Pineflow · Idea and PRD

searchcomplaints12 monthssummaryYour ideaApp Store + PlayReddit and XGoogle TrendsAI with web searchdocs/research.md

Plan

Idea and PRD

Check that people really have the problem, then write one page that says what you'll build.

The AI can do most of the work here, but the idea is yours. Know the people, their pain and why it matters better than anyone. That's the difference between a founder and someone just vibing. With experience you hand more to the AI, but you still decide.

2-3 h

1

Find a real problem

Paul Graham: How to get startup ideas ↗

Start from a problem people already complain about, not a clever idea. Something that bugs you every week, an app people hate, a job they still do by hand.

  1. Pick your problem

    Step 1 of 3
    1. 1
      No idea yet? Search Reddit, TikTok and the rest for wishes, not ideas. Add a topic you know, and try "I wish there was an app", "I'd pay for" or:Search"is there an app that"
    2. 2
      Or read the 1- and 2-star reviews of an app you use. Note what people keep asking for.
    3. 3
      Write down three problems you found, or that annoy you. Pick a problem you have yourself, or know well. You'll make better calls on every screen.
    4. 4
      Pick the one you understand best. A boring problem people pay to solve beats an exciting one nobody pays for.

    Done when You can say your problem in one sentence. Example: lifters who train alone never know when they're ready to try a new max.

    Stuck? Watch someone do this stepFind your next app idea in 10,000 bad reviewsCode with Beto
2

Check the demand

Official Google Trends help ↗

Look for proof that people want it: rival apps, searches and complaints. Ten rival apps is good news, it means there's a market. A friend saying 'nice idea' is not proof.

  1. Research the demand

    Step 2 of 3
    1. 1
      Create docs/research.md. Ask Claude Code:
      PromptCreate docs/research.md with five headings: Kill criteria, Searches, Complaints, Competitors, What we do differently.
    2. 2
      Write your stop rule first, before you look. For example: if I can't find 10 real complaints in 1 hour, I stop.
    3. 3
      Search the App Store and Google Play like a user would, for example gym log. Write how many apps you find.

      ✓ Dozens of apps means a real market. None at all is a red flag.

    4. 4
      Check Google Trends ↗ over 12 months: rising, flat or one spike?
    5. 5
      Find 3 rival apps and their weak spots. Their complaints are your feature list. Ask AI, with web search on, and paste the answer under Competitors:
      PromptMy app idea: [one sentence]. Find the 3 closest rival apps in the App Store and Google Play. For each, give one strength, one gap, and the 3 most common complaints in its reviews, with a real quote each. Then tell me which complaint none of them has fixed.
    6. 6
      Write one sentence under What we do differently: the thing none of the three does. Apple asks the same at App Review, since it rejects apps that look like copies.

      ✓ Example: ranks each lift against lifters of the same bodyweight, the others only log weights.

    7. 7
      Talk to 5 people who have the problem. Find where they hang out, Reddit, TikTok, Instagram or X, and send a message. Ask about the problem, not your app.

    Done when docs/research.md has real quotes, 3 rivals, what you do differently and a stop rule. The idea holds up, or you just saved months.

    Stuck? Watch someone do this stepHow to validate your app idea even with zero audiencepaulnocode
    research.mdTemplate to fill in
3

Write the PRD

Shape Up, the free book on scoping work ↗

A PRD is one page that says what you're building: who it's for, what it does and what it won't do. Your agent reads it every session, so every gap becomes a guess.

  1. Write your PRD

    Step 3 of 3
    1. 1
      Create docs/PRD.md. Ask Claude Code:
      PromptCreate docs/PRD.md with these headings: User, Goal, MVP, Out of scope, Success.
    2. 2
      User: one real person in one real situation.

      ✓ Example: a lifter who trains alone three times a week and wants to know when to go for a new max.

    3. 3
      Goal: what the app does, in one sentence. Can't say it in one sentence? You're not ready to build yet.

      ✓ Example: log a lift and see when you're ready for a new max.

    4. 4
      MVP: 5 features or fewer, each one a screen on the phone. Cut until it hurts. The first version does one thing well.
    5. 5
      Out of scope: at least 3 things you won't build. If it isn't written down, the agent builds it anyway.
    6. 6
      Success: one test per feature that you can check by hand.

      ✓ Example: a new user logs a lift in under a minute, on their phone.

    7. 7
      Check it: upload docs/PRD.md to ChatGPT, Gemini or Claude on the web, in a new chat, and ask:
      PromptRead this PRD. In your own words, what app should I build, and for whom?

      ✓ Wrong somewhere? Fix that part of the PRD and ask again.

    8. 8
      Save a snapshot of your plan:
      Terminal
      git add .
git commit -m "Add research and PRD"
git push

    Done when A new chat reads your PRD and describes your app back to you, correctly.

    Stuck? Watch someone do this stepHow to write a great PRD, step by stepClickUp
    PRD.mdTemplate to fill in