AI follows your framing Give an AI agent a bad idea with confidence and it will help you justify it. The same bias applies while building — it improves what you react to, not what matters. Pretty ≠ production-ready Agents keep polishing the UI while your data model, security, error handling, and tests stay invisible. That's how you get something that looks like an app but breaks with real users. First, decide if it's worth building Personal tool or money-making product? If it's the latter, your idea is a hypothesis. Talk to people about the last time the problem happened — don't ask if they'd use your app. Find signal before you build Read competitor reviews, put up a landing page, collect emails, or deliver the result manually. You're not proving the business works — just earning enough confidence to build the next small version. Should it even be a mobile app? Ask where the job naturally happens. If anyone can open a link, start on the web. Only go mobile if you genuinely need the camera, location, notifications, or offline access. Pick the right platform Cross-platform? Use React Native with Expo. Apple-only? SwiftUI. Android-only? Kotlin. Don't build all three at once, and don't wrap a website just to call it an app. Plan before you prompt 'Build my app' is not a plan. Write one page first: who it's for, what problem it solves, and what constraints the agent must work within.