Articles by Aarav Patel

AI agents are increasingly being used for tasks that require more than generating text. An agent may execute code, process files, call APIs, run data transformations, or perform multiple tasks at the same time. That creates an important production challenge: how do you let AI agents execute work wit

AI coding assistants are becoming more capable. Instead of only suggesting a line of code, an AI coding agent can inspect files, make changes, execute commands, run tests, and continue working based on the results. That additional capability is useful, but it also creates an important engineering qu

Learn how AI agents behave without users, including identity, authorization, tool limits, retries, idempotency, human approval, security, and monitoring.

AI coding assistants have moved well beyond autocomplete. Modern coding agents can inspect a repository, understand related files, modify code, run tests, investigate errors, and continue working across multiple steps. That changes how developers should evaluate an AI model inside an IDE. A simple q

Learn how to control AI agent network access in Microsoft Foundry using identity, egress controls, private networking, tool authorization, and least privilege.

Learn how to test an AI app after a GPT-6 model update using regression tests, structured outputs, tool calls, RAG, security checks, and production validation.

Learn how OpenAI Prompt Caching works, improve cache hits, reduce input costs, and design stable prompts for faster, more efficient AI applications.

Learn how to build interactive AI agents in C# with conversation state, tool calling, follow-up questions, human approval, and multi-step workflows.

Learn how to design AI agent memory in C# with short-term context, long-term preferences, semantic retrieval, expiration, and practical memory best practices.

Learn how to add human approval to long-running C# AI agents using workflow state, approval events, authorization, idempotency, and durable execution patterns.

Learn how Microsoft Agent Framework helps C# AI agents recover from failures using durable workflows, retries, checkpoints, and safe execution patterns.

Learn how GitHub Copilot handles million-line pull requests, including context limits, large diffs, generated code, AI review quality, and practical testing strategies.

Learn how GitHub Copilot code review works, what changed in its latest workflow, and how developers can verify AI suggestions with human and automated checks.

Learn how to test GitHub Copilot sandbox rules for filesystem, network, Git, credentials, builds, and permissions before using an AI coding agent on projects.

Learn how GitHub Copilot's local sandbox controls filesystem, network, and credential access, with practical commands to inspect and tighten permissions.

Learn how whole-codebase indexing in GitHub Copilot CLI helps C++ developers trace code, understand dependencies, refactor safely, and debug large repositories.

Learn how GitHub Copilot CLI uses whole-codebase indexing to understand large C++ repositories, trace dependencies, explore code, and support safer refactoring.

Learn how to prepare GitHub Actions workflows for the move to Ubuntu 26.04. This article explains how to identify affected workflows, test with ubuntu-26.04, compare Ubuntu runner versions, pin SDK and runtime versions, check system dependencies

Learn how to manage GitHub code coverage rulesets with the REST API, automate policies across repositories, test coverage requirements, and troubleshoot enforcement.

Learn how to test GitHub Actions workflow execution rules with Evaluate mode, policy insights, actor and event controls, workflow targeting, and safe enforcement.

Learn how GitHub Actions workflow protection controls actors and events, secures sensitive workflows, uses evaluate mode, and strengthens CI/CD security.

Learn how .NET 11 improves string.Split performance, manage allocations, choose the right overloads, and use span-based parsing for high-throughput workloads.

Learn why regex performance improves in .NET 11 through source generation, engine optimizations, non-backtracking mode, reuse, timeouts, and benchmarking.