There is a real argument hiding inside the idea that we should stop taking only text notes.
I do not read that as "never write anything down." I read it as a useful reminder that our brains do not produce clean little Markdown files on command. Sometimes the thought comes out as a voice memo. Sometimes it comes out as a messy diagram. Sometimes it comes out as a half-legible sentence on paper that made perfect sense when you wrote it and looks like a medical incident two days later.
That is the actual second-brain problem. Not whether text is good or bad. Text is great. Text is searchable, portable, and easy for AI to work with. But if your system only respects neat text, it is quietly asking you to think in the format the tool prefers instead of the format your brain is actually using.
I am trying to go the other direction.
My second brain is becoming more mixed-media
For most everyday capture, I use transcription. Walking, talking, pacing, muttering through an idea, explaining something out loud before I know what I think about it. That is where voice is unbeatable.
Voice lets me catch the momentum of a thought before I over-edit it. It is not polished. It is not always elegant. But it gets the raw material out of my head before it evaporates.
Then there are the things I still write by hand.
Those are usually the thoughts only my noodle can noodle.
That sounds ridiculous, but it is true. Some ideas need the drag of the pen. Some decisions need arrows, boxes, ugly circles, and the little side comments that only make sense because I was in the middle of thinking them. Handwriting slows me down in a useful way. It makes the idea feel physical for a minute.
The trick is not pretending those handwritten notes are already part of the system. The trick is getting them into the system without turning them into another pile.
So I take a picture.
Then Second Noodle stores the photo and transcribes what it can. Sometimes it does a good job. Sometimes the AI looks at my handwriting and basically says, "Jimmy, I support you as a person, but I cannot endorse these letterforms."
Fair.
Capture should not punish the way you think
The strength of the source argument is that plain text notes can flatten too much of the thinking process. A text note can record the words, but it may lose the shape of the thought: what was connected, what felt important, what was circled, what got starred, what was shoved into the margin because it was not ready yet.
That matters because a second brain is not supposed to be a museum of saved sentences. It is supposed to help future-you recover the useful parts of past-you's thinking.
If the original thought happened as a sketch, a whiteboard, a notebook page, a voice memo, or a screenshot, forcing everything into text too early can strip away context.
The better pattern is:
- capture in the form that matches the moment
- preserve the original artifact
- transcribe or summarize enough to make it searchable
- connect it back to the project, idea, decision, or question it belongs to
- let AI help with resurfacing, but keep the human signal intact
That is where I think second-brain systems are getting interesting.
Text is the index, not the whole memory
I still want text. I want transcripts. I want summaries. I want tags when they are useful. I want search to work. I want the system to be able to answer, "Where did I talk about this?"
But I do not want text to be the only surviving version of the thought.
A photo of a handwritten note contains more than the words. It contains emphasis. It contains layout. It contains the fact that I wrote one line bigger than the others, or drew a line between two ideas, or boxed something because it felt like the point.
The transcription makes it findable. The image preserves the thinking artifact.
That distinction matters.
If the AI only stores the transcript, it may recover the content but miss the texture. If it stores the photo and the transcript together, it has a better chance of helping me return to the actual moment of thinking.
That is the second-brain sweet spot for me: not perfect capture, but useful return.
The reader test
If you are building your own second brain, I would not start by asking, "What app should I use?"
I would start with a more annoying question:
What forms does your thinking actually take?
Do you think out loud? Do you sketch? Do you write by hand? Do you highlight? Do you screenshot? Do you make little lists and abandon them? Do you use voice notes because typing feels like a speed bump? Do you understand something better once you draw it badly?
Your system should respect those answers.
Then the job of AI is not to make all of that disappear into a neat database. The job is to help you find your way back.
What I am taking from the signal
The useful lesson is not that text notes are wrong. The useful lesson is that text notes are incomplete.
A practical second brain should handle multiple kinds of input because real thinking is not one format. It should let voice stay voice, paper stay paper, images stay images, and still give you enough transcription and metadata to retrieve the idea later.
That is the part I am paying attention to.
My workflow now looks more like this:
- transcribe the broad messy thinking
- handwrite the ideas that need slower attention
- photograph the handwritten notes for ingestion
- let Second Noodle store the image and extract what it can
- accept that sometimes even AI cannot read my handwriting, which is rude but not inaccurate
That feels closer to the actual goal.
Not a perfect archive.
A second brain that can work with the way a real human noodle noodles.
Original signal: MakeUseOf surfaced the argument in "Stop taking text-only notes — here's how I use Obsidian with a local AI to visualize everything."