A Whist scoring app, built mid-game on a ski trip from the rules said out loud.
Whist Scorer went from a spoken set of rules to a working app at the table, mid-game, on a ski trip. It tracks predictions, tricks and running scores, so there's no paper and no arguing about the maths.
It's still live at whistscorer.com. It has since grown logins, saved games and history, and it installs on an iPhone like an app.
The rules came from the one person at the table who actually knew them, so my job was translation: listen, say it back clearly through Wispr Flow and get Replit to build exactly that. After the first game I had five WhatsApp messages of fixes.
The steer that mattered most was hierarchy. It was a shared screen for the whole table, so the score had to be readable from across the room.
Wispr Flow dictation into Replit Design Mode for the first build. Since then: React, Vite, Tailwind and shadcn on the front, Express with session auth on the back, and Postgres through Drizzle for saved games. A domain bought through Replit Domains and a PWA install for iPhones.
Speed. It went from the problem to a spoken brief to a working app in one evening, and the people with the problem used it. That's still the best argument I have for building with AI.
We were on a ski trip, fourteen rounds into a game of Whist, and the admin was killing it. Predictions, results and running scores every round, on paper or in Notes. Fiddly and distracting.
I asked the one person who actually knew the rules to talk through what they wanted, then used Wispr Flow to push that straight into Replit Design Mode. Within minutes we had a first version we could use. Only three of us needed it, so there was no deployment pipeline and no distribution plan. It just had to be open on a screen and readable from across the table.
After the first game I got five WhatsApp messages on what to fix. I added those, plus my own tweaks to what stays visible at all times, because the app had become a shared display for the whole table rather than a private calculator. With the hierarchy right, people stopped asking each other what the score was.
It’s the clearest “have a need, build a solution” story I’ve got. The problem, the spoken brief and the working app all happened in one evening, at the table, for the people who had the problem.
I bought whistscorer.com through Replit Domains and put it live. I’ve no idea how big the market for Whist scoring software is, but the problem is real for anyone who plays the game properly. Next is working out whether it’s worth a real iOS app, possibly at a tiny paid price, once I’ve got an Apple developer account.
