Browse All Posts (378)

GitHub demonstrates how to extend GitHub Copilot code review using Model Context Protocol (MCP) and custom skills, so reviews can incorporate internal documentation and repository-defined checklists to produce findings aligned with a team’s engineering standards.

Microsoft Defender Security Research Team, Dor Edry and Amit Eliahu break down a prompt-injection pathway in Anthropic’s Claude Code GitHub Action that could leak CI/CD secrets by reading /proc/self/environ, and provide practical hardening guidance for AI-powered GitHub Actions workflows.

Charles Feddersen and Abe Omorogbe explain how AI apps and agents change database design, focusing on reasoning over operational data instead of only transactions. They demo new capabilities across Azure SQL Database, Azure Cosmos DB, and Azure HorizonDB (cloud-native PostgreSQL) to simplify architectures and reduce latency.

Seth Juarez explains how Azure AI Foundry Toolboxes let teams build, discover, and govern tools across multiple AI agents, reducing duplicated integration work around authentication, credentials, and endpoint wiring.

Scott Hanselman hosts a Microsoft Build live “vibe check” where AI-assisted demos are put under scrutiny: what the AI actually built, where the seams are, and whether the result is a clever prototype or something that could hold up in production.

Cassidy Williams joins Scott Hanselman and Mark Russinovich for a live Microsoft Build session where they “vibe check” an AI-assisted demo, digging into what the AI produced versus what required human fixes, and where the seams show up when you push a prototype toward something more real.

Scott Hanselman hosts a live Build 2026 “vibe check” with Simon Willison, reviewing AI-assisted demos and digging into what the AI actually built, where the seams are, and what it would take to move from clever prototype to production-ready software.

Pierce Boggan recaps day one highlights from Microsoft Build 2026, focusing on how VS Code and GitHub Copilot roles are evolving, what’s coming next for AI adoption in the editor, and how agent-style workflows are changing developer expectations.

Addy Osmani (with Burke Holland) explains what “agent skills” are and how to structure them so AI coding tools can follow repeatable workflows that reflect senior engineering judgment, producing more consistent, production-quality outcomes.

Patrick Nikoletich and Burke Holland introduce the Copilot SDK and show how it can be used to extend GitHub Copilot by building custom, agentic experiences on the same runtime that powers Copilot.

Evan Boyle and Burke Holland walk through what’s new in GitHub Copilot CLI, including a redesigned terminal interface, a “Rubber Duck” workflow for second opinions, recurring prompts with /every, and a hands-free voice mode aimed at making terminal-based coding and review faster to iterate on.

Milos Colic shares how Xoople scaled Python-based AI workloads on Azure using Ray via Anyscale, covering the distributed-systems challenges behind data ingestion, training, and inference, and why the team prioritized delivering outcomes over operating clusters.

Nikisha Reyes-Grange introduces Azure HorizonDB and Rayfin, focusing on how these Azure Data and Microsoft Fabric innovations aim to modernize PostgreSQL operations and simplify building and running data applications, including SQL-level AI functions and hybrid search concepts.

Jeff Hollan and Lee Stott explain how hosted agents in Microsoft Foundry help teams move from local agent prototypes to production-grade AI systems, with a focus on identity, isolation, evaluation, and lifecycle management so developers can deploy secure, scalable agents with clearer operational boundaries.

Vivek Bhadauria discusses how Microsoft built an end-to-end “observe → evaluate → optimize” workflow for AI agents, sharing practical lessons on agent observability, context-specific evaluation rubrics, and using inner- and outer-loop signals to continuously improve agent behavior in production.

Brad, an OpenClaw maintainer, shares how he uses the GitHub Copilot app to triage and prioritize large volumes of GitHub Issues and pull requests across his open source work, including using multiple major models under one subscription for cross-checking results.

Seth Juarez and Burke Holland introduce the GitHub Copilot app, a desktop experience aimed at agent-driven development where you can hand off an issue, watch agents work, review the diff, and merge changes from a single screen.

Marco Casalaina and Ayca Bas explain how to build grounded, enterprise-ready agents by delegating context to three “IQ” systems: Foundry IQ for knowledge, Fabric IQ for data, and Work IQ for human/work context, with a live demo showing how this approach avoids brittle, hand-wired pipelines.

Ana Schafer shares a quick update from Microsoft Build 2026 with Qualcomm, focusing on the shift toward AI agents and context-aware systems, plus what cross-device integration means for developers building experiences that span wearables, edge devices, and data centers.

Vivek Chauhan shares a quick on-the-show-floor update from Microsoft Build 2026 on Fireworks AI, including the scale they run at and how their partnership with Microsoft connects to Azure AI Foundry for governance, security, and reliability.

Adrian Macias discusses how open-source AI development is shifting across local AI PCs and Azure, covering agentic AI, AI-assisted coding, and the practical need for flexible deployment options as teams experiment and scale AI workloads.

Stephen McCullough shares a short NVIDIA “quick take” segment recorded at Microsoft Build 2026, highlighting Nemotron as a high-intelligence, open-source AI model and calling out the value of the latest NVIDIA and Microsoft-related innovations discussed during the event.

Chu Lahlou demonstrates how Azure AI Foundry can turn existing enterprise APIs, tools, and data into callable capabilities for AI agents, with grounding via Foundry IQ and runtime governance to keep tool usage observable and controlled.

Umang Sehgal and Lily Du show how to build agents that operate inside Microsoft Teams—participating in chats, channels, and meetings—so they can automate tasks, surface insights, and take action in context without forcing users to leave their workflow.

Tyler Leonhardt explains what it means for Claude to run as a coding agent inside GitHub Copilot in VS Code, focusing on how the integration behaves at the code level, how context is assembled, and what changes when you select Claude explicitly versus letting Copilot choose.

Burke Holland and Reynald Adolphe show how to use GitHub Copilot CLI inside VS Code for “rubber duck debugging”: having a second model family review and challenge the first during planning, implementation, and testing to help catch mistakes earlier.

Nish Anil, Hazem El-Hammamy, and Jeff Fritz show how GitHub Copilot’s modernization capabilities use agentic AI to analyze large legacy codebases, map dependencies, plan upgrades, and refactor safely at scale, including governance concepts like rule books and command-center style oversight.

Harald Kirschner walks through the new Agents window in Visual Studio Code, focusing on how it improves visibility across agent sessions, supports multi-workspace workflows, and reduces cost through token optimization and automatic model routing.

Courtney Webster and Burke Holland discuss how AI-driven, prototype-first workflows are changing the traditional PM-to-developer handoff, including PMs contributing directly via pull requests and teams iterating faster with tighter feedback loops.

Anthony Shaw explains what tensors are and why they matter for how ML models run, then connects that understanding to writing better prompts and benchmarking when using GitHub Copilot to optimize code.

Justin Chen and Burke Holland demonstrate VS Code’s integrated browser and how it fits into a real development workflow, including sharing browser tabs as agent context, inspecting page content, interacting with elements, running Playwright scripts, validating changes live, and debugging with breakpoints without leaving the editor.

Julia Kasper and Seth Juarez give an inside look at how the VS Code and Copilot teams evaluate and ship AI model updates, including how they test model quality, compare model behavior on the same prompts, and balance capability improvements with reliability during rollouts.

Seth Juarez, Pierce Boggan, Yang Liu, and Pengcheng He explain how Microsoft AI trains and evaluates code-focused models that power GitHub Copilot, including what makes coding models different, how developer workflows influence optimization, and how improvements show up inside VS Code.

Joanna Oikawa explains how the VS Code design team is adapting the editor’s user experience for more agentic workflows, sharing concrete UX changes, the trade-offs behind them, and lessons learned from what didn’t work.

Visual Studio Code hosts a Microsoft Build 2026 live stage session with demos and discussion spanning GitHub Copilot, the Copilot SDK, and VS Code workflows. It touches on agent integration, multi-model verification, security concerns in AI code review, and developer tooling updates shared by the teams building them.

Johnson Shi, Zoey (Zhuyu) Li, and Huangli Wu announce public preview support for regional endpoints in Azure Container Registry geo-replication, including the new Azure CLI and portal experience, endpoint URL formats, and practical guidance for pinning pushes/pulls and Kubernetes workloads to specific replicas.

Freddy Chiu demonstrates how to profile and tune agentic AI applications on Intel-powered Windows PCs, focusing on end-to-end performance across CPU, GPU, and NPU. The session shows how to collect telemetry, identify bottlenecks, and apply practical optimization techniques to improve responsiveness and power efficiency.

shijain13 explains what’s new in the Azure Monitor Health Model (Preview), focusing on expanded discovery options, faster health signal setup, and new aggregation rules that help teams reason about workload health with less alert noise and clearer troubleshooting paths.

Sam Foo explains how Pod CIDR expansion works for Azure CNI Overlay in Azure Kubernetes Service (AKS), and what to consider when planning pod IP ranges for long-lived clusters as they scale.

davidwright, Arnaud Lheureux, and Suzanne Daniels explain why architecture and governance frameworks only help when they actively change delivery decisions. Using Git-Ape as the example, they show how to turn Azure Well-Architected, Azure Policy (including NIST mappings), and CAF guidance into repeatable repo-driven assessments with prioritized findings tied to code and policy.