I’m an avid journaler. Journaling (on and off) since my college days (2010’s). I’ve been doing physical journaling (good old pen and notebook), and digital (Notion, Rome Research, Obsidian and lately VS Code). So far, I tried to keep some form of gratitude journal, weekly retrospectives, and habit tracking. AI, and in particular Large Language Models (LLMs), finally made me fully switch to digital journaling and unlocked amazing new possibilities. For example, talking with my data, summaries, and mood and trend/pattern analysis. Read more in this episode.
Some Thoughts on Digital Journaling, Future-Proofing and Tool Lock-in
I’ve been journaling on and off since my college days (2010’s). Especially intensely since early 2022, having a physical gratitude journal and different digital databases (e.g. ‘graphs’ in Roam Research or ‘vaults’ in Obsidian) and even digital gardens.
Journaling is a great tool for developing metacognition, and creating notes is (at least for me) a great way of sense-making. But creating all these notes and journal entries is quite an effort, even if you like doing it. Logically, you don’t want all this effort to go to waste. So it’s about making your journaling/note-taking output future-proof. This is how I started using Obsidian and physical notebooks. Let me explain.
A silly example: Imagine you used to store your notes on CD’s or, even worse, floppy disks. Nowadays, you would practically (almost) have no way of accessing your data anymore.
A more realistic example: imagine you were using a note-taking app that goes bankrupt. This sounds extreme, but we’re talking about decades, and nowadays things go sideways on much smaller time-scale. (This almost happened to Evernote, a note-taking app that was once all the rage.)
Physical (note)books are interestingly very resilient – taking the story to extreme, written text from 3500 BCE in ancient Mesopotamia (modern-day Iraq) is still readable and available today. However, unlike digital data, they are not easily searchable, editable, portable etc.
A tool is a tool, and in the end of the day you should choose what works best for your context, but you should be aware of the tradeoffs you need to consciously make. It boils down to:
- Future-proofing: will I be able to access this data in 10, 20 or 30 years? What about my (fictional) grandson?
- Easy to (re)use: can I easily copy, share, edit, and search my notes?
- Tool lock-in: If I switch, how easy is to take my data with me? Any other pain points or friction?
- Wow-factor: Does it really do a great job for your context specifically?
I made a table for tool scoring of the tools I’ve used – you can make your own to help you decide.
🟢 – Great
🟡 – Solid
🟠 – Passable
🔴 – Bad
How AI Changes the Game
As mentioned, journaling in itself is a great tool for metacognition, but for me, the activity is not a goal in itself. I’ve kept a daily journal for years, but recently I realized I was sitting on a goldmine of personal data I was barely ‘mining’.
Once you create all this data, you open up a lot of possibilities if you make it easily searchable, actionable, and informative. Large Language Models (LLMs) take this to another level. For example, with Claude Code, all of a sudden, you’re able to literally talk to your journals, query in natural language, automate tasks, and more.
Imagine having 2 years’ worth of journal entries, study notes, work stuff, and plugging it all into an LLM7. It’s… powerful. Let me show you what this looks like in practice.
Here are some examples of what Claude is doing for me and my journaling practice.
Claude Code + Markdown: The ultimate journaling stack
Recently, I’ve been using Claude Code in my journaling. These were, in fact, the first 4 skills I taught Claude: analyzing my mood, daily journal entries, weekly, and monthly reviews.
Mood Patterns – ‘/analyze-mood’
In my daily practice, I use dated daily journal entries (markdown files). Within daily entries, I write tasks I’m working on, focus areas, notes and remarks etc. Near the bottom, I always note down my mood. Claude’s ‘mood-analyzer’ skill uses this data to provide me with actionable insights, uncover patterns and summarize periods.
For my context, I’m following how far I deviate from the balance state (‘ok’). Whether it’s towards feeling euphoric (‘plus’) or feeling down (‘minus’). You can obviously choose any preferred scale or way to note down how you’re feeling (e.g. ‘How good was the day on a scale from 1 to 5’, or ‘Feeling bad, good, great…’ etc.).
Claude analyzes my mood patterns from the daily journal entries over a given time period. It searches for the ‘danas8: [mood]’ entries in my dated markdown files and reports on stability/deviation from the balance – ‘ok’.
Here’s how it looks in practice:
Beyond calling the skill, I could continue to ‘talk with my data‘ (or, more precisely, talk to Claude based on my data), e.g. ‘Tell me what tasks I did on the days I felt mildly down?’, ‘Any notes on the days my mood was mildly elevated?’ and so on.
Journal Entries – ‘/daily’
My daily notes include:
- Tasks I want to achieve with statuses
- Notes on work (‘posao’), life (‘zivot’), and other (‘ostalo’)
- Plus, minus, next (+, -, →)
- Mood for today9
The skill ‘/daily’ creates daily notes on my behalf. There’s a cron job (scheduled task) that runs daily at 09:00 (9 AM) to create the note for the day. I can also manually call it any time, and also optionally pass the date to create a note for a specific day. If a note exists, Claud will ask me before overwriting the one that already exists.
The unfinished tasks from the previous day’s note are automatically carried over to the new daily note, and everything the template is ready for me to fill it out.
/weekly-review
This skill generates a weekly review by scanning my daily notes for the
week. It summarizes tasks by status (done/partial/not done/delayed), analyzes my mood patterns, collects reflections and notes, and creates a review file.
A typical weekly review is a few pages long, but if you’re curious, check how it looks here.
/monthly-review
This skill generates a monthly review by aggregating my weekly reviews. It summarizes the weekly plus/minus/next items, mood trends, and task statistics across the whole month, leaving space for me to reflect.
A typical monthly review is a few pages long, but if you’re curious, check how it looks here.
Privacy – The Elephant in the Room
The obvious question: You’re sending your journal to an AI?
Yes, but with guardrails. I maintain a separate folder in my vault that’s not synced and is excluded from Claude’s access for truly private entries. The analyzed content is typically patterns in productivity, mood, and projects, not deeply personal details.
Why this over third-party journaling apps? Future-proofing and control. I choose what gets analyzed, API calls are auditable, and Anthropic doesn’t train on my data.
The tradeoff: I’m accepting calculated risk for powerful insights. This won’t be right for everyone, and that’s okay.
The New Era of (Self-)Reflection
Ultimately, the transition from physical notebooks to an AI-augmented digital “brain” isn’t about chasing the latest tech trend. For years, I was a data collector, filling pages and databases with reflections that rarely saw the light of day once the notebook was closed. By prioritizing future-proof formats like Markdown and leveraging the reasoning power of LLMs, we can turn those static archives into a living dialogue.
My journal is no longer just a place to write down; it’s a coach, co-pilot, and assistant that identifies my burnout patterns before I do, tracks my mood, ensures my long-term goals don’t get lost in the noise of daily tasks, and more.
As you consider your own journaling stack, I encourage you to look beyond the interface and think about the legacy of your data. Whether you’re using a pen or a terminal, the goal is the same: developing a clearer understanding of yourself. Tools like Claude Code have simply made that understanding more accessible, actionable, and… exciting.
The ‘wow-factor’ isn’t just in the automation; it’s in the realization that years of ‘sense-making’ are now finally ready to talk back.
Are you using AI for your notes and journaling? Would you like to try it out? Let me know in the comments!
P.S. If there are interested builders, I can do a more technical deep-dive, cover API costs etc.
* Featured image made with GenAI and Canva
- Physical notes in notebooks etc. are timeless and will survive thosands of years unless destroyed by fire or similar. But, they are not easy to search, copy, edit etc.
↩︎ - Notion has a lot of great features, it’s easy to use, search, copy etc. It is relatively easy to export as well, but you are quite locked-in, and they are not as ‘too big too fail’ like Google or Apple are so not very time-proof.
↩︎ - Roam Research has a lot of great features, it’s easy to use, search, copy etc. It is relatively easy to export as well, but you are quite locked-in, and they are not as ‘too big too fail’ like Google or Apple are so not very time-proof. ↩︎
- Obsidian has a lot of great features, it’s relatively easy to use, search, copy etc. It’s based on pure markdown (.md), which makes it much more time-proof than Notion or Roam Research, or even Google Drive or Apple Notes. However, the more plugins (for advanced features) you use, the more it strays from pure .md which makes it less future-proof. You have a tradeoff between functionalities and future-proofing. Personally, I almost use no plugins.
↩︎ - Google Docs are solid-to-passable future-proof because of the scale of the company behind them, and how core they are. Quite feature-rich as well, but very vendor-locked-in.
↩︎ - Apple Notes are is solid-to-passable future-proof because of the scale of the company behind them, and how core it are. Quite feature-rich as well, but very vendor-locked-in, even worse than Google.
↩︎ - Make sure you don’t share anything confidential or proprietary. In my case, I shared no work artefacts.
↩︎ - Danas means ‘today’ in Serbian
↩︎ - See about mood tracking under the mood-analyzer skill
↩︎