Be careful with introducing AI into your notes

This is my entry to this month’s IndieWeb Carnival, hosted by V.H. Belvadi who invited us to write about friction:

I am pleased to host the first IndieWeb carnival of 2025 and invite you to think about the importance of friction, primarily on the web but possibly also elsewhere. Feel free to consider the idea in as abstract or concrete a manner as you like.

As I started thinking about the topic, multiple ideas popped into my head. My first blog post of the year about human curation would definitely fit the topic (and is something Bevaldi wrote about in his entry). I had already written and scheduled that post by the time the invitation was published so I decided I wanted me entry to be about something else.

I also considered writing about the friction of physical media. A few months ago I watched MecklesFrog’s Youtube video essay The Importance of Inconvenience which sparked some ideas but my thoughts on that topic require bit more marinating in my notes.

So instead, my entry this month is about why reducing friction with generative AI might not be such a good idea when it comes to taking and reading your notes.

I’m a note-geek and as one, I spend some of my free time reading discussion forums about taking notes. Recently, I have seen a surge of discussion about AI (more specifically, large language models) helping out with one’s notes.

Often that discussion covers one or both of the following:

• Automating creation of notes through AI summaries or other note computer-driven creation methods
• Getting insights out from the notes system with the help of AI summaries or chat bots

However, I think these AI tools only seem as something that makes it better while only making it seem easier. I do see the appeal for them. It does sound enticing to skip some of the tedious process of capturing knowledge, summarising the main points to avoid reading through bunch of notes to get an outcome when you need it.

Both of the two use cases are tasks that require thinking and working with notes. That’s what I consider to be the real value of having notes in the first place.

The point of working with notes is not to create notes nor it is to create outcomes of from those notes. Both of these are the necessary side effects and artefacts of the main process.

That main process is thinking. I have written earlier about how notes are a tool of thinking. When you write a note, the friction and the slow process of thinking what to write, interpreting the source materials and converting them into a concrete note is where the magic happens. Writing thoughts down into tangible words on tangible notes is so valuable because it takes vague ideas and forces you to work on them.

The same happens when you extract insights from those notes. You dive into the notes and absorb what’s in them and how they connect with each other. You then do the thinking to turn those ideas into tangible outcomes like writing a report, a blog post or new notes.

Sometimes the big, serendipitous discoveries happen because I know what’s in my notes. And I only know what’s in there because I’ve processed thoughts into notes myself and I’ve done the work to interact with the rest of the notes to see how things relate to each other.

A notes system that’s even partially created by an AI doesn’t provide me that knowledge. It’s like working with someone else’s notes. Sure, you can probably find something valuable in there but it’s hard to see the really valuable insights and connections because you don’t know what’s related to what.

I also don’t remember every note in my system. When I create a new note or extract information from the existing notes, I need to explore and read what is in there. Every time I do that, I learn more. Sometimes I do it at the time of writing a new note initially and sometimes I quickly record thoughts into a temporary note and then process it when I have more time to sit and think.

It’s very possible that an AI model can already create better outcomes (for some metric of better) from my notes. But that only matters if the artefacts created from notes are the main point and I argue they are not. If I’d do that, over time I’d lose the ability to work with my notes because I wouldn’t know well enough what the computer put there and where it read the information from.

I could use a tool like that to create a report at work from my notes but then I also would not be able to have good follow up discussions about that with a colleague or boss because I didn’t actually know what happened and what were in those notes.

What may seem like friction is actually valuable thinking.