Syndicator.net Rehaul Swaps 34 Gemini Digest Posts for 12 Feed Guides

This morning I tore down the blog on syndicator.net and started it again. All 34 posts are gone. Twelve new ones are up, and from Monday a routine adds three a week, but only if they pass a quality check.

Some background. Syndicator is my feed aggregator. It pulls the newest items from 115 sites in my network into one river of headlines, and as of today that includes the new oqft.com. At the end of October I gave it a blog at syndicator.net/blog, on the same jollyroger template this site runs on. Then I hooked up a Gemini workflow to write it. The plan was a quirky first-person blogger, a vibe coder and digital nomad, picking stories from the Syndicator feed. One or two posts a day, a photo from my own media feeds, a few links. It started here on jollyroger.biz and moved over to Syndicator a day later.

For a week or two I liked it. Then I read a run of them back to back. The posts had nothing to do with feeds or syndication. Adobe Photoshop and quantum circuits in one post, bridges and caves in the next. Same cheerful voice every time, no point to any of it. Honestly, total garbage. A blog on a domain like syndicator.net should make the site look like it knows its subject. This one made it look like a content farm.

What got deleted

The workflow, its script and its state file all went, plus the CI rule that skipped builds for digest-only pushes. Then every post. I didn't set up redirects. The old URLs just 404 now, and I'm fine with that. There was nothing there worth sending anyone to.

What's there now

Twelve cornerstone posts, all about how content moves between sites. Some are plain explainers: RSS vs Atom vs JSON Feed, OPML files, WebSub, conditional GET. Two are news: Reddit ending its RSS feeds and Really Simple Licensing, which puts AI usage terms inside the feed itself. Three are about Syndicator, including how it's built with one Node command and no dependencies. One uses the aggregator's own data, what 114 sites show about Hugo and WordPress feed URLs.

The new routine

The blog now has its own content plan in the repo. Four lanes: feeds explained, syndication news, building Syndicator, network data. Never two posts in a row from the same lane. A scheduled task runs Monday, Wednesday and Friday, writes one post, and publishes it straight to the site with no draft step.

The gate before publishing is the part I care about. Each post needs a real news peg, a tested how-to, or numbers pulled from actual data. Every fact gets checked in the run. Code gets run. If nothing passes, the task publishes nothing that day and says why. I'd rather have a quiet Wednesday than another post about caves.

One smaller fix went in too. The GitHub Actions job that rebuilds Syndicator every 12 hours was set to run on the hour, and GitHub kept starting it hours late. It now runs at 02:17 and 14:17 UTC. Same idea as my old post on owning the publishing stack: you only notice the plumbing when it leaks.

Lesson from the Gemini run, for me anyway: automation is easy, a quality bar is the hard part. Without one you get volume, and volume on its own does nothing for a site.