GitHub’s Post

Tested this creating an application with Angular SPA fontend, dotnet backend using Azure and Azure SQL. I mainly used GPT 5 anx Claude 4.0. Not eveyhing worked 100% but after some (3-4 hours) tweaking I got up a working app with full infrastructure, ci/cd, uint- and integration tests and more. Not bad at all.

Really interesting... spec-driven dev with Markdown is a clever idea. How does Copilot handle complex state management in this setup though?

I genuinely like programming. I enjoy experimenting, learning and writing actual code in forms I consider elegant. I find it a nice creative outlet. I don't like prompting. I don't enjoy guessing the correct wording and tone that will make the LLM do what i'm thinking. I don't like the resulting code. This might be the first time i have a truly boomer opinion.

If you use Basic Vibe Coding, the AI is forced to fill the gaps with guessing. If tasks gets more complex, with tons of guessing. So for sure the way out is to profide more details. Spec driven development sounds logical. Pretty sure an additional key aspect is to force the LLM to fix its own errors, meaning to use test driven development. What‘s your experience about?

Good to know. Thanks for sharing

This is great new but how reliable is this in production environments?

Vibe coding on steroids 🚀

This is good capability using md. Until now I was using Pytest or Playwright MCP to create and execute my tests. Will look at it and check what all usecases it can enable in prod.

New cert: Markdown Architect. 🤣

That’s awesome! I really see huge potential in complete markdown-based approaches — everyone can write markdown, and it makes building and explaining ideas so much more approachable.