Fretwork

I started taking guitar seriously about three years ago and quickly noticed I was practising the things I could already play. Every practice app I tried was a timer with a streak, which measures the least useful thing — you can spend an hour playing the same song badly. I wanted something that tracked individual pieces and techniques, how long since I'd touched each one, and roughly whether I was getting better at it. I built it over a few months of evenings, mostly on the train.

It solves the gap between practising and improving. You add the things you're working on — a scale pattern, a section of a song, a chord change you keep fumbling — and after each session you log which ones you touched and rate how it went, and the app surfaces what you've been avoiding. That last part is the whole value; the neglected list is uncomfortable and useful. It's for me and about a dozen other people who play as adults with limited time, mostly friends and two of my teacher's other students.

Everything, including the parts I'd rather not claim. The Flutter app, the local database, the session logging flow, the neglect scoring, and the charts. I also did the icon, which is bad. The design I put actual effort into is the session logging screen, because if logging takes longer than about twenty seconds nobody does it after playing — it's now three taps for a typical session, and I rebuilt it twice to get there.

I made it local-first with no account at all, which means no sync between devices and I've decided I'm fine with that. The alternative was running a backend for a dozen people, and every hour spent on that is an hour not spent on the app or on the guitar. I chose a subjective one-to-five rating after each item rather than any kind of audio analysis, which I did prototype — pitch detection on a phone microphone in a room with a television on is unreliable enough that a wrong measurement is worse than an honest guess. The neglect score weights by how recently you touched something and how badly you rated it, so a piece you played yesterday and hated resurfaces faster than one you played last week and nailed.

The scoring was hard to get right, and it took two rewrites. My first version was pure recency, which meant it constantly pushed things I'd deliberately parked. The second added the rating and swung too far the other way, hammering me with the hardest material every session, which is demoralising and not how practice works. What landed was a mix with a cap — the list never shows more than three neglected items, and always includes one thing rated well, so a session doesn't start with a wall of failure. That last rule came from a friend telling me the app made her not want to open it.

I use it four or five times a week and my teacher noticed within about two months that I'd stopped neglecting the technical material, which was the entire goal. Twelve or so other people use it; two of them are consistent, the rest drifted, which seems like a realistic outcome for a practice app. I've had no crashes reported, mostly because there's very little in it to crash. I haven't published it beyond a link because I don't want to support strangers, and I'm aware that's a limitation and not a design choice.

My guitar teacher shaped the model more than I did — the distinction between a piece, a section and a technique is his, and my first version lumped them together in a way he said didn't match how anyone practises. A friend who plays piano was the one who told me the neglect list made her not want to open the app, which led to the cap and the one-good-item rule. Two of my teacher's other students have used it long enough to send real feedback, mostly about the rating scale being too coarse, which I haven't fixed yet.