Case study
Journal: Designing an AI‑Native Journaling Experience
Role
Solo designer, content designer & full‑stack developer
Tools
Node.js, Express, SQLite, Claude API, Web Speech API
Status
Live in production
Most journaling apps solve the wrong problem. They give you a blank text box and call it done — as if the hard part of journaling is finding a place to type, not finding the words, remembering to sit down, or ever making sense of what you wrote six months ago.
Journal is a privacy‑first journaling app built to solve the problems that actually stop people from keeping the habit: the blank‑page freeze, the entries that pile up unread, and the total absence of a way to talk back to your own history. I designed and built it end to end — architecture, interface, and every line of copy — treating Claude not as a gimmick bolted onto a text editor, but as a content layer woven through the entire writing and reflection experience.
This case study focuses on the content and product decisions behind it, not just the feature list.
The problem, more specifically
Three friction points kill most journaling habits:
The blank page. Nothing to react to, nothing to prompt a first sentence.
Write‑and‑forget. Entries accumulate but are never revisited, so the app captures a life without ever reflecting it back.
No way to ask a question. “What was going on with me last spring?” is a completely reasonable thing to want to know — and no journaling app lets you just ask it.
Every content decision in Journal traces back to one of these three.
Key content design decisions
01
Turning a blank page into a nudge
Rather than a static prompt library, Journal generates context‑aware writing prompts tailored to what someone has actually been writing about. The content design goal here was tone: a prompt that reads like a thoughtful friend noticing a pattern, not a productivity app nagging you to hit a streak. Voice dictation and handwritten‑photo transcription (via Claude’s vision capabilities) exist for the same reason — removing every excuse between a feeling and a captured sentence.
02
Making reflection a built-in feature, not an afterthought
Journal automatically tags mood, builds an incrementally updated thematic mind map connecting entries and recurring ideas, and generates AI‑written weekly, monthly, and yearly summaries. The content challenge wasn’t the AI generation — it was deciding how much interpretation to hand the user versus let them arrive at themselves. Summaries are written to surface patterns rather than assign meaning to them, because a journaling app that tells you how to feel about your own year has overstepped its role.
03
Conversational search: “Ask Your Journal”
The most content‑design‑heavy feature in the app is natural‑language Q&A over past entries. This meant designing not just a search box, but a conversation pattern — how the app phrases an answer when it’s confident, how it hedges when an entry is ambiguous, and how it cites which entries an answer draws from so the user can verify it against their own words rather than trust a black box. This is content design in its purest form: the words are the interface.
04
Privacy as a content commitment, not just an architecture choice
Journal is single‑user, session‑based, with no client‑side analytics or third‑party tracking, and a soft‑delete trash instead of destructive deletes. That’s a technical decision, but it’s also a content governance one: a journaling app is asking someone to hand over their most unguarded writing, so every piece of copy — from the login screen to the delete confirmation — was written to make that trust explicit rather than assumed.
What I’d highlight for a hiring manager or client
End‑to‑end ownership.I made the information‑architecture calls (how entries, search, and the mind map relate to each other), the interaction calls (dictation, backdating, autosave), and the content calls (prompt tone, summary phrasing, Q&A voice) — the same integrated thinking a content designer brings to a cross‑functional product team, just without a team to hand pieces off to.
Restraint as a design choice.The hardest content decisions here were about what the AI shouldn’t say — summaries that observe rather than diagnose, Q&A answers that cite their source rather than assert. That’s the same judgment call regulated‑industry content design requires.
Built and shipped, not just specced. The app is live in production, not a Figma file — every content and UX decision had to survive actually being used.
Tech stack