3-pass prompt test before you build with AI
Quick framework I’ve been using before asking AI to build anything: 1. Outcome pass — write exactly what the finished thing should do in one sentence. 2. Constraint pass — list limits: no backend, mobile-friendly, CSV upload, privacy, budget, time, etc. 3. Failure pass — ask: “What will probably break, confuse users, or make this harder than it looks?” Practical takeaway: make the model critique the plan before it writes code. You’ll usually save more time in the planning prompt than in the build prompt. Question: what’s one small AI tool or workflow you want to build, but haven’t started because the scope feels fuzzy?