A lot of these are so accurate it's uncanny.

Reminds me of the spam text generated for linkfarms and such --- probably because they use the same algorithm. It gives the same "uncanny valley" feeling of looking like it was written by a human, but upon closer inspection becomes much less believable.

Then again, some of them don't look "Markovian" enough and appear to be verbatim real article titles.

Is there an algorithmic way of detecting this? Sometimes it seems like some of the comments on HN could have been written by an ai.

I wish someone would write even half of these.