I trained an ML model on 15K headlines from the local (Dutch) populist newsletter and use that model to tweet a new headline every 30 minutes, for the past thre

I trained an ML model on 15K headlines from the local (Dutch) populist newsletter and use that model to tweet a new headline every 30 minutes, for the past three years. To make it look like tweets from the actual newsletter, I take the first few words, run those through google translate, and then use that to find an image on Unsplash to add to the tweet. The headlines frequently make no sense and often have grammatical issues. Yet a couple of times a month someone (often a boomer) interacts with the tweets. Their confusion bring me immense joy.