Memento AI
A private iOS journal that reflects only on your writing.
Personal Project · Designer & Developer · 2023
AI as a slow, therapeutic mirror for self-reflection.
I designed and engineered Memento, a native iOS journaling app utilizing a localized RAG architecture to provide contextual, fully cited AI reflections grounded exclusively in the user's own writing.
TL;DR: Psychological safety is the foundational requirement of any journal.
The product vision was to shift AI from a high-speed productivity utility into a slow, therapeutic mirror. I owned the entire stack, from UI design to the data embedding pipeline, building an experience that feels purely meditative rather than transactional. I rejected push notifications and streaks in favor of a minimalist loop where writing naturally begets more writing.
Challenge: Standard AI integrations turn intimate reflection into surveillance.
A journal requires absolute psychological safety, but standard AI integrations often turn intimate reflection into an intrusive experience. Generic language models generate synthesized, impersonal responses that can distort a writer's true voice, while un-cited AI insights leave users feeling exposed and uneasy.
Approach: The AI could only speak from the user's own words.
I prioritized slow, intentional reflection over rapid-fire transactional mechanics. Instead of artificial rewards or notifications, the intrinsic value of self-discovery drives the experience. I established a strict architectural constraint: the AI could only converse using the user's explicit historical entries, ensuring every interaction felt deeply personal and authentic.
Solution: Writing naturally opens the path for the next entry.
The interface centers on a minimalist editorial loop: after logging thoughts, the AI poses a single contextual follow-up question; tapping it pulls up a clean canvas for the next reflection. A dedicated "Dive Deeper" panel displays the exact historical entries the AI referenced, providing immediate data transparency and source citations.
Outcome: Absolute privacy is a powerful product differentiator.
By treating AI as a quiet mirror rather than an active content generator, the app has validated its core product vision: that intentional architectural constraint builds deeper engagement. Early testing shows that when users feel psychologically safe, their writing frequency and depth increase. A closed beta is underway ahead of public launch.
