Most weather apps answer the wrong question.
They tell you it's 12°C outside. That's accurate. But whether 12°C means you need a coat depends entirely on who you are — and weather apps don't know you.
That friction is what Samukunai ('Are You Cold?') is built around.
How it works
The mechanic is minimal: tap warm or cold after you check the morning forecast. That's the only input you give the app.
After 10 of those taps, Samukunai calculates your personal cold threshold using 75th/25th percentile statistics on your own feedback data. Once calibrated, the morning push notification shifts from a generic temperature to something like "on the cold side for you" — not a city average.
No machine learning. Nothing trained on external datasets. The personalization is your data — full stop. That was a deliberate choice; it keeps the logic auditable and the scope narrow.
Stack
WeatherKit handles weather data. SwiftData + CloudKit sync feedback across devices. The personalization logic lives in a module we call ComfortEngine — it's statistics, not a model. The UI follows iOS 26 Liquid Glass conventions.
What we got wrong
The cold-start problem is real. Below 10 feedbacks, Samukunai is a worse weather app than the one that came with your phone. No personalization, less information, no widget.
We knew that going in and shipped anyway, because the core loop — tap, accumulate, personalize — needed real users to validate. iOS 26+ only makes the addressable market smaller too. We're not pretending otherwise.
Who it's for
If you run cold, if you overdress every spring, or if you're the person who always asks "do I need a jacket?" — give it 10 taps. That's the buy-in.
Free. No ads. No in-app purchases.
Top comments (0)