Design Before Prompt: The DARE Method in Five Minutes

The trap of prompting first

The fastest way to waste an afternoon with an AI coding assistant is to open a blank chat and type "build me the thing." You get code. It compiles. It even looks right. Then you notice it assumed a database you don't use, invented an endpoint nobody asked for, and quietly skipped the one edge case that mattered.

The problem isn't the model. It's that you handed it a decision it couldn't make: what are we actually building? An AI fills gaps with plausible guesses, and plausible guesses compound. Every unstated assumption becomes a line of code you'll rewrite later.

DARE fixes this by refusing to prompt for code until the thinking is done. The name is the method: Design → Blueprint → Tasks → Execute. Four phases, four artifacts, one rule — no code before the plan.

The four phases

Design answers why and what. You write a DESIGN.md that captures the problem, the users, the constraints, and what "done" looks like. No architecture yet — just a shared, unambiguous statement of intent. This is the document you argue about before anyone writes a function.