4 min read

How I take my personal notes, with AI

Obsidian is awesome. It has always been my favourite way of note taking. But with AI taking over, it was the last missing piece now. Obsidian is I think the best way to take notes and work on inferences from our existing notes hands down. With some skills and some routines, this is as powerful as having a PA I feel. I'm not over hyping it, I promise.

Sync across devices

I happen to have an iPhone & an macbook pro, so iCloud is a given. Now Obsidian vaults can be placed in iCloud since they are just markdown files. Which means I can edit my notes on both my phone and my laptop. This is a crucial piece for the puzzle.

Philosophy

Now that sync is sorted, next we get into the philosophy of note taking.

Always keep the raw files

have the raw files stored up in a folder. LLMs can make mistakes and it always helps to have the sources handy. Also, LLMs do a great job at simplifying data from complex raw files. This is the most important step.

Write with AI

Voice as input has gotten really good. Now you don't need to sit down and type out every word, just dictate and AI will do a good job of cleaning stuff up and writing a nice update to a note or a new note. Make use of this convenience to add data into the vault as much as possible.

Daily notes as a journal

You could argue that daily notes are also a raw source. I don't disagree. That's one way to look at it. Just make sure to journal all the important points so that when looking back it helps to connect the dots. I also roll up daily notes monthly once into a large monthly note since I don't seem to like the clutter of having one note per day, but this is just personal preference.

Projects

Track all the projects, with all the attachments (in raw folder) under one folder. If you are shifting houses, that's a project. If you are planning to upgrade your CPU and analyzing where the bottleneck is , that's a project. Its funny, but every month you'll have some or the other projects. Its worthwhile to keep notes and findings around them in one place.

Hobbies

These are what are called living notes, they tend to slowly evolve over time with projects and other information. Worthwhile to keep them somewhere neatly.

And then everything else, I tend to keep health, wealth, community and self as other top level folders.

Data sources

The raw folder is really powerful. You can store pdfs, csvs, power points or whatever you feel like there. and then point the AI at it and extract valuable info. here are the data sources I update frequently or every month atleast

  1. Voice notes - I use voicenotes.ai extensively and have a obsidian plugin to sync all my voicenotes to this folder
  2. bank statements - If you want to understand something, follow the money. This contains the bank statements of my bank statements, credit cards etc.. really high signal info.
  3. investments - I connected the zerodha MCP to AI and it pulls down my current state of investments and stores a snapshot of it here. useful for audit trail.
  4. order history - I pull data from amazon using Amazon order history reporter in csv files and store it here.
  5. calendars - I pull data from my google calendar using gws cli. I tend to create events when I plan on doing something or when I'm doing something, for instance when I'm travelling, movie bookings etc..

and everything else just lives somewhere else. For example, there are other misc folders contain tickets, invoices, payslips etc.. the AI decides where to put what.

Rituals

The monthly review

The most high signal activity I do with my notes. Every month start I make sure all the data is populated for the last month and ask it to do a monthly review of the previous month. The broad structure is as follows.

# Health
## Data
- [[raw/calendars/2026-04/kalluribharat@gmail.com.ics]]
- [[raw/calendars/google/2026-04.json]]

## State
...

## Focus
1. ...
2. ...
---

# Wealth
## Data
- [[raw/bank-statements/hdfc/2026-04-01-hdfc-bank-statement.csv]]
- [[raw/credit-card-statements/diners-black/2026-04-01-diners-black-cc-statement.xls]]
- [[raw/investments/zerodha/2026-04.json]]

### Spending by Category 
... (a large table of category wise spends)

### Notable
...

## Focus
1. ...
2. ...
3. ...
---

# Community

## State
...
## Focus
1. ...
---


# Looking Ahead
...

Skeleton structure for monthly review

With the raw sources intact, the monthly summary is very high signal. Overtime I also roll up monthly summaries into quarterly summaries and yearly summaries. Its really nice to zoom out and get a view of the basics.

I've made this a skill now and I just invoke the skill every month end for this entire thng to finish. Its really powerful.

How this all translates

The obvious upside of having monthly reviews and visibility into all major aspects of life. Course correcting and measuring becomes easier since there is data to backup almost everything. There are other upsides which are very interesting, here are some examples

  1. Recently we were planning a roadtrip from Bangalore to Allepy. There is already a travelling hobby note stating where we want to go sometime, there was a note with a travel subscription where we could get discounted rates on hotels. So now with that context, codex was comfortable able to plan out a seven day road trip with stops every 300 kms with bookings where we could get discounted stays. Its really awesome how much it can stich data nicely once given the sources.
  2. Was considering if I should apply for & get a amazon icici credit card given I already have a diners black card. was able to quickly pull out yearly order volume, check cashback/reward point rates across cards and decide that with my historic order volume, the new credit card does not add any significant upside. so decided against it.
  3. Recently I was shifting houses, AI really came in handy to remember how much deposit I paid last time (bank statements), helped in remembering and shortlisting properties we searched for and for many other small things. AI really helps in project context (given raw files are rich)
  4. Some simple questions get really easily answered, some questions I just recently asked
    1. How many computer mice did I buy? (too many 🤣)
    2. How many times did we eat outside last month?
    3. How much did the standing desk cost? How much did the CPU cost? When did I buy the macbook? How much was the deposit we paid for the current rental property?
    4. When is the last time I ran a 10k? : I had strava data synced so this came handy

This is a living system, so as it evolves I'll keep this updated. But that's what I have for now. Cheers!

Share
Bharat Kalluri

Bharat Kalluri

Bangalore