Echo: voice instead of keyboard — and how I built it myself

A thought is always faster than fingers. You catch the exact phrasing — and lose it while you finish typing the sentence.

What hurt

I think out loud better than I type. But the dictation tools at hand kept letting me down: some need the internet and ship your voice to someone else's server; others stumble the moment an English word lands inside a Russian phrase. And for me every other sentence is bilingual.

What I built

Echo — a desktop app that listens and turns speech into text right on your own machine. Offline, no cloud. It doesn't break when you switch Russian↔English mid-sentence, and it runs on the GPU, so it isn't slow. It can even sit in on a meeting and write the brief itself — who said what, and what was agreed.

How — without a team

I don't "know" the Rust and Tauri it's built on. I took an open-source base and from there described to an agent, in words, what I wanted: "let it switch languages," "let it save the note to a file." It wrote the code, I tested it on myself. That's how a tool that would once have needed a whole team grew solo.

I now dictate emails, notes and code — typing took a back seat.

This is one of the four "doors" I wrote about in horizons. If you want to build your own tool for your own task — that's where the course begins.