unknown

Spec-Driven Development (SDD) involves writing explicit specifications in Markdown before coding, allowing AI agents to generate and validate software automatically. While this approach can improve code generation with AI tools, it risks reverting to waterfall methodology if specs become too extensive. The recommended practice is breaking specifications into smaller, executable deliverables that provide fast feedback loops and align with iterative development cycles, making AI agents more effective with concrete, focused context.

Spec-Driven Development: Is it a good idea?

Marcos Lobo

<p>Agree. For AI agents maybe we have to do it. But I’ve worked with too much waterfall, and worse, leaders using the word “agile” all the time, while 100% thinking, executing and doing in waterfall logic the whole time</p>
<p>Organisations often sacrifice so much release power and progress for stupid bureaucracy reasons, like being forced to plan a whole project exactly ahead, to be able to calculate the man hours correctly. Ironically, as expected, the calculations are very wrong 80% of the time anyway, but they keep doing it, because the finance department needs estimations in a waterfall style model. It’s insanity.</p>
