Code Like a PM: Using PRDs to Supercharge AI Coding in Cursor

When most devs talk about AI-assisted coding, they focus on prompts. But the real magic happens before you ever touch the keyboard.

With Goru, our goal-setting app built in React Native, the first thing we did wasn’t code—it was write a PRD. Then we handed it off to Cursor and asked the AI to generate a full development plan.

What we got back was more than helpful—it became our blueprint for a 6-week build.

Step 1: Writing the PRD

Before we even opened Cursor, we wrote a lightweight but detailed Product Requirements Document. It outlined everything from the user journey to the technical architecture.

We broke the app downinto clear experiences:

• A chat-based SMART goal creation flow
• AI-driven milestone scheduling
• A weekly timeline to visualize progress
• A journaling system with streaks and reflection
• A stats dashboard powered by AI insights
  

Then we added tech stack decisions, database schema, key UX flows, and UI principles—like keeping things minimal, clean, and energy-aware.

Step 2: Prompting Cursor with the PRD

Here’s where the magic kicked in.

We dropped the entire PRD into Cursor and prompted:

“Based on this spec, generate a development plan for building this app in React Native using Supabase, OpenAI, and Expo. Break it down by phases and tasks, with a week-by-week structure.”

Within seconds, Cursor gave us a full development outline:

The AI-Generated Development Plan (Highlights)

Phase 1: Project Setup & Infra

• Initialize Expo project with TypeScript
• Set up ESLint, Prettier, and Git
• Configure Supabase and OpenAI clients

Phase 2: Timeline UI

• Create horizontal month selector
• Build vertical weekly timeline
• Add view toggle + empty state

Phase 3: Chat-Based Goal Creation

• Build chat interface components
• Integrate OpenAI for goal analysis
• Add due date picker + busyness selector

Phase 4: Milestone Management

• Create milestone cards with CRUD actions
• Implement batch add to timeline
• Enable drag-and-drop reordering

Phase 5: Journaling & Progress

• Build journaling chat flow
• Add daily reflection logging
• Track streaks and challenges

Phase 6: Polish & Deploy

• Add loading states, error handling
• Run tests and deploy to App Store

Step 3: Letting AI Code with Context

With the dev plan in place, we could prompt Cursor  in context:

“Generate a ChatInterface component for React Native Paper that supports async AI responses, suggestion buttons, and a typing indicator.”

“Scaffold the milestone schema in Supabase and write the TypeScript types.”

Each prompt was sharper, because the AI understood the project.

Iterate with the Same PRD in ChatGPT

When I wanted to switch tools—for example, to ask ChatGPT to generate a better onboarding flow or a naming convention for tables—I could reuse the same PRD. This let me:

• Stay aligned across tools
• Reuse architecture decisions
• Avoid re-explaining everything
  

The PRD became our single source of truth.

Bonus: Turn It Into a Living Document

As we progressed, we updated the plan:

• Checked off completed phases
• Broke large features into GitHub issues
• Used the schema to seed database migrations
• Linked UI components back to Figma designs
  

It wasn’t static—it was our AI-enabled dev ops system.

💡 Final Thoughts: PRD + AI = Velocity

The biggest myth about AI coding is that it removes the need for planning. The opposite is true.

AI amplifies your thinking—but it can’t think for you. That’s why writing a PRD, even if it’s scrappy, sets you up to prompt smarter, build faster, and stay aligned with your product vision.

Cursor didn’t just write our code.

It followed our blueprint.

And that blueprint? That came from thinking like a PM first.

‍