Testing my prompt design method through real-world use cases.
For me, these prompts work exceptionally well. Through this experiment, I'm testing whether that's true for others too.
I'm comparing two approaches to prompt creation:
- Baseline: Auto-generated prompts from a structured Jobs-to-be-Done interview
- Crafted: Prompts I design using my method
What I'm learning:
- Does my method work exceptionally well for others too?
- Where can it be improved?
What you get:
- Your prompt in both versions (baseline + crafted)
- Real comparison in your own context
- Your feedback shapes how the method evolves
Copy the Job-to-be-Done interview prompt and use it with your preferred AI (ChatGPT, Claude, Gemini, etc.)
The interview will generate two things:
- Job-to-be-Done Documentation – A structured description of your task
- Baseline Prompt – An auto-generated prompt ready to use
Create a new issue using the "Submit Job-to-be-Done" template and include:
- Your JTBD documentation
- The auto-generated baseline prompt
- Optional: Context about your use case
- Phase 1 Voting: If many submissions come in, upvote (👍) the jobs you'd like to see WSPL versions for
- Phase 2 Voting: Once both versions are available, test them and vote on which works better for your needs
All submitted prompts (both baseline and WSPL versions) are collected in this repository for reference and learning.
- Be honest: Vote based on actual testing, not assumptions
- Be constructive: If something doesn't work, explain why
- Be curious: This is research – there are no wrong observations
- Be respectful: We're all learning together
Do I need to know anything about the method?
Nope! Just test both versions and tell me which works better for you.
Which AI should I use?
Any major LLM works (GPT-5, Claude, Gemini, etc.). Mention which you used if it's relevant.
How long does the interview take?
Usually 10-15 minutes, depending on the complexity of your job.
What if my baseline prompt is already perfect?
Great! That's valuable data. Still submit it so we can learn from it.
Do prompts get maintained?
No – this is an experimental snapshot, not a production tool library. Prompts are optimized for learning.
All prompts submitted to this repository are shared under CC BY 4.0 unless otherwise noted.
Experiment by Martin Haberfellner | neoncode.systems