I’ve been leading engineering teams for a few years now. I don’t write code day-to-day anymore. That’s fine, that’s the job.

Lately though I’ve been feeling a gap between what I understand about how software gets built and what’s actually changing on the ground. AI-assisted development is moving fast. I read the posts. I watch the demos. I hear my engineers talking about the tools they’re using but watching something change isn’t the same as working through the change yourself. I try to carve out the ability for this within my day job, but as many of you I’m sure know, that is just not a realistic expectation. Especially if I want to do meaningful learning vs. dropping in to build enough context for a hot take. If I’m going to make good decisions about how others work, I need to close that gap myself, not just observe it from a distance.

So I’m going to write about what I’m hacking on. I am not going to pretend to have many answers; most of the time I’m just trying to form better questions. I want to do this publicly, partly because I think it will motivate me and partly because I want to know if other people are asking themselves similar questions.

So that’s the motivation; my posts will likely add up to a hodgepodge of ideas.

This blog is the first experiment

This isn’t my first foray into blogging, about 4 years ago now I started writing about some DevOps specific topics. The old site was built with WordPress which just took WAY too long to get going and get looking the way I wanted. I did not have WordPress experience and I am not proud to admit it took me weeks after work to get it looking right. If only I had Opus at the time I have no doubt it would have saved me some headache by recommending I use something like Hugo which was around at the time. This time my motivations are very different than they were then and I didn’t want to spend days or weeks setting up infrastructure for a blog. I wanted to sit down, describe what I wanted, and see what came out the other end. So I opened Claude Code in an empty directory and started talking.

The whole thing took one session. Not a weekend. Not a series of evenings after the kids were asleep. One sitting.

I started by asking Claude to research the current landscape for static site generators and hosting. What are people actually using in 2026? What’s cheap, easy to maintain, and quick to get started with? It came back with a comparison of Hugo, Astro, Eleventy, and a few others, alongside hosting options like Cloudflare Pages, Vercel, Netlify, and GitHub Pages.

We made decisions together. Astro for the framework because it ships zero JavaScript by default but supports interactive components if I ever need them. Cloudflare Pages for hosting because the free tier has unlimited bandwidth and deploys happen automatically on every git push. Total hosting cost: $0. The only recurring cost is the domain registration that I already carry.

From there, Claude scaffolded the entire site. Layouts, pages, content collections with typed frontmatter schemas, an RSS feed, a sitemap, Open Graph meta tags. I steered on design. I wanted something clean and professional, not the default “developer blog” look that most static site themes produce. I like the aesthetic of my old site as at the time it was something I really iterated on, so I shared a screenshot and the old logo. We went back and forth a few times on header sizing, killed dark mode (I just prefer light), and tuned colors until it felt right.

Then I asked about comment functionality. We set up Giscus, which uses GitHub Discussions as a comment backend. Free, no database to manage. The blog’s source repo stays private and comments live in a separate public repo.

That was it. I had a blog.

The stack

For anyone who cares about what’s under the hood:

  • Astro 6 for the framework. Markdown and MDX support out of the box, typed content schemas, file-based routing. Writing a new post means creating a .md file with a title and date, then pushing to main.
  • Cloudflare Pages for hosting. Git-based deploys, global CDN, automatic SSL. Free.
  • Giscus for comments. Backed by GitHub Discussions.

The dev workflow is about as simple as it gets. I write a markdown file, commit, push. Cloudflare builds and deploys automatically. No CMS, no database, no servers.

What I took away

The blog itself isn’t remarkable. It’s a static site with some markdown files. What’s meaningful is how little friction there was between having the idea and having a working result. I put my daughter to bed at 8 pm, downstairs by 8:30 and was wrapped up by 11:30. I described what I wanted in plain English, made decisions when there were real trade-offs to consider, and had the AI walk through the project at the end to teach me the technical details. A day later and I am writing my first blog post.

That’s the kind of thing I want to understand better. Not “AI can build a blog,” but what it means when the cost of building something drops this dramatically. How that changes what’s worth attempting. What it means for the teams I lead and the way we think about what’s actually hard.

I don’t have answers to any of that yet but I’ll be writing about it as I attempt to figure things out.