>especially in these days where ML can be used to statistically narrow down and pinpoint who wrote a specific piece of text only based on things like use of pun

>especially in these days where ML can be used to statistically narrow down and pinpoint who wrote a specific piece of text only based on things like use of punctuation, vocabulary, sentence structure and style. I think you can fight ML with ML - for example, use GPT-like algorithms generate text for you. But then you must also be careful about when you post - I remember some of the Russian trolls were exposed because their bursts of activity coincided with 9am-6pm Moscow time. So you have to use a random number generator decide when to appear online if you want to hide your location. There's always something which can narrow down their search. One small mistake and you are busted. They don't even need to pinpoint you exactly, if it's narrowed down to 1000-10000 people who meet the criteria, they already win because they have the capacity to go through the list one by one and eventually find you.