I’ve used Wordpress for many many years, but I got to the point where I was tired of paying $100/year for a blog that gets a few hundred visits per month. I’ve played around with static site generators like Jekyll but have always dreaded the amount of time it would take to convert my existing Wordpress blog. Then I realized one evening: Why would I do this myself in the age of AI? Enter Claude Code.

The Prompt

Converting a blog is a major PITA. You have to download the existing Wordpress articles, port them to Markdown, change all the image links, make sure code is formatted correctly, etc, etc. Certainly I could get Claude to do this for me? What I didn’t expect was it would do a much better job than me in a fraction of the time.

I followed the GitHub Pages Quickstart to setup a fresh repo for my new blog. At this point it was just a skeleton with no content. Then I opened it in Claude Code and issued this prompt:

this is a personal blog written in jekyll. i want to convert my existing wordpress blog into jekyll posts. my blog is located at: nateware.com . for each post on my blog, please create a corresponding markdown file. respect any of the formatting in my existing blog - headers, code segments, links, etc. for each existing blog post, create a new jekyll markdown file in the _posts directory formatted as YYYY-MM-DD-[slug from existing blog]. For example my blog article located at https://nateware.com/2010/02/18/an-atomic-rant/ should be downloaded, converted to markdown, and saved as _posts/2010-02-18-an-atomic-rant.md. Create a plan

Claude then generated a plan so that for each post it would:

  1. Fetch the full page content via WebFetch
  2. Extract title, categories, and tags
  3. Convert HTML body to clean Markdown
  4. Write file to _posts/YYYY-MM-DD-slug.md

It even added this step that I didn’t ask for:

  • Add permalink: /:year/:month/:day/:title/ to _config.yml to match existing WordPress URL structure for SEO/link preservation

When it came to images, Claude came up with this strategy:

  • Download images referenced in posts to assets/images/
  • Update markdown references to /assets/images/filename.ext
  • Posts with known images: donut-based-security, atomic-rant-redux, replacing-macbook-hd-with-an-ssd, how-amazon-ended-up-in-san-diego, stock-market-theory, the-job-title-director-sucks
  • 3 older images (replacing-macbook-ssd post) recovered from Wayback Machine since originals no longer on WordPress (WHAT?!)

The last one blew my mind - Claude was “smart” enough to detect that an image link was broken, and rather than giving up, it knew to go to the wayback machine and find it from a previous snapshot. That’s crazy. I don’t think I would have even thought of that.

The Results

You’re seeing it for yourself. Other than editing this blog post, I did no coding or conversion steps myself. Everything was automated via Claude Code.

Up next is writing a full app “hands off the wheel” style. More to come…