Building Morrow Self, a privacy-first wellness app

By Malik Chohra

We designed and shipped Morrow Self, a privacy-first wellness app with the tagline "live your life like a startup." Daily wins, structured journaling, and an on-device AI that reads your journal without the journal ever leaving the phone. This is our wellness app development case study: what we built, the privacy decision at the center of it, and why it is built entirely on our own stack. It is live at morrowself.app.

What is Morrow Self?

A self-mastery app that borrows startup language for personal life. You log daily wins, the way a founder logs progress. You journal in a structured format instead of a blank box. And an on-device AI helps you reflect: surfacing patterns, prompting follow-ups, drawing connections across entries. The premise is that the discipline that makes a startup work also makes a life work, so the app gives you the same instrumentation.

Why does a privacy-first wellness app keep the journal on device?

Because a journal is the most sensitive data a person will put in an app, and the standard pattern of shipping it to a cloud model is the wrong default for that content. So we ran the AI locally. Morrow Self uses an on-device model (local Phi-3) to read and reflect on entries. The journal stays on the phone. Nothing in the reflection loop requires uploading what you wrote.

This is a design decision before it is a technical one. Privacy-first means the private data physically does not leave the device for the core AI feature. It is not a setting buried in a menu, it is the architecture. The trade-off is that an on-device model is smaller than a frontier cloud model, but for journaling reflection that is an acceptable trade and the privacy is worth more to this user.

How is the onboarding adaptive?

Morrow Self uses AI dynamic onboarding: fixed base questions, then AI-tailored follow-ups based on the answers. We did not hand-code every branch. The model decides what to ask next and the app renders it as a real native component. That is our Wire RN generative UI stack doing the work. Morrow Self is where we dogfood it in a real consumer product.

What is it built on?

Our own boilerplate, end to end. Morrow Self runs on the AI Pro tier of AI Mobile Launcher: Expo and React Native, strict TypeScript, Restyle tokens, Redux Toolkit and RTK Query, MMKV, Supabase, RevenueCat for the subscription. The on-device LLM runs through llama.rn. The adaptive onboarding runs through Wire RN.

Building it on our own stack is the point. It is the proof that the boilerplate ships a real product, and it is where we find the rough edges before a client does. When the on-device model or the generative onboarding breaks, it breaks in our app first.

FAQ

What makes a wellness app privacy-first?

The test is where the sensitive data goes. A privacy-first wellness app keeps the private content, the journal, on the device rather than shipping it to a cloud model. Morrow Self runs its reflection AI on-device with a local Phi-3 model, so the journal never leaves the phone for the core feature.

Does on-device AI work well enough for a real app?

For the right tasks, yes. Reflection, pattern-spotting, and follow-up prompts on journal text are well within a small on-device model's range. You would route to a cloud model only for heavy reasoning, which the journaling loop does not need. The privacy win is worth the slightly smaller model.

Can CasaInnov build a wellness or health app with on-device AI?

Yes, Morrow Self is the proof. We build privacy-first mobile apps with on-device inference as a standard option, not a special request. See our AI mobile development service to start.

Why build it on your own boilerplate?

Speed and proof. The plumbing was already done, so the budget went to the product. And shipping a real consumer app on the same stack we sell is how we catch problems before clients do.