Last week, I went to GitLab Connect Sydney. About 10 minutes in, I wanted to write about it — mostly to share some reflections, partly to stay visible in the industry, and maybe (just maybe) to get on GitLab’s radar a bit — do you guys need an influencer? haha.
The post I ended up publishing felt clean, thoughtful, and in my voice. But getting there was... not a straight line. This post is about how I got from raw input to polished output, using a combination of note-taking, AI collaboration, and a bit of back-and-forth editing. Spoiler: you’re reading a post about how I wrote another post, using AI. Very meta.
Step 1: Dump Everything Into ChatGPT
During the conference, I started a new chat, dropping photos and bullet points into ChatGPT. Things like:
- Snapshots of slides
- Quotes from speakers
- Quick reactions ("this was a good point" / "this felt rushed")
I told ChatGPT, 'Don’t say anything — just reply with noted until further notice,' because it kept trying to make sense of my weird stream-of-consciousness comments.
This felt a little like live tweeting which I would do at other conferences back before when twitter... was better.
The goal wasn’t structure — just capture.
It also let me stay present. I wasn’t rewriting slides or typing long thoughts during the talks. I just fired off observations and moved on.
Step 2: Work With Context
After the event, I had ChatGPT summarise everything I’d dumped in — and help shape it into an outline. But here's where it got interesting:
- The official conference page didn’t load for the AI (500 error), even though it loaded fine in my browser. I had to copy key info in manually. Perhaps GitLab's website is blocking it? I dunno. Weird.
- Some speaker names and quotes needed correcting.
- I asked it to match the style of my existing blog posts — and gave it links to a few of my previous blog posts to read and study. I asked it what it noticed about my writing style and adjusted the draft to suit.
The tone shifted immediately. Less corporate recap, more me.
Step 3: Edit Together
From there, I dropped into edit mode. I asked for tighter headings, cleaned up phrasing, and cut anything that felt redundant. It was iterative:
- “Can we make this punchier?”
- “This stat doesn’t add anything — can we replace it?”
- “Let’s change the order — that demo feedback comes too early.”
I still made final calls, but ChatGPT became a second brain. It remembered the flow, understood my tone, and let me steer.
Step 4: Polish & Publish
Once the post was shaped, I asked for:
- SEO title + description
- A short excerpt for the preview.
- A LinkedIn version (with tags and tone matching my intent)
This made the actual publishing part feel easy. The writing was already done. The context-setting and delivery just snapped into place.
A Note on Transparency
I’m experimenting with using AI to help generate blog posts. I don’t want to flood the internet with generic, AI-written content — that’s not the goal. What I’m trying to do is reduce the friction it takes to get my thoughts into the world. Thoughts that I believe are (hopefully) novel and valuable.
I work hard to make sure the final product still feels like me. I guide the tone, edit every section, and stand behind each post as if I typed every character myself. It won’t be perfect — no writing is — but I want to be upfront about the tools I’m using, and how I’m using them.
So What?
This isn’t a post about how AI is magic. I still had to think. I still had to edit. But it let me:
- Capture fast, without interrupting the moment
- Collaborate without context loss
- Write in my voice, without starting from zero
And I think that’s worth sharing. Not because it's the future of writing — but because it was just a really good way to write this thing.
So yeah. That’s how I wrote the last blog post.
This one too, actually.