
I’ve always struggled to keep a written journal, even though I found it genuinely helpful whenever I managed to stick with it.
Over the last year I started recording short video reflections instead using Photo Booth on my Mac. Since I’m already on my laptop most of the day, it was much easier to keep the habit and talking for five minutes felt much easier than writing.
The habit finally stuck. But I ended up with a growing collection of video files that I never revisited, so I wasn’t really learning anything from them.
So I built InnerArc, a private audio and video journaling app designed to be revisited. Rather than treating each journal entry as an isolated note, InnerArc analyses them together. It identifies recurring themes across weeks and months of reflections, groups related entries, and helps surface patterns that are easy to miss.
I’m particularly interested in helping people answer questions like:
What keeps coming up?
What has changed with time?
What themes define this season of my life?
The frontend is React Router/Remix and the backend is Ruby on Rails.
The most interesting part has been building the analysis pipeline. Each journal entry is embedded so semantically similar reflections can be clustered together, allowing recurring themes to emerge even when they’re described in completely different words.
I’m also experimenting with force-directed graphs as a way to explore those connections visually instead of scrolling through a chronological list of entries.
Today it analyses recurring themes across entries and visualises the connections between them. It’s built around the idea that the value of journaling isn’t just recording your thoughts but being able to return to them with perspective.
I’d genuinely love feedback from people who keep any kind of journal.
The shift from "recording memories" to "discovering patterns" stood out to me.
Capturing thoughts is already a solved habit for a lot of people.
Helping them notice something they wouldn't have noticed on their own feels like a completely different kind of value.
I'd be curious which insight has surprised early users the most.