A €100 order and one button
A friend commissioned a dead-simple dictation app for his elderly client. My shortcut runs into the background-recording wall — and a prototype of the new app came together in seventeen minutes.
By Georg Malahov
Today I talked with a friend, and he told me a story that both amused and inspired me. He has a client, an elderly man: typing is uncomfortable for him, his eyesight is failing — and he keeps trying all sorts of workarounds to use ChatGPT for dictation. What he wants is one specific thing: a dictation app for Android and for Windows. And my friend effectively commissioned the project from me — build a dead-simple app for this grandpa. Roughly speaking, one button: press record, dictate, look at the cleaned-up text, copy it or share it. Budget — €100.
If I build it and wrap it up, it will work very simply, and maybe I'll earn my €100. But that's not the point. This is a real person with a concrete pain, who is already today devising workarounds just to be able to dictate. One more confirmation that the dictation idea is a very good one.
About the shortcut I put together in March I'll say it straight: it works, but not great. The main limitation is the same everywhere — background audio recording. Everyone wants to dictate for a long time, and no workaround — not an HTML page, not a PWA, a web app you install on your phone — gives you background recording, especially on iOS. I barely use my own shortcut myself: translation is awkward there, background recording doesn't work.
So I decided to try again — all the more since new, smarter models have come out. A universal app that works as simply as possible: press, dictate, finish — the text is in the clipboard, go and paste it. And I've already done it. In seventeen minutes. I talked it through with Claude — the AI assistant I write code with: it drew up a plan and then implemented it itself. It doesn't look very pretty or very stylish, but it's very simple: a couple of buttons, you paste in your own model keys — and everything runs right on the device. I haven't tested it across platforms yet, but that's not the main thing.
My thoughts from here go like this. Build it on a PWA after all, but wrap it in Capacitor — a wrapper that turns a web app into a native one for phones — and possibly also in Electron or Tauri, to get a desktop app. Tauri I don't even know yet what it is — I'll figure it out. And if the new version turns out genuinely comfortable and I like it myself, maybe I'll finally stop begrudging the money for an Apple developer account and publish my first app in the App Store. Word is, with today's models that's practically a turnkey job. But it's too early: without traffic, paying for accounts and stores makes no sense.
Today's conversation breathed new life into the dictation project. Everything I usually lack in pet projects came together here: a concrete customer, concrete money and a concrete person who will actually use it. What's left is to get it to the point where the grandpa presses the button — and it simply works for him.