Weekly Update 22 September 2025 – Teams AI Library, Visual Studio Enhancements, Inside a datacenter

T-1 month: Skype for Business Server 2015 and Skype for Business Server 2019 End of Support

Overview of Teams AI Library

Getting the most out of MCP in Visual Studio with Prompts, Resources, and Sampling

Make Sense of Your Output Window with Copilot

Inside the world’s most powerful AI datacenter

You can also listen to the audio-only version: Thoughtstuff Podcast – Tom Morgan on Teams Dev: Weekly Update 22 September 2025

Find all my videos at thoughtstuff.co.uk/videos. You can also subscribe to the audio-only version of these videos, either via iTunes, Spotify or your own podcasting tool.

Transcript (AI-generated)

Hello and welcome to another weekly update. Hope you’re doing well.

Skype for Business End of Support

A real blast from the past – and probably the last time I’ll mention it: Skype for Business Server 2015 and 2019 are ending their support in one month’s time. This deadline has been coming for a long while. October 14th, 2025 is the end for:

• Skype for Business Server 2015
• Skype for Business Server 2019
• Skype for Business 2016 client
• Skype for Business 2019 client

If you’re still on these platforms, Microsoft recommends migrating to Microsoft Teams or the Skype for Business Server Subscription Edition. Farewell to an era—we once talked about Skype development weekly!

Teams AI Library Now GA

The Teams AI Library, which has been in beta, is now generally available for C# and TypeScript—with Python still in preview. This SDK provides a declarative and streamlined way to create AI-powered agents and apps specifically designed for Microsoft Teams.

It’s a bit confusing with overlapping tools, like the agents SDK and declarative agents, but the Teams AI Library offers out-of-the-box scaffolding and options, including support for messaging extensions and more. A great option for pro-code developers focusing on the Teams ecosystem.

Visual Studio Embraces AI Context with MCP

Visual Studio has added support for MCP (Managed Common Prompt) prompts and resources, significantly enhancing Copilot’s potential through custom plugin servers. For example:

• Install an Azure DevOps MCP server to give Copilot access to your repos and pipelines.
• Install a Figma MCP server to let Copilot understand your design assets.

This creates a more seamless and enriched developer experience, reducing the cognitive and procedural load of context-sharing across tools.

Visual Studio Output Window + Copilot

Debugging just got smarter. Visual Studio Copilot can now read the output window during your debug sessions. That means Copilot can analyze logs in real-time and help spot issues when you’re stuck, even if you don’t know exactly what to look for. It’s like finding a needle in the haystack—except now Copilot is helping you search with a magnet.

Inside Microsoft’s AI Datacenters

Scott Guthrie published an excellent blog post titled “Inside the World’s Most Powerful AI Data Center”. It highlights Microsoft’s newly announced AI data center in Wisconsin, with others in Norway and the UK. Some stats:

• Each rack: 72 Nvidia Blackwell GPUs
• 1.8 TB/s of GPU-to-GPU bandwidth
• 14 TB pooled memory per GPU rack
• Functions as one unified accelerator—not individual chips

This post is filled with great visuals and dives into Microsoft’s cooling systems, scalability architecture, and approach to hyperscale AI infrastructure. Well worth a read.

Upcoming Talk at Collab Days

This Wednesday, I’ll be speaking at Collab Days at Bletchley Park (The National Museum of Computing, UK). I’ll share some hands-on experience building declarative agents for Microsoft 365 Copilot, including tips and lessons learned. It’s a packed day of fantastic speakers—hope to see you there!

Whatever you’re working on this week, have a great one – and I will speak to you again next time.