← back
Using Spec-Driven Development for Production Workflows - Erik Hanchett, AWS
Takeaway
Use reviewed specifications as durable guidance for coding agents, with context discipline and human accountability throughout implementation.
Summary
- Spec-driven development creates structured requirements and design documents before implementation to steer coding assistants on complex work.
- Hanchett recommends concise steering files and on-demand skills, warning that both missing context and oversized instruction files can undermine results.
- Humans should iteratively inspect generated requirements, design decisions, and code, with normal review tools supporting that responsibility.
- Kiro packages vibe and spec modes in an IDE and CLI, but the workflow can also be implemented manually by asking another assistant to draft and refine requirements and design documents.
spec-driven-developmentkirocoding-agents
Original description
AI coding assistants are great at completing small tasks or features. However, what do you do when you are working with more complex code bases, and you need to build in-depth features that need upfront planning? This talk explores spec-driven development as a solution to this problem. I'll show you how modern AI coding assistants (like Kiro) can help break down complex tasks into three distinct phases. We'll look at the real-world tradeoffs of this approach, and most importantly and how you can use it in your own projects right away. Speakers: Erik Hanchett (Amazon Web Services): Erik Hanchett is a Senior Developer Advocate at AWS who teaches developers how to build with modern web, AI, and the cloud. X/Twitter: https://x.com/erikch LinkedIn: / erikhanchett GitHub: https://github.com/erikch