I’m building a journaling system that remembers what you already survived
Most journaling apps give you a text box.
Then they expect you to do the hard part:
- Remember old emotional patterns
- Connect past experiences to current struggles
- Detect recurring cycles
- Extract meaning from hundreds of disconnected entries
After journaling daily for a long time, I realized something strange:
The most valuable insights already existed in my entries, but I almost never revisited them at the right moment.
So I started building a memory system inside Deeditt that tries to solve that problem.
The idea is simple:
What if your journal could detect patterns across time and remind you that you already survived something similar before?
How it works
Every journal entry goes through a signal extraction pipeline:
- Emotions
- Themes
- Narrative stage
- Cause/Effect relationships
- Semantic embeddings
Those signals are then used to build a graph of memory connections between entries.
Some examples:
- Recurring emotional cycles
- Unresolved patterns
- Thought evolution over time
- Emotional recovery arcs
- Same feelings recurring every winter or every Sunday
- Semantically similar experiences separated by months
The interesting part is that this is not keyword search.
People describe the same emotional state in completely different ways:
- “I feel exhausted”
- “Everything feels heavy lately”
- “I think I’m reaching my limit”
Traditional search treats those as unrelated, semantic retrieval does not.
The architecture
Current stack:
- Go APIs/services
- Python analysis pipeline
- PostgreSQL for surfaced data
- SQLite per-user memory stores
- embeddings + cosine similarity
- FTS5 pre-filtering before vector comparison
One unexpected lesson:
The hardest problem is NOT embeddings, the hardest problem is retrieval.
Finding the right memory at the right moment is significantly harder than detecting semantic similarity.
The system currently generates candidates using:
- Similar experiences
- Past resolutions
- Recurring temporal patterns
- Same-day-across-years recall
But deciding which memory is actually useful to surface today is still an open problem.
And honestly, that problem feels closer to cognitive architecture than traditional software engineering.
Biggest realization so far
I originally thought this project was mainly about AI.
It isn’t.
It’s mostly about:
- Memory systems
- Time-aware retrieval
- Narrative structure
- Emotional context
- Meaningful resurfacing
Most journaling software is optimized for capture, very little is optimized for understanding.
I think there’s a huge opportunity there.
Especially now that semantic systems are finally cheap enough to experiment with at small scale.
Curious if anyone here is building similar “personal memory” systems or experimenting with long-term semantic retrieval.
This feels stronger than a journaling app. The real product is emotional memory retrieval: helping someone recognize old patterns, recovery arcs, repeated states, and moments they already survived but would normally forget when they need them most.
That is a much more differentiated frame than “AI journal.” Most apps stop at capture, reflection prompts, or chat. Deeditt sounds like it is trying to become a memory layer for emotional continuity, which is a deeper and more defensible direction if you can make the resurfacing feel timely instead of random.
One thing I’d think about early is whether Deeditt carries that emotional trust layer well enough. It sounds functional, but the product direction feels softer, more personal, and more recovery-oriented. If this becomes a serious personal memory and emotional support system, Lyriso.com would probably fit the tone better.
I really appreciate this perspective. Interestingly, this is also something I've been realizing while building Deeditt: the most interesting part is no longer the "journal" itself, but the continuity between experiences, memory, reflection, and shared human knowledge across time.
The post focused mostly on the private memory/retrieval side because that's the part I'm currently deep into architecturally, but Deeditt itself is actually broader than journaling. The bigger goal is creating a space where personal experiences can become structured knowledge others can learn from as well, not just content to consume.
And now you have mentioned this, kind of agree about emotional trust. I think systems dealing with memory and reflection need a very different tone than traditional social platforms or productivity apps, but also it's a challenge to communicate the duality.
Hi Jack, one practical thought after our Deeditt thread.
You mentioned the hardest part clearly: communicating the duality between private memory, reflection, recovery, and shared human knowledge.
That is exactly the kind of naming/positioning problem where a sharp outside audit may help before the product gets more public memory around the current frame.
I do focused naming and positioning audits for early products: current name risk, category framing, domain weakness, emotional trust fit, and what stronger naming direction I’d take before more users, launch assets, or public perception build around the current name.
It is not a long consulting thing. Just a clear written breakdown with practical recommendations.
I’m doing a few at $99 while refining the format. If useful, I can do one for Deeditt and give you a direct outside read on whether the current name can carry the bigger memory/knowledge direction or whether a softer trust-led brand path makes more sense.
Best place to discuss is here:
https://www.linkedin.com/in/aryan-y-0163b0278/
That duality is exactly why I’d take the name seriously now, not later.
If Deeditt stays a private journal, the name can probably work because it describes an action.
But the product you’re describing is bigger than that. It is memory, reflection, recovery, lived experience, and shared human knowledge. That is not a normal productivity app category.
The risk is that Deeditt may keep making the product feel like something users “do,” when the deeper promise is something they return to and trust.
That matters early because names train perception before the product gets explained. If people first place it in the “AI journal / social reflection app” bucket, it becomes harder later to reposition it as an emotional memory and knowledge layer.
That is why Lyriso.com feels like a stronger direction to me. It has a softer, more human tone, and gives the product room to become a trusted emotional-continuity brand instead of sounding like a functional app.
If Lyriso feels aligned with where Deeditt is actually going, this is the stage where I’d secure it before the current name becomes harder to undo.
Happy to talk privately if useful:
https://www.linkedin.com/in/aryan-y-0163b0278/