Each one works in Claude, GPT-5, Kimi, Gemini, Hermes, or whatever model your builder uses under the hood.
I'm starting a small web app. Please write a short, beginner-friendly PRD I can paste into Lovable, Base44, or Replit.
Product name: [Name]
One-line goal: [What it does in 1 sentence]
Audience: [Who it helps]
Top 3 user jobs: [Job 1], [Job 2], [Job 3]
Must-have features (v1): [3–6 bullets]
Nice-to-have (later): [2–4 bullets]
Success in 2 weeks looks like: [e.g., 1 working screen + 3 users test it]
Tone/constraints: keep it non-technical, short sections, action steps, no jargon.
Include these sections exactly:
1) Problem & Outcome
2) Users & Jobs-to-be-Done
3) Scope (v1 vs later)
4) User Flows
5) Screens
6) Data
7) Non-Goals
8) Risks & Assumptions
9) Success Metrics
10) Next 5 Actions
Keep the first version simple. Propose one simple first screen with clear buttons and inputs.
Help me rewrite my app idea as one sentence using this template:
"My app helps [audience] do [useful task] so they can [better outcome]."
Then suggest 3 ways to make it sharper, and ask 3 clarifying questions if anything is still vague.
My rough idea: [paste your idea here]
Demo post — [App name]
1) One-sentence pitch: My app helps [audience] do [task] so they can [outcome].
2) [Attach 1 screenshot of the most important screen]
3) Feedback I want: [specific question — e.g. does the empty state make the next action obvious?]
4) Out of scope right now: [things you don't want feedback on yet]
Hey — I'd love quick feedback on [specific part of my app].
Context: [1 sentence — what it is, who it's for]
What I'm testing: [the one thing you actually want feedback on]
Constraints: [what you can't or won't change right now]
Timebox: [e.g. 2-minute reaction, not a full review]
If you only have 60 seconds, the single most useful thing you could tell me is: [your best guess at what would help].
Just change [one specific thing] in [file or component name].
Constraints:
- Don't touch anything else on this screen.
- Keep existing components, layout, and design tokens.
- Don't introduce new dependencies.
- If you have to refactor more than 1 file, stop and ask first.
Acceptance: [what 'done' looks like in 1 sentence].
Given this app and this stage, recommend a builder and a model. Explain each in 2 sentences. Then suggest 1 backup option.
App: [one sentence]
Stage: [idea / first screen / iterating / launching]
My comfort with code: [none / a little / comfortable]
Budget sensitivity: [low / medium / high]
Need self-hosting? [yes / no]
Consider these options:
- Builders: Lovable, Base44, Replit, Windsurf, Cursor, Bolt.new, v0
- Models: Claude Opus 4.5, Claude Sonnet 4.5, GPT-5, Gemini 2.5 Pro, Kimi K2.5, Hermes 4, Grok Code, DeepSeek V3
Review my v1 app spec against an ethical-builder checklist. For each item, mark PASS / FAIL / N/A and suggest a fix in 1 line if it fails.
Checklist:
- Doesn't collect more user data than it needs
- Asks for consent before collecting personal info
- Doesn't expose API keys or secrets
- Doesn't impersonate a real person or brand
- Has a clear "what happens to my data" answer
- Doesn't enable spam, scraping, or unauthorized access
- Has a way for users to delete their data
- Discloses that AI helped build it (where appropriate)
My app spec: [paste your PRD or summary]
Walk me through a v1 launch readiness check for my app. For each of the 8 items below, ask me a yes/no question. Stop after each one and wait for my answer.
1) Does the first screen load without errors?
2) Is the primary action obvious within 5 seconds?
3) Are error states handled (empty, loading, failure)?
4) Have you tested on mobile width?
5) Is there at least one happy path that works end-to-end?
6) Have 2 real people tried it?
7) Do you have a 1-sentence pitch ready?
8) Do you know what feedback you actually want?
App: [one sentence about your app]