Devlog: Blogging Tools - Post Classifier
📘 Devlog

Blogging Tools - Post Classifier

Decided to go all the way with Jev and TypeSafe, and worked with Claude to add a post classifier to Blogging Tools. This taps into the RSS feed ingester, already in place to poll my blog for new posts. When one is detected, it will pull the post title, body, and any image alt texts. It will send this to Jev alongside a set of questions, sent as Nouls, entered by the user.

Each question has an associated category, and two thresholds: an “ask” threshold and a “set” threshold. If the returned score is above the “set” threshold, the category is set on that post right away. If it’s below “set” but above “ask”, it will formulate an in-app notice asking if I’d like to set the category. Anything score lower than that is ignored.

Auto-generated description: A form titled Edit Question contains fields for Name, Question, Category, Ask threshold (%), and Set threshold (%), with entered values and a blue Save button.

Because the questions require some calibration, there’s also a test feature which will allow me to enter a URL, and it will run the classifier.

Auto-generated description: A webpage displays a test for extracting and classifying a post about a trip to Auckland and Waiheke Island, including the post's title, body, image alt text, and classification results.

I do have some other ideas for this tech. A sentiment classifier would be nice: something to track how often I’m complaining about this, verses talking about things I like. Maybe something to track how often I’m talking about certain topics. Too much about AI? not enough about Go? This will all be advisory of course, and I may not build any of them, lest I succumb to Goodhart’s_law. Let’s just see how the category classifier goes for now.