Pylimitics

I was reminded of the memoir What I Talk About When I Talk About Running by Haruki Murakami. I read it a few years ago, mostly because I found the title intriguing. What reminded me was this post from Herman Martinus, the developer and maintainer of Bear. He cited this passage from Murakami:
“…I can’t grasp much of anything without putting down my thoughts in writing.“

When he continued to say “I’ve also found that the best way for me to understand a concept or idea is for me to write about it,” I realized this is also true for me. It’s the reason I take notes, and it’s the reason behind the writing I do that I have no intention of publishing or distributing. It’s not writing to communicate with an audience. It has a different purpose: clarification for myself.

One of the biggest lessons I’ve learned in writing technical content is that the first draft is usually for me; it’s how I discover what I understand and don’t about a subject. It’s the first draft that enables me to ask the questions I need answered in order to advance my understanding. Once I arrive at a draft that has all (or most) of my questions answered, I’m ready for reviews.

The writing tools I use for different projects affect how the writing helps me understand. For creative, freeform, and I’m-not-sure-what-this-is writing, manual notes on paper work best for me. The next step up, when I have a clearer idea what I want to write, is a minimal text editor. And for writing I expect to need formatting and embark on with some confidence, I sometimes go straight to a writing tool like a word processor.

I know I’m not alone in this, and I’ve read that for some people, even the pen or pencil they use with paper makes a difference for them. I haven’t found the writing implement itself is significant for me; as long as it makes legible marks it’s generally okay with me. I have a small jar of pens and pencils on my desk, but only because one of them occasionally runs out of ink or has a broken point, so I have backups at hand.

Like Herman Martinus, I generally don’t go back and read the things I write for my own understanding. The real output from that writing can’t be seen on the paper; the real output is kept invisibly in my mind. Even so, I’m often reluctant to discard my notes from a project. Some of this, I think, goes back to the practice of keeping an engineering notebook during a development project. In at least one company I worked for, those notebooks were collected and archived in some way. For all I know, they’re still in some temperature-controlled storage bin somewhere.

I wonder how different the galaxy-class document that is an LLM would be if it included not only the published and digital content written by millions of people, but the unpublished analog content as well.