Sheet 02 · Web app, live · 2026
Founders often ask friends "would you use this?" and hear a polite yes. This app turns one sentence about an idea into research that can prove it wrong.
"Fall in love with the problem, not the solution." The rule the app is built on
Plan
The first screen asks a single question: what are you building? Before any research, the app rewrites the idea from the customer's side, so the founder starts from the problem people have, not the feature they want to build.
There is no account. Projects are saved in the browser only, and "Restore from a backup" brings them back. People can try it without handing over an email.
Plan
Then the founder picks a method. The four cards follow the order I learned in my Google UX course: start at your desk, check the money, count with a survey, and understand why with interviews. Each card says what it costs in people and time, so nobody runs interviews when a desk search would answer the question.
Research
The audit explains direct and indirect competitors in plain words, with an example, then fills a table with two of each: what they offer, who it's for, price, strengths and weaknesses. Every cell can be changed, and the whole audit copies out in one click.
Research
The survey screen starts with the decision, not the questions: what you'll decide, who should answer, when to send it, how many answers to aim for, how long it takes, and how answers stay private. Then come the questions, grouped by what they test.
Bias check, live while you edit
Would you use an app that reminds you to drink water?
Future question · yes/no question
Tell me about the last time you forgot to drink water during a workday.
Click any line to change the wording. If the new wording leads people to an answer, the app flags it while you type.
General notes
I wrote the product spec, designed the flow and built it in plain JavaScript with Claude Code, deployed on Vercel. Every screen asks one thing at a time, and the copy stays in plain words, because the people using it are founders, not researchers.
I made it while working through my Google UX Design certificate courses. Each screen puts one lesson into practice: write a research plan first, ask about the past instead of the future, and pick the cheapest method that answers the question.
I also turned my course notes into a Claude skill: my own research method, written down so an AI assistant follows it step by step, from plan to research to synthesis. I built it from my notes for course 4, on conducting UX research, and from my dating-app survey. I used it for this app, and I keep using it for every study since.
ux-research-helper · SKILL.md
Plan → research → synthesize. Always say which step we're on.
Before any data comes in
Write 2 or 3 "I'm wrong if…" rules with numbers, and check the results against them first.