Browse All Posts (155)

Andrea Griffiths explains how to package repeatable tasks as agent skills that work across GitHub Copilot and other clients. Using independent verification and analytics pipeline examples, she shows how to separate project context from instructions, share execution paths, and enforce human approvals and CI checks without confusing instructions with permissions.

GitHub explains October updates to GitHub Copilot, covering self-learning automations for issue triage, Projects for coordinating work across repositories without local cloning, and Fusion for routing tasks across frontier models. The episode connects these capabilities to routine pipeline work and discusses what software factories mean in practical development workflows.

davidwright and Arnaud Lheureux explain how Git-Ape community skills package specialist procedures for AI-assisted platform work. The walkthrough covers discovery, workspace-scoped installation, contribution requirements, and proposed skill packs, emphasizing pinned source revisions, explicit approval, and the distinction between installing procedures and validating or executing their output.

Allison announces that GitHub Actions hosted runners will end support for ARM32 job execution on Windows and Linux by December 8, 2026. The change also removes native AArch32 user-space execution support and follows the end of Node.js 20 support, as Node.js 24 does not support ARM32.

Adrian Twarog explains why visibility into AI spending matters before expanding agent workloads. Microsoft Foundry runs inside Azure, bringing model usage, credits and spending into one place so teams can understand the costs associated with their AI applications.

Rutvi Trivedi, Jeff Bloomfield, and Murali Ambati examine why portable ML compilers struggle to deliver hardware-efficient execution. They explain operator fusion, multi-level intermediate representations, scheduling, and memory constraints, then outline how Windows ML and DirectX could help balance cross-device scalability with hardware-specific performance for local AI.

Carlos_Mayol explains how Azure Local 2610 improves alignment with CIS and DISA STIG benchmarks while allowing individual security settings to be customized without disabling drift control. The update adds audit metadata and preserves cluster policy intent across solution updates, with safeguards preventing changes to settings essential to platform security.

Microsoft Developer explores fuzzy string matching in Azure SQL and SQL Server for application search that tolerates misspellings and inconsistent data. Practical demonstrations compare fuzzy matching with LIKE, Full-Text Search, and regular expressions, showing where each approach fits customer lookup, product search, duplicate detection, and data cleansing without adding AI complexity.

Microsoft Security Research, Microsoft AI Red Team, Amanda Minnich, Amit Eliahu, Blake Bullwinkel, Dor Edry and Giorgio Severi examine how poisoned training data and model checkpoints can hijack coding agents. Their experiments show backdoors transferring across repositories and tool interfaces, reinforcing the need for provenance checks, deployment-specific evaluation, and restricted tool permissions.

Sameen Karim demonstrates how GitHub stacked pull requests split complex changes into smaller, dependent reviews. The session covers building stacks with GitHub CLI, reviewing on github.com, adding pull requests, atomic merges, branch protection, auto-merge, and fixing CI with stack rebasing.

Allison reports that GitHub secret scanning now detects Cloudflare service tokens in repositories. The update adds the cloudflare_access_service_token pattern, extending supported secret detection. The post also explains how partner secrets are reported to issuers and how user secrets generate alerts in public or private repositories.

Allison explains how direct links make individual GitHub public monitoring alerts easier to revisit and share. Clicking an alert number opens the finding, whose URL can be bookmarked or shared with teammates who have access. Public monitoring is available to eligible GitHub Enterprise Cloud customers at no additional cost.

John Edward explains Microsoft’s planned in-conversation connection setup for Copilot Studio agents, expected to reach general availability by December 31, 2026. The article outlines conditional connection reuse, unchanged authentication and access controls, and the changes developers must make to custom web interfaces using the Standard harness.

Jay Soni and Jordi Rey explain two public previews in Azure Application Gateway: advanced routing rules and advanced load balancing. They cover request matching, weighted traffic splits, priority-based failover and per-pool algorithms, with examples for canary releases, multi-tenant applications and AI endpoints, plus important preview limitations.

Mads and Dustin explore upcoming C# features, focusing on union types for modeling closed sets of data shapes, such as those found in wire protocols. They discuss how unions express intent and support consuming code, and also look at closed classes and extension indexers.

DeepGanguly introduces Azure Container Apps as an assessment target in Azure Migrate. The article explains how to discover web applications, evaluate readiness, estimate resources and hosting costs, organize migration waves, and use GitHub Copilot app modernization to prepare and deploy Java and .NET applications.

Michael_Flanakin details FinOps toolkit 15 updates, including a cross-platform AI plugin, local FinOps hubs, and NAT Gateway support for private deployments. The release also improves reservation and savings plan datasets, corrects cost data issues, and updates PowerShell commands, reports, and Azure Optimization Engine ingestion.

dotnet explains how the .NET 11 CLI became a native application, sharing the work involved and the gotchas encountered along the way. The session focuses on lessons developers can apply when building their own Native AOT applications, alongside the CLI's responsiveness and flexibility.

dotnet presents a demo-focused tour of ASP.NET Core and Blazor in .NET 11, covering performance, security, and observability alongside asynchronous validation, richer API contracts, and server-side rendering. The session also explores scalable Blazor applications, deeper Aspire integration, and building blocks for distributed applications and agentic user interfaces.

dotnet explores how WinForms in .NET 11 modernizes existing applications without breaking behavioral or designer compatibility. The session covers opt-in visual styles, accessibility and UI responsiveness improvements, and architectural boundaries that support safer AI-assisted modernization and the addition of intelligent workflows to established business applications.

dotnet explains how to try multithreaded MSBuild to speed up frequent builds and shorten development feedback loops. The video covers the new execution mode, what it means for existing builds, and how to prepare projects and tools to take advantage of multithreaded execution.

dotnet explores Runtime Async and the updated .NET memory safety model. The video covers moving async orchestration and code generation from Roslyn to RyuJIT for better performance, alongside stricter compile-time enforcement and improved auditability of unsafe code use.

dotnet explores the rebuilt foundations of C# Dev Kit 11, including a single Native AOT server, persistent caches, and faster incremental builds. The session explains the architectural choices behind shorter workspace load times and considers how these foundations could support .NET workflows beyond VS Code.

Maddy and Damian from the Aspire team discuss improvements to the Aspire experience for .NET developers. They explain how supporting additional languages prompted the team to revisit early .NET-centric assumptions, with changes to the Aspire + .NET workflow that are available today or coming soon.

Jose and Maddy explore how Aspire uses .NET to support developers and AI agents across programming languages. They cover the CLI, AppHost, Dashboard, and tooling, explaining how Blazor, Native AOT, and code generation underpin workflows for running, debugging, observing, and deploying distributed applications.

dotnet explores the .NET 11 numerics toolbox, including decimal floating-point types, BFloat16, generic complex numbers, and random-number APIs. The video covers precision and range choices, rounding, parsing and formatting, and layout-aware tensor operations, while distinguishing current improvements from planned .NET 12 tensor work.

dotnet explores the expanded post-quantum cryptography capabilities in .NET 11, looking at how these features can be used today and how the platform is preparing for future cryptographic needs.

dotnet explores how .NET 11 brings dotnet run, dotnet test, and dotnet watch to .NET MAUI on mobile devices and emulators. The talk covers device selection, app installation, test results, and live updates, alongside the platform challenges and practical workflows for local development and CI.

Barry returns to review .NET security bugs from 2025–2026, focusing on what developers can do to avoid similar pitfalls. The session connects the bugs with practical precautions for developers working with .NET.

dotnet explores how model abstractions, retrieval, tools, AI agents, and local inference fit together in the .NET AI stack. Through examples and demos, the session examines available building blocks, emerging patterns, and how developers can compose them into useful AI-powered applications.

EllaHassan_MSFT explains how Confidential Data Disk Encryption extends Azure confidential VM protection to temporary and attached data disks. Currently in private preview, the feature combines guest-side encryption, customer-managed keys, and attestation-based key release to protect application data without retaining the Azure host within the encryption trust boundary.

Wiliam_Rosa explains how to detect stolen-credential attacks by correlating Microsoft Entra ID login risk with Azure Databricks audit events. SQL examples illustrate suspicious event sequences, while the discussion highlights logging prerequisites, diagnostic coverage gaps, and the limitations of rule-based detection before teams rely on it in production.

Maira Ladeira Tanke, Mark Roy and Vivek Bhadauria outline engineering practices for moving Foundry agents into production. They explain when deterministic logic beats model reasoning, how to manage context and evaluation costs, and why tracing, delegated identity, session isolation, and capacity planning belong in the architecture from the start.

Jerry Nixon connects three SQL AI scenarios—agent-assisted database administration, controlled data access through MCP, and retrieval-augmented generation—to the DP-800 certification. He outlines learning resources covering database design, security, deployment, embeddings, and search for developers building AI-enabled solutions with SQL Server, Azure SQL, and Microsoft Fabric.

sbaynes describes how Microsoft and Novo Nordisk moved a clinical research assistant from custom Azure infrastructure to Foundry hosted agents. The case study explains session-isolated execution, managed scaling, private networking, and practical lessons for subnet planning, tracing sensitive workloads, and preserving expert-backed evaluations during migration.

Adelina Balasa presents a six-pillar framework for moving AI agents from demos to production. The session covers architectural foundations, governance, security, operational reliability, and cost optimization, explaining why production readiness requires more than model selection and how teams can approach deployment and scaling.

Yash Gund, Sumith Maniath and Microsoft Security Research examine how attackers use GSocket to retain encrypted remote access to compromised Linux servers after application patching. The analysis covers relay-based command and control, persistence techniques, Microsoft Defender detections, advanced hunting queries, and host remediation needed to remove continued access.

Devi_Priya walks through deploying a Streamlit-based AI application on Azure Container Apps, from container packaging and registry access to secrets, ingress, scaling and monitoring. The guide explains secure application and pipeline identities, revision-based releases, and why increasing container replicas does not increase Azure OpenAI model capacity.

Joe Filcik introduces GPT-5.6, GPT-6, and GPT-6.1 support in Azure Content Understanding, with evaluation results for document, video, and speech workloads. He explains how to compare accuracy and token costs, improve extraction with labeled examples, and test model deployments using Content Understanding Studio or the REST API.

Sue Black and Kent C. Dodds discuss how developers decide what is worth building when AI makes software creation easier. Their GitHub Universe 2026 conversation explores responsible software, tool choices, maintenance, and the trade-offs between shipping faster and ensuring that what gets built is fit for purpose.