Browse All Posts (178)

devanshirastogi explains upcoming changes to Azure Firewall explicit proxy and provides a migration walkthrough for PAC file–based setups, including moving PAC retrieval to customer-managed Azure Storage and using Managed Identity with the right RBAC roles. The guide includes portal steps plus PowerShell and Azure CLI examples for configuring Firewall Policy.

WSilveira explains an upcoming Azure Logic Apps hosting-model migration tied to the move toward .NET 10 support, including when the change is automatic and when customers need to update NuGet-based deployment processes to avoid unexpected out-of-proc migration.

Microsoft Developer shares a short clip featuring Amanda Silver on how Microsoft manages open source projects, focusing on how collaboration, community input, and real-world usage feedback influence how projects evolve over time.

Manfred Riem explores Spec Kit in an Open Source Friday replay, focusing on how clear specifications and shared context help contributors collaborate effectively and build better tools in the open.

varghesejoji introduces the ARM MCP Server and a GitHub catalog of 24 proof-of-concept agents that use Model Context Protocol tools to query Azure Resource Graph and (optionally) deploy ARM templates. The post focuses on safe, repeatable “determinism” patterns for governance, FinOps, and SRE workflows in Azure.

rhack summarizes an Azure Essentials Show episode with Thomas Maurer and Raffaele Garofalo on what typically derails AWS-to-Azure migrations and how to avoid it, focusing on stakeholder alignment, migrating like-for-like before optimizing, and using a blue-green cutover to keep rollback options open.

Nikita Bajaj explains how Azure Migrate integrates with GitHub Copilot Modernize (public preview) to generate portfolio-level code insights across multiple applications and repositories, helping teams assess Azure readiness, review remediation guidance, and plan modernization work using a shared workflow between migration admins and developers.

Allison announces an update to the GitHub Copilot usage metrics API that adds per-user AI credit consumption, letting orgs and enterprises track daily usage patterns and relate consumption to adoption and budgeting.

Natalie Guevara explains how GitHub built Qubot, an internal GitHub Copilot-powered analytics agent that lets employees ask plain-language questions over warehouse data and get answers quickly, with results captured as markdown reports in pull requests and validated through an offline evaluation framework.

OpeSam-Olowu explains what Microsoft announced at Build 2026 for running AI locally on Windows in telecom environments, covering Microsoft Foundry on Windows (Windows AI APIs, Foundry Local, and Windows ML), new on-device Aion models, and OS-level controls for secure, governed agent execution.

John Savill runs through the weekly Azure update for 19th June 2026, covering a VS Code Azure Functions project-creation refresh, Azure Migrate’s GitHub Copilot integration, networking and storage feature updates, Azure Databricks integration with OneLake, and new Log Analytics summary rules.

GitHub shows a quick demo of the GitHub Copilot app by asking its agent-first desktop experience to create a custom canvas workspace and run a WebAssembly port of Doom, illustrating what Copilot canvases can execute inside the app.

Jesse Houwing explains how to connect Azure Pipelines to GitHub Enterprise Cloud with data residency (ghe.com) by manually installing the Azure Pipelines GitHub App, then creating a service connection and rewiring existing pipelines to use the new GitHub repo as their source.

John Edward shares a practical list of 25 project ideas for building custom AI agents with Microsoft Copilot Studio, ranging from personal productivity assistants to customer support, onboarding, IT help desk, and document review bots.

ShubhraS announces the general availability of Microsoft Signing Transparency (MST), a SCITT-aligned transparency ledger that logs production builds for selected Microsoft cloud services. The post explains how customers can independently verify code integrity using cryptographic proofs and audit trails, and introduces the Ledger Explorer tool for offline verification.

The Microsoft Defender Security Research Team explains AutoJack, an exploit chain in AutoGen Studio that turns an AI browsing agent into a remote code execution path by abusing localhost trust, missing authentication on MCP WebSockets, and unsafe command parameter handling, plus practical hardening guidance and Defender hunting queries.

Julia Kasper (VS Code Eval Team) breaks down what an ultra-simple “write HELLO.txt” agent eval revealed after 50,000+ runs: models vary widely in tool-call discipline, planning overhead, and output-token cost, and those differences matter for latency, billing, and automatic model selection in VS Code.

Allison announces that the Opus 4.6 (fast) model will be deprecated across GitHub Copilot experiences on June 29, 2026, and points teams to Opus 4.8 (fast) as the replacement, including notes for Copilot Enterprise admins on enabling the model via policy.

jovanpop-msft introduces preview features in Fabric Data Warehouse that add approximate string-matching functions and modern string operators for T-SQL. The post shows how to use distance and similarity scoring to find misspellings and variants (like “Hongkong” vs “Hong Kong”), plus new concatenation and Unicode helpers for clearer SQL.

Gaurav Bhardwaj describes a hybrid Azure-based document extraction architecture for construction drawings and project documentation, combining deterministic field extraction with bounded LLM verification. The post breaks down the event-driven pipeline, confidence gating, cost trade-offs, and the security controls needed to run this kind of document intelligence workflow in production.

Govind Kamtamneni explains how to build an outcome-driven “learning system” in Microsoft Foundry using OpenEnv environments, rubric-based evals, and a closed-loop optimizer. The post contrasts non-parametric harness tuning with parametric post-training (ECHO), and shows how Azure Container Apps sandboxes provide an isolated, enterprise-ready runtime for agent rollouts.

GitHub features guests Maddy Montaquila and David Fowler on Open Source Friday to discuss .NET Aspire, a code-first orchestration and observability layer for distributed apps, including how it models services and dependencies and streamlines running and troubleshooting multi-service systems.

Allison announces that MAI-Code-1-Flash, Microsoft’s small coding model tuned for GitHub Copilot, is rolling out across more Copilot surfaces, including IDEs and the Copilot CLI, with availability expanding gradually and Business/Enterprise access coming soon.

arindamc shows how to build a multi-city, real-time weather monitoring pipeline in Microsoft Fabric using Eventstream and Eventhouse AI Skills, turning a repetitive portal setup into a single natural-language prompt that deploys sources, filtering, and KQL-backed alerting in minutes.

Jay Li explains how to migrate large Azure hub-and-spoke environments away from ExpressRoute hairpin routing through Microsoft Enterprise Edge (MSEE) and toward Azure Virtual Network Manager (AVNM) mesh. The post covers scale limits, enabling High-Scale Private Endpoints (HSPE), rollout/rollback strategy, and how to keep security inspection and segmentation intact.

Michaela Isaacs outlines how Microsoft Fabric is tightening security and predictability for Power Query connectors by bringing connectors in-house, defining a clearer connector lifecycle (Preview → GA → migration → retirement), and highlighting recently shipped V2 connectors for major data platforms.

Allison announces updates to GitHub Copilot code review, including support for repository-level AGENTS.md instructions and UI changes that make it easier to request reviews on draft pull requests while reducing timeline noise.

Daniel Roth and guests introduce dotnet/skills and explain how it helps AI assistants produce more accurate, idiomatic .NET code by grounding responses in reusable skills/tools and testing against real-world evaluation scenarios.

Allison announces a public preview feature in GitHub Issues that suggests potential duplicate issues during issue creation, plus new MCP server support that lets AI tools read and write issue fields to create more fully triaged issues automatically.

Dona Sarkar shares a standout Build 2026 takeaway for developers: AI agents don’t necessarily need their own standalone app, and can instead be built to show up in the places users already work.

Erin Stellato introduces GitHub Copilot Agent mode (preview) in SQL Server Management Studio (SSMS), showing how you can give Copilot a high-level goal in natural language and have it iterate by running queries and working through results, with controls for read-only vs read/write behavior.

Allison announces an update to GitHub pull request search: the author: qualifier now returns both human-authored PRs and PRs opened by Copilot cloud agent on a user’s behalf, making it easier to track everything you’re responsible for across the UI, mobile, and APIs.

Allison announces general availability of GitHub’s new repository switcher in the global navigation, which lets you jump between repositories from a dialog without leaving the page—useful when working across many repos in the same organization or owner namespace.

GitHub demonstrates the new GitHub Copilot CLI integration in JetBrains IDEs, showing how ask, plan, and agent modes can drive multi-step execution for tasks like working on a Spring Boot app, plus where to review token usage and debug logs.

Natalie Guevara introduces GitHub’s new pull request limits for maintainers who are dealing with high volumes of low-quality or spammy contributions, including AI-generated PRs. The post explains how the limits work today and outlines upcoming controls like PR archiving, issue limits, and smarter trust signals.

Allison announces new GitHub Actions capabilities for custom images on GitHub-hosted runners, including layered image builds (building a custom image from another custom image) and conditional control over when new image versions are generated via the snapshot keyword.

samcogan explains how to use Azure API Management (APIM) as an MCP gateway to control which tools an agent can see and invoke when connecting to an MCP server. The post covers governance patterns (allowlist vs deny-list), plus auth, logging, network isolation, and streaming-related trade-offs.

Nitin Jadhav demonstrates how pg_duckdb brings DuckDB-style analytical performance into Azure Database for PostgreSQL, including running faster OLAP queries on existing Postgres tables and querying Parquet files from object storage without loading them into Postgres first.

Chun Lin Goh explains why burstable cloud instances can cause non-linear PostgreSQL slowdowns when CPU credits run out, and shows how to model the saturation point with a lightweight discrete-event simulation approach so you can right-size instances and connection pools before production timeouts hit.

Sakshi Nasha explains how to secure PostgreSQL by addressing common blind spots that show up when moving from development to production, including role design, schema isolation, SQL injection risks, and safer use of functions and extensions.