The "most popular" (and maybe click-bait like) the content, the more crap you'll get. Especially when people think they know the subject but they don't.
Most common offender nowadays: ML and Data Science. I've found articles with glaring errors, techniques that don't make sense, very shallow tutorials, etc (Disclaimer: nobody is immune from making mistakes, not even me)
And then you get "oh but you start training and then your loss go to zero, right?! This means it works, right?" Yeah. No.
If your training loss goes to zero, your are probably overfitting or the problem you're trying to solve is too simple. If your testing loss goes to zero, you're probably doing something wrong. Or my favourite: your loss is very low but what you're trying to predict has result A 99% of the time and result B 1% of the time. What's your loss if you only predict result A?
I mean, if even the myths about /dev/random and /dev/urandom keep circulating around, can you imagine about other stuff?