Spec-Driven Development
The biggest risk in building with AI is losing control of what the system is supposed to do. Spec-Driven Development keeps the specification in charge.
-
1
Specify
What it must do
-
2
Plan
How we will build it
-
3
Tasks
Small, reviewable steps
-
4
Build
AI-assisted, engineer-owned
-
5
Verify
Checked against the spec
Every change traces back to a specification.
Watch the talk
Beyond Vibe Coding: How Engineering Leaders Scale AI in Regulated Industries
Justin Grammens walks through Spec-Driven Development on a real medical device project.
One source of truth
Expected behavior is written down once, so every AI-assisted output can be checked against it.
AI stays on the rails
Nothing is left to interpretation or hidden in opaque AI output.
Audit-ready by design
Regulators and auditors can trace any feature back to the requirement behind it.
One of the toolkits we use is Spec Kit, GitHub's open-source framework for Spec-Driven Development. We pick the tools that fit your process and regulatory needs. Spec Kit on GitHub →
We speak on Spec-Driven Development at conferences and industry groups. See speaking →