Three of the twelve install on a device rather than open in a tab. Android, cross-platform, and desktop.
A phone is a slower computer with a worse connection and a person holding it with one thumb. Everything that is merely inelegant on a laptop is disqualifying there, which is why we build the risky part on a real device in week one rather than in a simulator in week six.
Cross-platform when the product is mostly screens and sync, native when it is not — DroidDoodle runs a language model on the handset itself, and that is not a thing you reach through a bridge.
Whichever the product actually needs, and the question is usually settled by one feature rather than by preference. If it is screens, forms and sync, Expo and React Native get it onto both platforms with one team. If it needs the hardware — on-device inference, a custom canvas, anything with a frame budget — it is native, and we say so before you have paid for the other answer.
Yes: signing, listings, review notes, staged rollout. It is the least interesting part of the work and the part most likely to add two weeks if nobody planned for it, so we plan for it. Worth knowing that a first review can take days and a rejection can take a week.
No, and sometimes you should not. DiscVault ships as an apk and a desktop binary alongside its CLI and its website, because its audience already sideloads and a store listing would have bought nothing. A sideloaded app can still update itself — there is a post on this site about exactly how.
It is a data-model decision, not a feature you add at the end. Who owns the truth, what happens to an edit made on a train, what the second device sees when it comes back. We answer those three before drawing a screen, because every interface question downstream falls out of them.
We have. DroidDoodle drives a drawing canvas from a model running on the handset — no round trip, no key, no per-token bill. It is a real trade: you get privacy and latency, you pay in binary size, battery and a much smaller model. It is worth it more often than people assume and not always.
If you want us to. An app is the one thing we build that keeps needing us — OS releases break things a website never has to care about, and a build that shipped fine in June can be rejected in September. We would rather agree a small ongoing arrangement than be a phone number you call when the store emails you.