# 5 Microsoft Tutorials: Copilot Model Swapping, DeepSeek on Azure, Rust, .NET 10 RC, AI Agents

> Source: <https://ainexusdaily.vercel.app/article/2026-09-13-5-microsoft-tutorials-copilot-model-swapping-deepseek-on-azure-rust-net-10-rc-ai>
> Published: 2026-09-13 13:03:06+00:00

# 5 Microsoft Tutorials: Copilot Model Swapping, DeepSeek on Azure, Rust, .NET 10 RC, AI Agents

This week's Microsoft-ecosystem signal splits into two buckets: AI plumbing you can wire up in an afternoon (model routing, serving open-weight models, agents), and platform shifts with long shelf life (Rust going tier-1, .NET 10 RC). Here are the five tutorials that came out of the research, each w

This week's Microsoft-ecosystem signal splits into two buckets: AI plumbing you can wire up in an afternoon (model routing, serving open-weight models, agents), and platform shifts with long shelf life (Rust going tier-1, .NET 10 RC). Here are the five tutorials that came out of the research, each with a five-minute first action. What it is: An open-source VS Code companion that runs a local proxy (9Router) so you can swap the AI model behind Copilot Chat mid-conversation — no editor reload, no restart. Why now: The repo was created on 2026-09-10 and already sits at 327 stars in this week's GitHub new-repo crawl. It's the fastest-moving brand-new repo in the signal. Tutorial angle: Clone the repo, start the 9Router proxy, point Copilot Chat at the local endpoint, add two provider configs, then switch models mid-conversation and confirm in the router logs which model actually answered. The verification step matters — it's the difference between "I think it switched" and proof. Difficulty: Beginner Start today: git clone https://github.com/9router/routevscode.git and read the README end to end — the tool is small enough to fully understand in five minutes. One caveat: the repo is two days old and it proxies Copilot Chat traffic. Skim the README carefully and check the Copilot terms before pointing a work machine at it. What it is: AI Foundry is Microsoft's managed platform for deploying AI models — including open-weight ones like DeepSeek — behind a single secured endpoint. Why now: DeepSeek v4.1 Flash launched this week and pulled 993 points and 565 comments on Hacker News. AI Foundry's model catalog is the fastest Azure-native way to serve it, and the Foundry "What's new" page confirms the newest releases land there first. Tutorial angle: Create an AI Foundry project in the portal, find the newest DeepSeek release in the model catalog (or any fresh open-weight model if it hasn't reached your region), deploy to a serverless endpoint, and call it from a ~20-line script with the Azure AI SDK. Serverless means no GPU quota to manage and per-token billing. Difficulty: Beginner Start today: Open https://ai-foundry.azure.com, go to Model catalog, and search "DeepSeek" to see what's actually available in your region. windows crate What it is: The standard Rust toolchain — rustup, cargo, and rust-analyzer in VS Code — set up for building Windows apps that call native Windows APIs. Why now: "Rust is a tier-1 language at Microsoft" hit 717 points and 501 comments on Hacker News this week (rustfoundation.org). It's the biggest Microsoft dev-platform story in the signal, and it moves Rust-on-Windows from experiment to expectation. Tutorial angle: Install rustup on Windows, wire rust-analyzer into VS Code, scaffold a cargo project, and call a native Windows API through the windows crate — a first Rust-on-Windows app in under 10 minutes. Difficulty: Intermediate Start today: winget install Rustlang.Rustup — one command gets you rustup, cargo, and the default MSVC toolchain target, with the installer walking you through the rest. What it is: The release candidate of the next major .NET version, installable side-by-side with your current SDK for early testing before the November GA. Why now: .NET majors ship every November, and early September is exactly the RC-announcement window (dotnet blog / dotnet/core releases). Starting now buys you two months of soak time before GA. Tutorial angle: The safe RC workflow: install the RC SDK side-by-side, pin existing projects with global.json so nothing silently retargets, retarget one real app with upgrade-assist, and smoke-test it end to end. Difficulty: Intermediate Start today: winget install Microsoft.DotNet.SDK.Preview, then run dotnet --list-sdks — you should see the RC listed next to your stable SDK while your existing projects keep building against the old one. What it is: Microsoft's SDK for building production AI agents and multi-agent workflows in .NET and Python. Why now: Agent tooling dominated this week's GitHub trending — anything2explainer (934★), dream-loop (850★), screenwriting-skills (748★) — and dev.to's top AI article this week is about agent workflows. Meanwhile, microsoft/agent-framework keeps shipping on a frequent release cadence. Tutorial angle: Scaffold a new agent project, register a custom tool, connect it to a model endpoint (Azure OpenAI or Foundry), run it in the debugger, and inspect the tool-call trace to see how the agent decides when to call your tool. That trace is where most "my agent is dumb" debugging actually happens. Difficulty: Intermediate Start today: dotnet new install Microsoft.AgentFramework.Templates (Python folks: pip install agent-framework). If you only do one thing from this post, do the Foundry catalog check. It costs nothing, installs nothing, and answers a question every shop is asking right now: Open https://ai-foundry.azure.com and sign in. Click Model catalog, search DeepSeek. If v4.1 Flash shows up in your region, hit Deploy and pick the serverless option. That's the whole five minutes, and step 3 leaves you with an HTTPS endpoint you can call from anything — run pip install azure-ai-inference while it provisions and you're one short script from your first completion. Best weekend hack: routeVSCODE — a two-day-old repo that swaps models mid-conversation and shows you in the logs which one actually answered. Best career move: Rust on Windows. Tier-1 at Microsoft doesn't mean rewriting everything; it means the windows crate is a safe bet for the next decade. Best prep work: .NET 10 RC behind a global.json pin. Two months of soak time before November GA is free bug-finding. Question for readers: which of these are you trying this weekend — and if you've already put a proxy in front of Copilot Chat, did the logs ever show a different model answering than you expected? Tell me in the comments. If this kind of post is useful, the easiest way to support the work is to: Star / follow on dev.to (you're already here 🙂) Follow on X: @devkingov Reach out for HK-based dev work — .NET / Azure / system integration / IT security: studio.resurrects.co or email devkingov@gmail.com Subscribe to weekly HK tech posts → studio.resurrects.co/blog (one email a week, no spam)

## Key Takeaways

- •This week's Microsoft-ecosystem signal splits into two buckets: AI plumbing you can wire up in an afternoon (model routing, serving open-weight models, agents), and platform shifts with long shelf life (Rust going tier-1, .NET 10 RC)
- •This story was reported by **Dev.to** , covering developments in the**dev** space.
- •AI advancements continue to reshape industries — read the full article on Dev.to for complete coverage.

📖 Continue reading the full article:

[Read Full Article on Dev.to →](https://dev.to/karleeov/5-microsoft-tutorials-copilot-model-swapping-deepseek-on-azure-rust-net-10-rc-ai-agents-5790)
