Spec-driven development (SDD) is a workflow where detailed specifications are written before any code, with AI agents implementing from those specs. After hands-on testing, the author finds SDD genuinely useful for complex features, team projects, legacy codebases, and async agent runs — but overkill for small tasks and no substitute for developer expertise. The key distinction from Waterfall is the feedback loop: minutes instead of months. Tools like GitHub Spec Kit (72k stars), AWS Kiro, and Tessl are emerging around this workflow. SDD is framed as a subset of the broader discipline of context engineering. The practical recommendation is to use SDD selectively based on task complexity, not as a universal methodology.
Table of contents
What Spec-Driven Development Actually IsThe Waterfall ComparisonWhat I Actually TriedWhere Spec-Driven Development WinsWhere It Falls ApartThe Tools LandscapeContext Engineering: The Bigger PictureMy Honest Take: It Depends (But Here Is When)The Real Question Nobody Is AskingGetting Started If You Want to Try ItSort: