Case study · Product design

Between the Lines

A journal that reflects your own patterns back to you, in your own words and never as a diagnosis.

RoleProduct & UX Designer
ScopeDesign and build, solo
ContextAI Product Masterclass · Nueve Design
Year2026
BuildReact · TypeScript · Vercel
An evening with the journal open: the week's entries on the left, the Insights card reading between them. The running build, composited into a photo.
An evening with the journal open: the week's entries on the left, the Insights card reading between them. The running build, composited into a photo.
At a glance
Problem
Journaling tools sit at two extremes, a blank page or a mood emoji. More and more people take their inner life to a general chatbot instead: in a 2025 US survey, ~49% of people with an ongoing mental-health condition used one for support. The tool they confide in only agrees with them.
My role
Product & UX Designer, solo design and build. During the AI Product Masterclass (Nueve Design, 2026, five-week intensive) I designed the journal and built it into a working product: writing separated from rating, mood as a two-axis grid, dictation rebuilt around Whisper, insights computed on the device, and a Companion chat grounded in the journal.
Value
  • Shipped live: a working product you can open and try, not a clickable mockup
  • Write first, rate later: writing and judging are separate steps on one distraction-free screen
  • On-device by default: text leaves the device only on an explicit “Reflect deeper”
Try it
Try it live
Stack
ViteReactTypeScriptTailwindshadcn/uiTiptapVercel functionsxAI GrokGroq Whisper
Solo course project · working demo · no real users yet
01 · Origin

A course exercise became a product with a point of view

The starting point was a course project, suggested by a mentor as a way to learn a more complex, technical toolchain. From the first sketch I steered it toward something I believed in, adding the features it actually needed. What could have stayed a throwaway exercise grew into a working tool for self-knowledge.

02 · The problem

People are already treating a chatbot as a therapist that only agrees with them

Therapy is widely seen as expensive or hard to trust, so people improvise. More and more of them bring their mental health to a general-purpose chatbot, detached from any wider context. The risk is quiet. These tools tend to affirm rather than ground or challenge, and they can state things with unearned confidence.

The quiet risk, sketched: a chat that nods along, states things confidently, and knows nothing about the person's week.
The quiet risk, sketched: a chat that nods along, states things confidently, and knows nothing about the person's week.
US survey of people with an ongoing mental health condition. Sentio University, 2025.
~49%
use a general-purpose language model for mental-health support
03 · The landscape

Most journaling apps are either an empty text box or a mood-emoji tracker

The tools meant for reflection sit at two unhelpful extremes. The blank page gives you nowhere to start. The mood tracker shrinks a whole day to one tapped face and never says anything back. Neither closes the loop between writing and understanding, which is the reason to keep a journal at all.

04 · Context

In a country that funds mental health heavily, it is still the costliest condition

Norway invests in mental health more than almost anywhere, and it remains the single costliest condition there, close to three percent of GDP. Care exists, but access is uneven, and the everyday, low-stakes layer is thin. There is room, before and beside professional care, for something inexpensive and private that helps a person notice their own patterns.

Public Norwegian sources. Illustrative, 2026.
~3%
of GDP: mental disorders, the single costliest condition in Norway
~20%
of new depression diagnoses referred to specialised care
05 · Diagnosis

The first build made the same mistake it set out to fix

The first version worked, but it had never really been designed. Writing, mood rating and tagging were crammed onto one screen, with no focus and no payoff for the effort. It was the exact shape of the problem, staring back.

v1 from the first build, audited finding by finding: writing, rating and tagging crammed onto one screen, with demo text.
v1 from the first build, audited finding by finding: writing, rating and tagging crammed onto one screen, with demo text.
06 · Guiding policy

One rule held the scope: an adjunct, never a replacement

Every decision ran through a single commitment. Complement care and self-knowledge, and never impersonate a therapist. Each reflection carries the same line: patterns in your own words, not a diagnosis.

The evidence base was the partner

Rather than invent its own claims, the design defers to validated psychology and cites it.

  • Russell’s circumplex model of affect, for mood.
  • Pennebaker’s expressive-writing research, for the language reflection.
  • CBT vocabulary, for the prompts.
  • Low-validity personality quizzes refused, such as MBTI and the Enneagram.
The evidence base as the design partner: borrowed psychology, one refused shortcut, and the line that runs under everything.
The evidence base as the design partner: borrowed psychology, one refused shortcut, and the line that runs under everything.
07 · The build

A focused two-step flow: write first, rate later

Writing and judging are different mental modes, so they were separated. Step one is a distraction-free editor and nothing else. Only after writing does step two ask for mood and themes. With one home and one task, the navigation and the clutter came out, and nothing floats over the writing surface.

The two-step entry in motion: write first, then the reflect step appears.
The two-step entry in motion: write first, then the reflect step appears.
The Journal home: the entry timeline with the insight card and a single New entry action.
The Journal home: the entry timeline with the insight card and a single New entry action.
08 · Component

Mood became two axes, because calm and exhausted are not the same low

A single sad-to-happy scale hides half of how a day feels. The emoji gave way to an Affect Grid, a square where one axis is mood and the other is energy, drawn from Russell’s circumplex. One placed dot captures pleasant or heavy and wired or worn out at once.

The reflect step: the Affect Grid places one dot and reads it back as mood and energy.
The reflect step: the Affect Grid places one dot and reads it back as mood and energy.
09 · Iteration

Dictation worked in Chrome and nowhere else, so it was rebuilt

Voice input was added to lower the barrier to writing, first on the browser’s built-in speech API. It failed silently outside Chrome. It was torn out and rebuilt to record audio and transcribe it server-side with Whisper, which works everywhere, including mobile. A small, honest loop of build, test and adapt, with a concrete cause.

10 · The payoff

It reads between the lines only when asked

The reward for journaling is the Between the Lines card. By default it shows a plain-language conclusion computed entirely on the device, comparing this week to last, next to a swipeable mood-and-energy trend. The deeper AI reflection is one deliberate tap away, and only then does any text leave the device. Privacy stays on the writer’s terms, which is also how personvern and GDPR ask it to be.

A reflection in detail: the on-device conclusion with the honest privacy note underneath.
A reflection in detail: the on-device conclusion with the honest privacy note underneath.
The Insights card from the running build. The circled Saturday is the reason mood became two axes.
The Insights card from the running build. The circled Saturday is the reason mood became two axes.
11 · The Companion

The Companion answers the chatbot problem instead of ignoring it

A journal that reads between the lines invites an obvious next wish: someone to talk to about it. Instead of pretending that wish away, the product ships a conversation built the other way round from the generic chatbot. The Companion opens from the page, quoting what the week actually says, and hands a question back before it answers one. When the language crosses a crisis threshold, it drops the conversation and shows Norway’s real help lines, right in the thread. The same rule as everywhere else, an adjunct and never a therapist, only now it can hold a conversation without breaking it.

The Companion in the running build: a grounded opening, a question handed back, and the crisis guardrail with Norway's real help lines.
The Companion in the running build: a grounded opening, a question handed back, and the crisis guardrail with Norway's real help lines.

The reply to a chatbot that always agrees is not silence. It is a companion that asks back.

12 · Delivery

From concept to a clickable product

The point of the technical layer is a design that ships, not just a mockup. It runs end to end on React and TypeScript, a Tiptap editor, Vercel serverless functions, Whisper for dictation and Grok models for the on-demand reflections and the Companion’s replies. These tools were new, learned under a mentor’s guidance, and the thing worth showing is the distance covered, from concept to a genuinely usable product, with AI as an accelerator throughout.

Four screens from the running build: the journal home, the two-step entry, the Affect Grid, and the Insights card.
Four screens from the running build: the journal home, the two-step entry, the Affect Grid, and the Insights card.
Between the Lines in the hand: the journal home with the week's trend, in a lifestyle photo.
Between the Lines in the hand: the journal home with the week's trend, in a lifestyle photo.

Not a mockup to imagine, a product to open.

13 · Lessons

What the project taught, and where it goes next

Building it, rather than only drawing it, made every trade-off concrete.

  • Subtraction did more than addition. Removing the navigation, the clutter and the crude single mood axis.
  • Constraints from the evidence base sharpened the work instead of narrowing it.
  • The most honest feature is also the most trustworthy. On-demand, clearly labelled, never pretending to diagnose.
  • With no real users yet, every choice is still a hypothesis. Real journalers are the next test.

The product will keep developing: real journalers first, then deeper analysis of language over time.

14 · Sources

The public material behind the framing

Cited as references, not claimed as fieldwork. US usage data and Norwegian market context are kept separate, and figures should be re-verified against the primary texts before any public use.

  • Sentio University (2025). Survey of people with an ongoing mental health condition; about half use language models for mental-health support.
  • Healthline; Toronto Metropolitan University; Psychiatric Times (2025 to 2026). Expert warnings that chatbots affirm rather than challenge, can state false information confidently, and are an adjunct, not a replacement.
  • Norwegian GP-DEP Study; NCDNOR. Depression treatment and referral patterns in Norwegian general practice, and uneven access to specialised care.
  • ScienceDirect welfare-state review (2024). Mental disorders as the costliest condition in Norway, close to three percent of GDP.