Note · November 2026 · 4 min read

10,000 steps shouldn't take 1,000 taps

I built a habits app for myself. The lessons that mattered had nothing to do with features and everything to do with friction.

I added a walking goal to my own app: 10,000 steps a day. The "+" button added 10. Logging an ordinary day meant pressing it about 1,000 times.

I didn't notice while building it. I noticed the first evening I used it. That's the whole lesson, really, but here are the specific ones.

01Count taps, not features

The fix wasn't clever. The step size now scales with the goal, so any goal takes about ten taps, in round numbers (+1,000 for steps, +1 for glasses of water). And you can tap the number and type it, because sometimes you just know it's 7,400.

Since then, every change gets the same question: how many taps does the common case take now? Ticking a habit is one. If something takes more, it needs a reason.

The best habit app is the one that's faster than skipping the habit.

02Feel instant, recover honestly

A tick shows up the moment you tap, before the server answers. If saving fails, the tick rolls back and the app says so. Waiting on a spinner to log a glass of water is how people stop logging water. Saying "saved" when it didn't save is worse.

03"Not today" is not a miss

Sick days, travel days and rest days happen. A skip with a reason keeps the streak alive and doesn't count against you. The point of a streak is to keep you going, and one that punishes honesty ends up teaching you to lie to it.

04Put things where you'll look

Habits are grouped by morning, afternoon and evening, so the screen matches the day. Overdue todos jump to the top on their own. Finished ones clear in one tap, with an undo, because "delete all" without an undo is a trap.

05Two places, one truth

My AI assistant reads the same habits to write a weekly "what's actually working" note. That only works if the phone and the assistant agree on what a streak is. Same rules, written twice (once in JavaScript, once in Python), and a test that runs both on the same data and fails if they ever disagree.

It's a plain installable web app: no framework, no build step. Each person's data stays in their own account, locked to them at the database level.

06Use it before you ship it

Every bug above came from using the app, not from planning it. I built it for one user, me, and that user files very specific complaints.

Try the real app with demo data →

Same app I use every day, running on invented habits so nothing personal leaves my phone. Want to use it for real? Say hello.