I scanned the list and attached a subjective clickbaitiness to each 'headline'. Of course without reading the article. Yours got 100%.
I learned two things, this network knows how to make HN visitors click. It is hard to extract knowledge from a NN.
Some pretty hilarious and/or thought-provoking headlines! LSTMs capture so much about the input (compared to N-grams, say), it can be almost eerie.