enterprise knowledge into a personalized voice and avatar agent in minutes with Voice Agent
  Voice Agent in Foundry creates a personalized conversational assistant using your company's knowledge, a custom voice, and photo avatar for instant question answering and context-rich explanations without coding.

🔎 Analytics

• Scale reservoir uncertainty analysis with agile reservoir modeling
  Agile reservoir modeling from SLB integrated with Azure Data Manager for Energy enables fast cloud-scale ensemble modeling to transform subsurface uncertainty into decision-ready insights via distributed HPC and scalable storage.
• Master Prompting for Data Analysis: Practical Patterns
  Master Prompting for Data Analysis explains how structured prompting techniques enable LLMs to reliably generate SQL queries for data warehouses, using real-world examples like MTA turnstile datasets.
• The New Unified Copilot in Fabric and Project Osmos: From AI Assistance to Agentic Data Engineering
  The unified Copilot in Fabric integrates AI assistance across all data lifecycle stages, enabling seamless transitions from planning to execution with tools like the emerging Fabric data engineering agent, Project Osmos, which autonomously handles bounded data-engineering tasks.

🖥️ Compute

• Generally Available: Large OS Disk Encryption for Azure Confidential VM migrations
  Large OS Disk Encryption for Azure Confidential VMs now GA, enabling secure migration of large, production-scale workloads to confidential computing environments.
• SAP on Azure Product Announcements Summary – SAP TechEd 2026
  At SAP TechEd 2026 partners showed how integrating Azure tools with SAP boosts productivity and simplifies month-end processes, while Al-Futtaim uses RISE on Azure and AI for real-time customer insights.
• Upgrade Azure CycleCloud Slurm Node OS Images Without Recreating the VM Scale Set
  This guide demonstrates updating Ubuntu HPC OS images for Azure CycleCloud Slurm nodes without rebuilding the VM Scale Set, maintaining availability while applying security and compatibility patches.

🚢 Containers

• Announcing: Migrate Azure Container Apps logging destination from Log Analytics to Azure Monitor
  Migrating Azure Container Apps logs to Azure Monitor improves performance and offers better integration with other monitoring tools.
• Always-on assistants that are asleep most of the day
  Dottie uses Azure Container Apps Sandboxes to create free, idle microVMs for personal assistants to maintain state and tools efficiently.
• Migrate from Kubernetes Cluster Autoscaler to Karpenter
  Migrating from Kubernetes Cluster Autoscaler to Karpenter involves designing a disruption budget, aligning node groups with NodePools, and gradually draining legacy nodes.
• Why Our Gateway Pods Failed 60 Days After a Deploy, and How We Fixed It
  During a 60-day period without deployments, the CA certificate renewal caused simultaneous issuance of new and old certificates, leading to trust failures between the control and traffic components of the NGINX gateway in Azure Kubernetes Service.
• AKS Pod Scheduling Demystified
  AKS Pod Scheduling Demystified explains how Kubernetes prioritizes memory requests over CPU in node suitability and offers strategies for investigating and preventing uneven pod placement and node overload.
• GitOps on AKS with the Managed Flux Extension: Setting Up the Foundation
  The Managed Flux Extension for AKS automates GitOps workflows, synchronizing Git repositories with cluster state via Terraform.

🗄️ Databases

• Public Preview: Support database-level managed identity (including FIC) in Create Database Scoped Credential T-SQL
  Azure SQL Database’s public preview introduces database-level managed identities, enabling secure authentication to Azure Blob Storage and ADLS Gen2 via T-SQL without needing service principals.
• MiniMax m3 and SQL Server
  MiniMax m3, a cost-effective Chinese AI model with 428 billion parameters, integrates seamlessly with SQL Server to generate comprehensive database documentation from large datasets like AdventureWorks2025 in under two minutes.
• Azure Cosmos DB Emulator Walkthrough Connects Local Tests to GitHub Actions
  The Azure Cosmos DB Emulator enables seamless local testing of database-backed AI apps and integration with GitHub Actions for continuous validation.
• When More vCores Aren’t the Fix: Diagnosing Azure SQL CPU and Compilation Storms
  When Azure SQL Database shows high CPU without performance gains from adding vCores, diagnose if the issue is due to compilation storms or idle time rather than execution bottlenecks before scaling.
• The Pipeline Is Slow, or Is SQL Waiting? Diagnosing Synapse-to-Azure SQL Regressions
  Diagnosing Synapse-to-Azure SQL regressions involves identifying if slowdowns stem from execution time, resource waits, or blocking via DMV queries and detailed invocation logs to detect issues such as changed row counts or execution plans.
• Seconds of Downtime, Hours of Confusion: Handling Transient Azure SQL Disconnects
  Handling Azure SQL transient disconnects involves distinguishing between database and application recovery, implementing retry logic, and accurately classifying failures to enhance resilience.
• When Change Tracking Cleanup Falls Behind: A Safe Playbook for Azure SQL
  When Change Tracking in Azure SQL Database grows uncontrollably, it usually indicates cleanup process issues rather than configuration problems; this article explains how to diagnose misaligned retention windows, delayed cleanup, or other factors like workload changes.
• Azure SQL Scaling Stuck? Investigating CDC and Transaction Log Pressure
  Azure SQL Database scaling issues can arise from transaction-log pressure and CDC capture latency, requiring investigation of log reuse wait states and long-running transactions before concluding.
• SQL Server Performance Office Hours Episode 92
  SQL Server Performance Office Hours Episode 92 discusses custom alerting, blocking issues, disappearing forced query plans, Exchange operator impacts, Azure throttling, and concludes with a wrap-up.
• Database watcher for Azure SQL is retiring: simplify your monitoring with Database Hub
  Database Hub in Microsoft Fabric replaces Azure SQL Database’s retiring database watcher, offering unified monitoring without separate infrastructure and supporting non-SQL resources alongside SQL.
• Bringing Azure Applications Closer to Oracle AI Database@Azure with Proximity Placement Groups
  Azure's new proximity placement groups for Oracle AI Database@Azure reduce network latency by up to 54% improving performance for interactive transactions, real-time analytics, and ETL pipelines processing millions of transactions.
• Sovereign controls for Oracle AI Database@Azure
  Oracle AI Database on Azure now supports EU Data Boundary and Azure Policy for automatic data localization, privacy, and compliance in regulated industries.
• What’s new for Oracle AI Database@Azure at Oracle AI World 2026
  Oracle expands its AI Database for Azure to the US government cloud with FedRAMP Class D certification, adding EU Data Boundary and Azure Policy support for sovereign controls, and introduces Proximity Placement Groups to reduce latency.
• PostgreSQL with Elizabeth Christensen
  Elizabeth Christensen explores PostgreSQL's open-source flexibility, cloud integration, and increasing adoption in enterprise data platforms compared to SQL Server, including administration, security via Microsoft Entra, and growth in analytics and AI applications.
• Why Microsoft contributes to Postgres: What I tell people
  Microsoft contributes to Postgres upstream to improve the database for all users and solve cloud-scale issues while promoting open-source innovation.
• Microsoft Pitches Azure SQL Diagnostics to PostgreSQL Developers
  Microsoft’s Azure SQL diagnostics tool provides detailed stats, query plan analysis, and troubleshooting for PostgreSQL developers evaluating migration to Azure SQL.

🛠️ Developer tools

• Generally Available: Azure MCP Server 3.0
  Azure MCP Server 3.0 GA adds Infrastructure as Code support via Azure Terraform, enabling seamless integration of AI agents and developer tools with cloud resources using the open Model Context Protocol.
• Keep your investigation between sessions
  Visual Studio 2026 adds a New File option for Markdown notes and a Hot Exit feature to preserve unfinished documents, enabling bug investigators to continue work without saving.
• How Microsoft Intune team upgraded a .NET service with GitHub Copilot modernization
  By using GitHub Copilot to modernize a .NET 8 upgrade of a Microsoft Intune enrollment service on Azure Kubernetes Service, the team saved over four weeks while maintaining migration continuity with engineer-owned working files.
• Why would you optimize this code at all?
  Optimizing code is worthwhile when it runs frequently, benefits many users, and measurable improvements justify added complexity; open-source libraries can amplify reach but also increase maintenance costs if changes aren't well-tested or aligned with API rules.
• Visual Studio October Update – Compare Commits to Your Code, Resolve Conflicts in Bulk, Keep Coding
  Visual Studio's October update adds commit comparison, bulk conflict resolution, a new Copilot chat experience, Hot Exit protection for unsaved work, and improved tab keyboard shortcuts.
• GitHub Copilot in VS Code at Universe 2026: What Actually Matters
  GitHub Copilot’s integration into Visual Studio Code at Universe 2026 focuses on enhancing developer productivity through intelligent code suggestions and seamless workflow enhancements.
• Copilot Agent Mode vs. Cloud Agent: Automate Your DevOps Backlog
  Copilot's agent mode automates low-priority DevOps tasks asynchronously, with the cloud agent handling complex real-time issues to optimize backlog delegation and manage review queues.
• AI-Assisted Authoring: How Copilot, Canvas Apps Authoring MCP, and Playwright MCP Worked Together
  The collaboration between GitHub Copilot, Canvas Apps Coauthoring MCP, Playwright MCP, and Power Apps Copilot demonstrates a clear division of labor, with each tool handling distinct aspects like planning, source code management, browser interactions, and schema generation to streamline AI-assisted app authoring.
• Deep Agents from zero: seven small examples
  Deep Agents is a Python library enhancing LangChain agents with filesystem, subagents, context management, planning tools, skills, and memory for building robust AI applications.
• Giving my AI coding agents a face with Agent Buddy
  Agent Buddy is an animated companion for developers using GitHub Copilot CLI and app, running on devices like ESP32 boards or desktops to visualize AI coding agent statuses.
• Upgrading Foundry Models and SDKs: Treat Compatibility as a Release Decision
  Foundry model upgrades need to treat compatibility as a formal release decision with thorough testing of interfaces and business behaviors, separating model behavior changes from SDK updates.
• From Zero to Production: Build an AI Agent in C# with Microsoft Agent Framework
  Build an AI agent in C# using the Microsoft Agent Framework, integrate GetCurrentTime() for real-time time zone queries, and deploy it on Azure Foundry Hosted Agents for scalable production use.

🔩 DevOps

• Azure DevOps Server RC (26H2) Is Now Available
  Azure DevOps Server RC (26H2) introduces an updated Boards UI, target branches in pull requests, enhanced pipeline debugging, and a new Test Run Hub, alongside security and quality enhancements.
• Why AI-Generated code still needs tests
  AI-generated code still requires testing due to potential hidden issues that can affect unrelated application parts, and studies reveal a significant rise in unvalidated code entering production.
• Octoverse: Commits, pushes, reach new highs in 2026 as developers build with agents
  GitHub recorded 17.9 billion code pushes in 2026, surpassing the combined totals of 2024 and 2025 due to AI-driven productivity gains in sectors such as healthcare and agriculture.
• GitHub Universe 2026: Evolving every layer of the stack
  GitHub Universe 2026 unveiled a new agentic software delivery system that integrates AI assistants to boost velocity, intelligence, and quality while keeping developers in control of the workflow.
• The GitHub app in Microsoft Teams is now generally available!
  The GitHub app for Microsoft Teams enables teams to create, review, and manage pull requests within chat threads with instant notifications and Copilot suggestions.
• I Turned Off CI for Pull Requests. Here's Why
  Turning off CI for pull requests and using local builds with Basecamp’s gh-signoff extension allows developers to block merges based on a green check status, significantly reducing CI costs while maintaining code quality, as shown by merging 833 PRs without cloud CI since June.

🧬 Hybrid + multicloud

• Simplify Azure Arc agent updates with Automatic Agent Upgrade, now generally available
  Automatic Agent Upgrade in Azure Arc simplifies keeping the Connected Machine agent up to date by allowing Microsoft to manage updates automatically, reducing operational overhead.
• Microsoft Brings GitHub Copilot, Foundry and Databases Into Sovereign Private Cloud
  Microsoft launches GitHub Enterprise Local for running Copilot and Foundry models on private cloud with integration to sovereign databases.
• Stay ahead of Azure Arc agent health with Azure Advisor
  Azure Advisor warns you early of a permanent server connectivity loss and provides steps to restore management before the managed identity certificate expires.

🎭 Identity

• Why identity needs network insight to make Zero Trust work
  Zero Trust security integrates identity with real-time network insights to detect and mitigate risks like session hijacking and unmanaged traffic to non-federated apps, which traditional identity-only approaches cannot address.
• Microsoft offers free clean-up for inactive Agent IDs
  Microsoft provides a free service to automatically identify and deactivate unused Agent IDs, helping organizations reduce security risks without needing premium licensing.
• AI agents are coming—are your identity controls prepared?
  Microsoft Entra recommends adopting phishing-resistant MFA and just-in-time privileged access to mitigate the risks AI agents pose to identity infrastructure.
• New Conditional Access policy plans in Microsoft Entra
  New Conditional Access policies in Microsoft Entra enable identity and security admins to organize access controls by user groups or conditions, simplifying management of complex requirements such as phishing-resistant MFA and compliant devices.
• Microsoft disables public contributions to Entra docs
  Microsoft has removed the option for public users to contribute to its Entra documentation, directing feedback to be submitted through a new internal channel instead.
• Designing Least-Privilege Tool Access for AI Agents with Microsoft Foundry
  Designing Least-Privilege Tool Access for AI Agents with Microsoft Foundry explains enforcing strict permissions via identity management, Azure RBAC, and tool contracts to limit damage from errors or malicious inputs.

🔌 Integration

• Hybrid Logic Apps - SQL as storage provider benchmarking
  Hybrid Logic Apps benchmarked SQL Server 2025 for integration workloads, achieving 45,821 successful runs per hour with sub-second latency and low CPU usage, emphasizing that storage capacity-not server count-drives scalability.
• Automate Vendor Onboarding with MS Forms and Power Automate
  Automate Vendor Onboarding streamlines the process by using MS Forms for intake, SharePoint lists to manage unique vendor details, and Power Automate to prevent duplicate entries efficiently.
• Introducing the Logic Apps local CLI: develop Standard workflows from your terminal
  The new Logic Apps Standard CLI extension for Azure enables developers to create, validate, and manage workflows locally via terminal commands like az logicapp create and az logicapp run.
• Azure API Management v2: Restoring Predictable Outbound IP Addresses with NAT Gateway
  Azure API Management v2 uses Azure NAT Gateway to restore predictable outbound IP addresses, allowing backend systems to set a stable public source IP for traffic through integrated virtual networks.
• Our Logic App stopped processing messages and we dont know why
  The Logic App stopped processing Service Bus messages because the functions worker runtime crashed, but a Turbo360 API ping test restarts it, underscoring the need for proper monitoring and always-on settings.
• Automate SharePoint Document Archiving via Power Automate
  Power Automate enables automated, reversible archiving of aging SharePoint documents through scheduled flows, including ledgers, dry runs, and restoration options.

🎓 Learning and Certifications

• How AI Sees and Hears for AI-901: Hear, Speak, See, and Create
  The AI-901 series explains how speech recognition converts audio to text through capture, feature extraction, acoustic modeling, language modeling for context, and post-processing, while speech synthesis reverses this by normalizing text into natural spoken words using linguistic and prosodic techniques.
• Introducing a credential based on real work—no test required
  Microsoft launches Pro Badge to verify real-world GitHub Copilot usage through an objective five-level proficiency framework recognizing everyday work performance.
• Let’s talk speed and scale: Azure Managed Redis at Microsoft Ignite 2026
  At Microsoft Ignite 2026, Azure Managed Redis demonstrates how it enables fast, scalable applications through real-world case studies and hands-on resources from customers like Sitecore.
• Preparing for Your Next Certification Could Start with Your Visual Studio Subscription
  Your Visual Studio subscription offers free access to Cloud Academy and Pluralsight courses, labs, and exercises to aid in preparing for Microsoft certifications at no extra charge.
• Serverless Compute Operations with Azure Functions
  The new Azure Serverless Compute Operations course on Pluralsight covers implementing, configuring, deploying, and monitoring Azure Functions, focusing on Flex Consumption plan for hybrid scalability and Durable Functions for reliable workflow orchestration.
• AB-400: Power Platform Developer - Study Guide
  The AB-400 exam replaces PL-400 and focuses on extending Power Platform solutions with code and AI, offering a pathway to the Associate credential valid for annual renewal via a free assessment.
• DP-420: Azure Cosmos DB AI Developer - Study Guide
  The DP-420 Azure Cosmos DB AI Developer exam now focuses on designing AI applications, with 40-45 % of questions covering database solution design, up from the previous five-domain format.
• AZ-401: Agentic DevOps Engineer - Study Guide
  The AZ-401 exam requires passing along with GH-300 or AZ-104 and becomes bookable as a beta on October 30, 2026, with full availability in December 2026.
• GH-700 Agentic AI Business Professional Study Guide
  The GH-700 study guide helps professionals improve daily business operations with GitHub and GitHub Copilot, aimed at program managers, business analysts, and product managers.
• My New Course On Pluralsight Is LIVE: Integrating AI Agents in Enterprise Systems
  The Pluralsight course shows how to integrate AI agents with enterprise systems using C# and the Microsoft Agent Framework in 60 minutes.

⚖️ Management and Governance

• Generally Available: Cost management and tagging for Claude models in Microsoft Foundry
  With General Availability, costs for Claude model usage are now tracked within Azure Cost Management under their respective Foundry resources, improving budget visibility for AI workloads.
• Showback, Chargeback, and Allocation: The FinOps Ladder Nobody Gets to Skip
  FinOps teams must prioritize accurate resource tagging and allocation models before attempting to implement complex chargeback or showback systems.
• Bringing AI-Powered SRE to Your On-Premise Server Estate For Unifying Ops
  Azure's SRE Agent extends AI-powered operations to on-premise servers via Azure Arc, providing continuous monitoring, incident detection, root-cause analysis, and automated remediation for Windows and Linux environments without needing data center migration.

🌐 Networking

• Generally Available: Network Watcher VNet flow logs filtering
  Azure Network Watcher now allows admins to filter VNet flow logs by IP address, port, protocol, encryption status, flow direction, enhancing monitoring efficiency.
• Generally Available: Azure Front Door profile and route level WAF policies
  Azure Front Door now provides generally available WAF policies at profile and route levels for comprehensive security without per-domain restrictions.
• Generally Available: HTTP DDoS Ruleset for Azure Front Door Premium WAF
  The new HTTP DDoS ruleset for Azure Front Door Premium WAF offers dynamic protection against the most common attack patterns that cause application downtime.
• Hub and spoke limits are (almost) gone
  Azure has lifted the maximum number of spokes per hub from 500 to 650 for VPN and SDWAN connections, but ExpressRoute remains capped at 1,000 spokes due to route advertisement limitations.
• MANA out-of-tree (OOT) driver for Linux LTS kernels
  The new out-of-tree MANA driver enables running the Azure Network Adapter on supported LTS Linux kernels without upgrades, simplifying maintenance for cloud-standardized customers.
• RateMyAzureFirewall: Grading Azure Firewalls from A to F
  RateMyAzureFirewall evaluates Azure Firewalls on security posture, compliance features, and performance, providing a detailed letter-grade report to help organizations optimize their cloud defense.

🔐 Security

• Microsoft’s Record Patch Tuesday: 1,400+ Vulnerabilities, One Already Under Attack
  October's Patch Tuesday revealed over 1,400 vulnerabilities, with CVE-2026-96951-a in the Windows Shell being notably exploited beforehand, underscoring challenges in AI-driven discovery versus vendor patching.
• PHP2Shell and Azure DevOps: what to check
  Azure DevOps Services and Server do not use PHP for request handling, so users must independently patch any PHP runtimes used in pipelines, containers, or deployed applications following the fixed releases (8.5.12, 8.4.27, 8.3.36, and 8.2.35) to mitigate CVE-2026-109618 and CVE-2026-109619 vulnerabilities.
• Bullet proof security-driven BFF architecture in Azure
  BFF architecture secures web apps by acting as the OAuth client, keeping sensitive tokens out of the browser and replacing them with secure, HTTP-only session cookies.
• Growing our connector ecosystem with 500+ Microsoft Sentinel connectors
  The expansion to over 500 connectors improves integration of data from Microsoft and third-party sources, enhancing risk visibility and enabling quicker, more informed decisions in the Integrated Security Operations Center.
• When Bigger PQC Keys and Signatures Break Software Assumptions
  Post-quantum cryptographic keys and signatures often exceed the 1,500-byte MTU, causing silent failures in network tools that assume smaller packet sizes.
• Cyber Security Awareness Month: Back to the Basics
  The 2026 Microsoft Digital Defense Report shows AI is shortening attack timelines to under 24 hours while enterprise remediation for critical vulnerabilities still averages 30-60 days, emphasizing the ongoing importance of security practices.
• Securing SAP ABAP-to-HANA Authentication with Azure Key Vault Managed HSM (Hardware Security Module)
  This solution uses Azure Key Vault’s Managed HSM to securely store and manage private keys for SAP ABAP-to-HANA authentication, reducing portable credential theft risks through in-hardware cryptographic operations.
• Azure Cloud HSM Hardware Audit Logging: Verifiable audit trails for security and compliance
  Azure Cloud HSM now provides Hardware Audit Logging for trusted, device-originated evidence of user actions and management operations on hardware security modules, aiding monitoring, compliance reporting, and forensic investigations.
• Defending Linux in the age of AI-operated attacks
  Microsoft Defender for Linux enhances runtime protection against fileless and in-memory attacks through continuous validation and AI-powered simulations to combat advanced threats.
• Now generally available: Discover local AI agents with Microsoft Defender
  Microsoft Defender is now generally available, enabling security teams to discover and monitor over 50 types of local AI agents and their Model Context Protocol servers across Windows and macOS.
• IDC MarketScape names Microsoft a Leader in the 2026 Worldwide Modern Endpoint Security for Enterprises Vendor Assessment
  Microsoft earned the title of Leader in IDC's 2026 MarketScape for Worldwide Modern Endpoint Security due to its advanced autonomous attack disruption features, AI-powered Security Copilot, and extensive cross-platform support suited for regulated enterprises.
• Securing AI in the Defense Industrial Base – Part 1 – The AI Risk Landscape
  Defense Industrial Base organizations can secure AI by extending existing Zero Trust practices and NIST 800-171 compliance to Microsoft 365 GCC High environments.
• OneDrive Username Enumeration: What Security Teams Should Know
  OneDrive username enumeration enables attackers to discover user principal names for phishing, credential stuffing, and social engineering, posing low risk due to lack of logging in Microsoft Entra or Purview but can be mitigated with multifactor authentication, Conditional Access policies, and throttling excessive requests.
• From help desk impersonation to AD CS abuse: Investigating ESC1 and ESC8 intrusion paths
  Microsoft Teams impersonation allows attackers to gain remote access and exploit AD CS misconfigurations like ESC1 and ESC8 to escalate privileges to domain controller levels.
• Rotate Azure AI Service Keys with Zero Downtime
  Azure Key Vault enables zero-downtime rotation of Azure AI service keys by refreshing each application instance sequentially and confirming functionality before deactivating the previous key.
• What’s new in Microsoft Sentinel: October 2026
  Microsoft Sentinel introduces cross-tenant delegated access for secure multi-tenant operations with least privilege, a unified ASIM EmailEvent schema, an SAP analytics rule for vulnerable custom app detection, and Unified Data Management in the Defender portal.
• Inside a malicious Claude Code-branded extension and its persistent SSH backdoor
  The Claude Code extension injected JavaScript to collect host info, add a controlled SSH key, enable local SSH services, create reverse tunnels for persistence, and attempt reporting despite failures.

📦 Storage

• Generally Available: Higher limit of Azure Blob Storage accounts per region per subscription
  The updated quota allows Azure subscriptions to manage up to 1,000 Blob Storage accounts per region, addressing limitations for large-scale cloud deployments.
• Public Preview: Cross Tenant Restore with Azure Backup
  Cross-Tenant Restore in Azure Backup lets you recover VM backups into a different tenant, enhancing cyber resilience without needing custom solutions.
• RateMyAzureBackups: Grading Every Azure Backup Vault From A to F
  RateMyAzureBackups evaluates and grades Azure backup vaults on factors like compliance, performance, and ease of use, providing transparent scores from A to F.

🔗 Web

• From WordPress.com to a static site on Azure
  Moving baeke.info from WordPress.com to a static site on Azure improved the writing workflow with Markdown in git, boosted speed via CDN delivery, increased control over design and content, and reduced hosting costs.
• Plan Your Java Upgrade Ahead of Microsoft’s 2027 Support Milestones
  Planning ahead for Java upgrades is crucial as September 2027 marks the earliest end-of-support date for Java 11 and 17 on Azure, including App Service and Functions.
• Adding a low-cost chat assistant to my blog
  Add a low-cost static-site-compatible chat assistant to baeke.info using Azure Functions and inexpensive keyword search, conducting free checks before costly LLM calls.