.NET Rocks! GitHub Spec Kit with Den Delimarsky
18 snips
Oct 30, 2025 Den Delimarsky, a principal product engineer at Microsoft, dives into the GitHub Spec Kit and its role in specification-driven development with LLMs. He highlights how granular specifications maximize coding efficiency and serve as guardrails for AI models. Den explains the importance of detaching requirements from technical stacks and chunking tasks for successful AI-driven outcomes. He also discusses the benefits of generating tests first and maintaining developer oversight to ensure quality and security in code. It's a fresh perspective on enhancing software development with AI!
AI Snips
Chapters
Transcript
Episode notes
Early Coding Memories With VB6
- Den remembered coding in Visual Basic 6 and missing modern UI styling updates.
- He nostalgically described dropping form designers and progress bars while lamenting lack of sign-in requirements.
Specs Separate Requirements From Tech Stack
- Spec-driven development detaches requirements from implementation so models can regenerate code across stacks.
- Denailarsky says specs let you generate multiple tech variants from the same requirements to compare options.
Granularize Specs And Use Clarify Prompts
- Spend time granularizing specs to avoid under-specification and AI guessing.
- Use the SpecKit "Clarify" prompt to have the model probe your blind spots before coding.