The Dictation Shortcut My Parents Liked

By Georg Malahov

I haven't written anything in a while — so here's a small experiment that made me happier than the big projects. A couple of weekends ago, on March 14–15, I tested an idea for AI dictation: I built an iPhone shortcut — a mini-script that lives right in the system, no separate app needed. Tap, talk — and get clean text, without the "ums" and broken-off sentences.

I set it up for my parents. It really clicked with them. It works well, genuinely well. What I especially liked is the distribution mechanic: a shortcut can be shared directly via a link — the person opens it, and everything just works. No store, no installation, no three screens of instructions.

This week I set up the same dictation for myself on my MacBook. And that turned out to be truly useful. I talk a lot with Claude these days — the AI assistant I write code with — and English works better with it. Now I put thoughts into words however feels comfortable: in Russian, in English, mixed — and out comes a clear, structured message in English. The dictation translates into any target language, works fast, and the barrier of "sit down and formulate it" disappears entirely: I speak on the go — and the thought is already sitting there as finished text, ready to send.

The contrast with my main project is noticeable. Call Copilot — my browser extension for transcribing calls — is quiet right now: no updates, no payments. I ran ads for a day, because Monday is usually the most effective day for installs — and nothing changed. Some people install it, some use it, some remove it; I don't see demand for more yet. And here's a shortcut assembled over a weekend — and it lives: my parents use it, I use it myself. That says nothing about demand — people close to you don't count, as I wrote myself about payments in Call Copilot.

So I'm thinking about how to add dictation to my product. Call Copilot already has payments, tokens and all the plumbing around them — dictation would slot in very neatly: shared infrastructure, more value.

But there is still work to do there. All these small improvements are deceptive: I know how to build them, and it's still not something I can just knock out. Each one takes a context switch, getting back into the code, and hours at the desk. I keep not sitting down to it — progress is slow, and for now I'm giving myself a short break.

Meanwhile the dictation keeps working with no further tuning: on my parents' phone, on my laptop. The test is simple — will they still be using it in a month, once the novelty wears off. Beyond that there are no conclusions: two people from my own family are not a demand check, only a reason to take the idea as far as a product and test it on strangers.