How this site publishes itself
The pipeline behind workhacker.com: MDX files in git, agents drafting, humans approving, and a podcast page that reads its own feed.
Workhacker
We tell clients we run our systems on our own properties first. This site is one of those properties, so it seems fair to show the plumbing.
Content is files in git
Every post here is an MDX file in the site's repository. No hosted CMS holding the content hostage, no database to migrate. There is an admin interface for editing in the browser, but it writes to the same files and commits to the same repo. Git is the source of truth; the deploy is just a build.
That sounds austere until you want to know who changed a sentence and when. Then it's a git log away.
Agents draft, humans approve
Most first drafts here start with an agent. That's the part that got cheap. What didn't get cheap is the part where someone who will be embarrassed by a wrong claim reads the draft and decides whether it ships. Nothing goes live without that read.
This is the same shape we build for client systems: the pipeline does the volume, a person holds the veto. The review step isn't overhead on the automation. It's the reason the automation is allowed to run.
The podcast page keeps itself current
The podcast section of this site doesn't get hand-edited when an episode ships. It reads the show's RSS feed and renders what it finds — titles, dates, players. The feed is already the canonical record of the show, so the page treats it that way. One less thing to forget.
What's next
Season one of the podcast ran to 36 episodes. Season two is coming. When it arrives, this site will notice on its own — which is roughly the point of everything above.