🤖 Blazor + GPT-5 Agent Mode: Does it actually work?

Published 9 months ago • 1 min read

Hey Reader,

I just published a new video where I put GPT-5’s Agent Mode to the test. 🚀

Instead of asking it for something simple, I went all in:

A .NET 9 console app calculator
A Web API with endpoints and docs
A Blazor Server app with extra counter buttons
And even a Blazor app that calls a public weather API

🎥 Watch it now:

In the video, I didn’t just type “make me an app” and hope for the best. I gave GPT-5 very specific prompts, like:

“Please create a .NET 9 console application that works like a simple calculator providing addition, subtraction, multiplication, and division.”
“Build a .NET 9 Web API with an endpoint that says hello GPT-5 agents. Add Scalar so I can test the API easily.”
“Create a Blazor Server app that extends the default counter example with a +5 button and a reset button.”
“Make a Blazor Web App using the interactive server render mode that calls a public weather API and shows the current temperature.”

This was a perfect reminder: AI can speed things up, but you still need to know what’s happening under the hood.

If you’re learning .NET, Blazor, or Web APIs, GPT-5 Agent Mode can be a great coding buddy.

But it’s like using a calculator in school - you still need to know the basics first. Otherwise, when something breaks (and it will), you won’t know how to fix it.

That’s why in my upcoming course, .NET with AI, I’ll show you how to work with AI tools the right way - so you stay in control of your code.

👉 Join the waiting list here to be the first to know when it’s live.

Take care,

Patrick

P.S. Want to go deeper into real-world .NET and Blazor projects? Join me inside the .NET Web Academy. Check it out here.

I built a Blazor app in 30 minutes (here's the part that matters)

Hey friend, This week, I built a working Blazor app from scratch in about 30 minutes. But that's not really the point. The point is how. Not by throwing a prompt at Claude and hoping for the best. But by thinking about architecture first, defining instructions and skills for the AI upfront, and then letting it implement while you stay in control and review the output. That's the difference between AI doing your job and AI doing the boring parts of your job. The video walks through exactly...

3 days ago • 1 min read

I published a video a year ago saying AI coding wasn't the future. I was wrong.

Hey friend, A year ago, I said AI coding tools weren't the future. I want to correct that. The distinction I missed: there's a real difference between vibe coding (copy-paste until it compiles, no understanding required) and skilled AI-assisted development (you own the architecture, AI implements it, you review the output). I was criticizing the first and accidentally dismissing the second. In 2026, skilled AI-assisted development is the professional standard. Not using these tools is now a...

10 days ago • 1 min read

We debated a delete button for 5 days

Hey friend, We once spent five days in meetings debating a single delete button. Whiteboard sessions. A compliance guy nobody invited. A product manager who scrolled Slack for what felt like forever. And still no decision. Until Tom walked in on Monday morning, asked one question, and shipped it that afternoon. In this video, I tell the full story and walk you through the four questions that should end this debate on any team, in any codebase. Take care, Patrick P.S. We go deeper on exactly...

17 days ago • 1 min read