I was reading an old journal entry when I felt something and I did not have a name for the feeling. Not nostalgia exactly, more like running into a past version of yourself on the street and realizing you've been carrying something they left behind without even knowing it.
The entry was from three years ago. I barely remembered writing it. And reading it, I understood something about how I make decisions that I had never been able to articulate out loud.
That moment was the reason I built Pensio. But it took me a while to get there.
Years of my live in a folder
It started with a stubborn kind of curiosity. I wanted everything in one place, so I did the boring, slow, slightly obsessive work of converting all 400+ entries into Markdown. Different formats, different apps, different export quirks. It took a while. But eventually I had it: years of my life, sitting in a single folder, in a format I owned.
Then I started reading some of the old files. It was funny to see the worries of the time, things that now seem so small I didn't even remember them. But also my thought process for solving some of them.
It was like looking at an analog photo you forgot you took. Not like the thousands of digital photos you curate and filter, more like a raw shot from years ago, where you naturally captured something real without thinking about it. Some entries gave me strong feelings while reading about hard moments, conflicts, people I met, and many past versions of myself.
The files were there. But insight about specific feelings, places, and people wasn't.
Why the obvious solutions didn't work
The first thing I tried was obvious: paste some entries into a local AI and ask what patterns it noticed. It could summarize what I wrote, but the answers often didn't make sense. I asked about goals I had set, and it just brought random pieces of text together without understanding the context of any of them.
So I started reading about RAG, Retrieval-Augmented Generation, the technique that lets you point an AI at a set of documents and ask questions. It sounded exactly like what I needed.
It was a disaster.
RAG was built for FAQ-style knowledge bases. You have a question, it finds the chunk of text that answers it. But journal entries don't work like that. They're emotional, non-linear, full of context that lives between entries rather than inside any single one. When I asked "how did I feel about my job last year?", the system pulled three random entries that happened to contain the word "job," written years apart about completely different things.
RAG doesn't understand that two entries about the same person, written six months apart, are connected. It doesn't understand you. It just finds text that looks similar and hopes for the best.
I knew I needed something different.
The moment it actually worked
I started building a pipeline. Ugly, hacky, not something anyone else could use, but it was just for me.
Instead of dumping entries into a database and hoping for the best, I'd first read every entry with an AI and extract structured meaning: emotions, intensity, themes, people mentioned, whether I was reflecting on the past or anxious about the future. Then I'd store that alongside the raw text, building something like a graph where I could navigate between entries and their connections.
Once the data was structured, I could teach an AI to navigate my journal, not just search it. I could ask "what do I consistently write about before a hard decision?" and it could actually reason across my history. That is what I was looking for all along, not a search engine for my journal, but something that understood the connections between my emotions, my decisions, and the people in my life.
The first time it really worked, I asked something like: "Have I ever felt this way before about this topic?" It came back with an entry from two years earlier, made an emotional connection I hadn't seen myself, and pointed me to the sources so I could verify it. I sat back and just looked at the screen for a moment.
The aha moments kept coming. Contradictions I had never noticed. Things I believed strongly one year that I had quietly abandoned by the next. Patterns around specific people, specific seasons, specific types of decisions. The journal wasn't just a record anymore. It felt like something I could walk back into, and keep discovering more of myself.
Building it for other people
For a while, it was just mine. A collection of scripts that worked beautifully for one person and would have been incomprehensible to anyone else.
Then I showed it to a few people, carefully, because I wasn't about to share 400 private entries. My partner was the first. Her reaction told me this wasn't just for me. Some friends had the same reaction, and a few of them became the first beta testers. Watching other people have their own aha moments, about their own lives, their own patterns, their own past selves, made it clear that this was worth building properly.
So I made a decision: do it right. Build it as an actual app, with a real interface and privacy I could stand behind.
I built it to solve my own curiosity about myself, and the more I refined how the AI navigates journal connections, not as documents but as a life, the more I realized it could be useful for a lot of other people too.
If it helped me understand years of my own story, I think it can help others understand theirs.
Pensio is free to try. Your journal already knows things you've forgotten. You just need a way to ask.