Okay, I’ll admit it: I’m a bit weird. My weirdness? I love writing documentation. I know, not exactly rock and roll. But there’s something about putting things down in writing that helps me understand them better. It’s like my brain needs to see it on paper (or a screen) to really click.
Turns out, this isn’t just a personal quirk – it’s becoming a superpower in the age of AI.

Years ago, I took a Gallup StrengthsFinder test, and one of my top strength was “Input.” Basically, I’m wired to collect and organize information. Which, hey, explains the documentation thing!

Writing Isn’t Going Anywhere

Even with all the cool new tech, writing is still a big deal. We still have books, articles, and all kinds of written stuff. I’m the kind of person who likes the nitty-gritty, the real details. That’s why I subscribe to O’Reilly – tons of in-depth knowledge packed into one place. I use it to feed my brain with stuff to work on.

If you’re reading this, you probably value writing too. That’s awesome! In a world of short attention spans, taking the time to read is a win.

But here’s the big reason why writing matters now more than ever: AI. Specifically, LLMs (Large Language Models). Think of it this way: AI doesn’t magically create awesome stuff. It learns from massive amounts of text. So, someone needs to write that text!

This is a big deal. Even big enough that AI companies are having to pay up to authors whose work was used to train their models. Knowledge in books is crazy valuable.

And it’s not just books. Code is text too! That’s why I’m so into GitOps. I won’t get too technical, but basically, when you manage your infrastructure with code, you can let AI agents help make things better. LLMs are good at understanding text, so they can understand code. This means we can focus on the big picture – building the right mental models – instead of getting bogged down in boring tasks. But how does all this connect to my documentation obsession? Here’s a story.

Big Projects, Big Problems

Everything’s great in a demo, right? And startups can be pretty smooth too. But when you’re working on a huge project, it’s a different ballgame. Lots of moving parts, lots of people, lots of… well, you get the idea.

I’m working on one of those projects now. And while we generally know how to get the job done, the real problem is a lack of information. Finding the specific detail you need? It can feel impossible. It’s like this “tribal knowledge” thing – where only a few people know how something really works.

I hate meetings that waste time. They suck energy and stop me from getting stuff done. And honestly, most meetings could be emails (or better yet, documented!). Now, there’s some documentation. But even okay documentation is better than nothing, because AI can help you use it. The worst case? No docs, or docs that are wrong.

AI to the Rescue (Again!)

It’s cool to think about AI agents automating stuff (and I’m working on a project like that myself!). But we can use AI to fix problems today.

Not everyone loves writing documentation (I get it, I’m the weird one!). So, let AI do it! Give it some pointers, or better yet, let it look at the system itself. It can write up a description. You could even skip the whole “document” thing and let AI answer questions directly.

That last one is a bit risky from a security standpoint, so creating a snapshot of info as documentation is good enough. But even with docs, things can get scattered. That’s where AI comes in again. Instead of reading through pages of stuff, you can chat with AI. Ask it a question, get a clear answer. No meetings, quick answers, less frustration. It’s a win-win-win. And it’s a pretty easy win – a low-hanging fruit.

To really use AI to its fullest, to build those mental models, we need to understand the systems we’re working with. And AI is the perfect tool to help us do that. So, let’s write stuff down. Or better yet, let AI write stuff down for us.