Reinventing Journaling With AI
TL;DR
- Problem. Journaling is a high-value habit that almost nobody sustains. The friction is always the same: blank page, finding the words, finding the time. Existing apps do not remove that friction, they just redecorate it.
- My role. Solo creator and director. I shaped the product thesis and the interaction model, then directed the build. The original was a web app I built end-to-end; the native rebuild was planned by Claude Fable and implemented by Claude Opus, under my direction.
- What changed. A voice-first journaling system that turns raw speech into a well-written entry that still sounds like you, then layers on tagging, mood analysis, weekly summaries, semantic search, and retrieval-grounded chat across your full history. It started life as a Progressive Web App and is now Afterword, a native iOS app.
- Result. A native iPhone app built on a dependency-free Swift package (JournalKit) with 94 passing tests, GPT-5 pipelines, and end-to-end encryption via Apple's CryptoKit — verified byte-for-byte compatible with the original web crypto. Heading to the App Store, on TestFlight first.
- Why it mattered. It is proof that a privacy-respecting AI app can make an unloved habit feel effortless, and that you can rebuild a mature product natively, better in every dimension, by letting one AI plan the work and another execute it.
Journaling has always seemed like a powerful tool.
It helps with reflection, memory, and understanding what's happening in your life. But traditional journaling has one big problem: it requires discipline most people don't have.
Writing for thirty minutes every day sounds good in theory. In practice, it rarely happens.
That realization is what started this project.
I wanted the benefits of journaling without the friction.
And the obvious solution was voice.
The Insight: Speaking Is Easier Than Writing
There are many moments in a normal day when reflection would be easy if the barrier were low enough.
Driving home from dropping my kids off at school. Walking somewhere. Sitting in the car before going into a meeting.
In those moments, speaking for two minutes is effortless.
Writing for ten minutes is not.
Voice journaling removes most of the friction that prevents people from capturing their thoughts. You can simply talk.
But once AI enters the picture, something much more interesting becomes possible.
Instead of storing raw entries, the system can turn spoken thoughts into structured knowledge.
That insight became the foundation for the product.
A Journal That Understands What You Said
The product experience is intentionally simple.
When you open the app, the main screen has only one primary action: record.
You tap the record button and start talking.
When you're finished, several things happen automatically:
- The recording is transcribed into text.
- AI shapes it into a clear, well-written entry that keeps your voice.
- Metadata is generated.
The system extracts:
- a fitting title
- themes and tags
- mood and sentiment
The entry is then stored in a searchable archive.
Instead of a collection of disconnected notes, the journal gradually becomes a structured record of your life.
Weekly Reflections Generated by AI
One of the most powerful features emerged once the system started accumulating entries.
Every week, the AI generates a weekly summary.
This acts as a "week in review" built entirely from your own recorded thoughts.
The summary includes:
- mood trends across the week
- recurring themes
- major highs and lows
- meaningful quotes from entries
- gentle, optional life-coach suggestions drawn from what you wrote
These reflections often surface patterns that are difficult to notice in real time.
It's one of the moments where the system starts to feel genuinely intelligent rather than just automated.
The Moment It Became Real
The turning point during development was seeing the weekly insights for the first time.
Instead of simply storing journal entries, the system was interpreting them.
It could identify recurring themes I hadn't consciously noticed.
It highlighted moments that mattered.
It created summaries that made the week easier to understand at a glance.
That was the moment the project shifted from being an experiment to feeling like a genuinely useful tool.
The AI wasn't a gimmick. It was providing perspective.
A Personal Knowledge Archive
Over time, the system becomes something more than a journal.
It becomes a searchable archive of your thoughts, experiences, and intentions.
You can ask questions like:
- What was I worried about last month?
- What goals did I say I wanted to work on?
- What patterns keep repeating in my entries?
Because the entries are structured with metadata and themes, and embedded for semantic search, the archive becomes queryable by meaning rather than keywords. You can even chat with it — ask a question and the app searches your whole history to answer, like a conversation with your past self.
If you mentioned a goal two weeks ago and haven't revisited it, the archive can surface it again.
From a Web App to a Native iOS App
The first version shipped as a Progressive Web App. It was local-first, installable on any device, and it worked. But a browser is always wearing someone else's clothes. Real journaling on a phone deserves real native foundations: instant launch, a microphone that behaves, a live waveform driven by an actual FFT of your voice, home-screen widgets, a Live Activity while you record, and the fit-and-finish of iOS itself.
So I didn't patch the old app. I rebuilt it, and renamed it Afterword.
The interesting part is how. Before a line of Swift was written, I handed the entire problem to Claude Fable and asked for a grand master plan: a full port to native iOS that shared the existing backend, the same wire formats, and the same encryption, byte-for-byte. What came back wasn't a sketch. It was an architecture — phased milestones, a cleanly split Swift package, frozen contracts so the new app could talk to the existing server on day one, and a test strategy to prove the new crypto matched the old exactly.
Then I handed that plan to Claude Opus to build. Because the thinking was already done, implementation moved fast. The result is JournalKit: an eight-module Swift package — Swift 6 with strict concurrency, SwiftData, Apple's CryptoKit, and zero third-party dependencies — with 94 passing tests under a SwiftUI app designed for iOS 26. The encryption layer is verified byte-for-byte compatible with the original web crypto in both directions, so a journal moves between platforms without a single byte out of place.
A plan good enough that the build felt like reading from sheet music.
Private by Architecture
Privacy was non-negotiable from the start, and going native made it stronger. Your saved entries are end-to-end encrypted on your iPhone with a key only you hold — the server only ever stores an unreadable blob. (Transcription and the AI that shapes your words run on the server to turn your voice into an entry; your saved journal stays encrypted.) You can keep everything local to your phone, or turn on encrypted cloud backup for sync across devices. No ads, no tracking SDKs, nothing sold, and you can delete your account and all your data whenever you decide.
Why This Kind of Tool Matters
Most journaling apps today fall into one of two categories:
- simple text notebooks
- heavily guided self-improvement apps
Neither approach felt right for what I wanted.
I didn't want therapy prompts or motivational coaching.
I wanted something neutral.
A modern version of a traditional journal, enhanced with AI.
Something that helps you:
- capture thoughts easily
- reflect on patterns
- explore your own history
AI makes that possible in ways that weren't feasible before.
Building Something I Wanted Myself
This product started with a simple motivation: build the tool I wished existed.
That's often the most reliable way to discover useful products.
If a problem feels obvious and frustrating to you, chances are many other people feel the same way.
After analyzing existing journaling tools on the market, it became clear that most of them either focus heavily on self-improvement frameworks or lack meaningful AI-driven insight.
This product aims to strike a different balance.
Simple voice capture.
Powerful AI interpretation.
A personal archive that becomes more valuable over time.
The Bigger Vision
AI is changing the kinds of tools we can build, and the way we build them.
For the first time, software can help people reflect on their lives at scale. And, as the rebuild showed, it can also do much of the engineering itself: one model planning the architecture, another writing the code, with a human directing both.
Instead of static journals, we can create systems that:
- understand what we say
- identify patterns in our thoughts
- surface insights we might miss
- build long-term personal memory systems
This project explores what happens when you treat journaling not as a notebook, but as a thinking tool.
A system that helps people understand themselves a little better over time.