Its entirely possible to boost productivity and have the LLM do your work for you, but you need to spend lots of time training it and getting it ready. Time that the big players have, but you don't.

Those same players can throw millions of dollars at mistakes. You can't.

Automating those crappy, papercut style tasks that eat away at your soul on a daily basis; THAT'S where LLMs shine and where they are best used.

Stop pushing the "we boosted out productivity by x%," nonsense.

(𝟯/𝟯) Whether you're building REST APIs or AI agents, this episode delivers practical insights.

Full transcript available for those who prefer reading: dotnetcore.show/season-8/bui...

(𝟮/𝟯) Building MCP servers with the C# SDK uses familiar patterns:
app.MapMcp() works like MapGet()
Host locally or in Azure Functions
Define tools, resources, and prompts
Mike literally asked Copilot to generate MCP tools. Used AI to build AI tooling. 🤯

100% this.
We've talked about automating the boring stuff for decades, and a goal of many people is to automate their job away. But as soon as it's given to us, those same folks start screaming about how the sky is falling.
Seems like it's impossible to please ANYBODY these days.

(𝟯/𝟯) Start small: Weekly lunch, psychological safety first, "no stupid questions" rule, end with action items.

The transformation happens when people start talking to each other instead of about each other.

Full guide: rjj-software.co.uk/blog/communi...

(𝟮/𝟯) Communities of Practice aren't just for developers. When you invite everyone, you get:

- Real user insights that prevent wasted development
- Cross-pollination of ideas
- Solutions to problems you didn't know existed
- Innovation through collision

Thank you, again, for being on the show. It was a blast.
I only hope that you, and everyone who listens, can look past the fee flubs of mine that made it past the edit (one instance of "MVP" in place of "MCP" and a few instances of "OpenAI" instead of "OpenAPI").

"I fight for the user"

(𝟰/𝟰) Mike's take: "AI all the way down" is already happening.
Whether you're building REST APIs or exploring AI agents, this episode has practical insights you can implement today.

Full transcript available: dotnetcore.show/season-8/bui...

(𝟯/𝟰) Model Context Protocol (MCP) is how you build tools that LLMs can use. The clever bit? Humans stay in the loop.
Build MCP servers with C#, host in Azure Functions. GitHub already has one that manages PRs and issues.
Mike used Copilot to generate MCP tools. 🤯

(𝟮/𝟰) OpenAPI isn't just documentation. From a single definition, you can generate:

SDKs in multiple languages
CLI tools (Azure CLI does this!)
Tests
Terraform configs
Interactive UIs

Modern ASP .NET Core extracts this from your code automatically.

The entire conversation is worth a listen, and the above quote is just one piece of the 🔥 in the conversation.