For someone using sed often enough inventing prettypath won't make sense. However, if producing correct sed command, be it by remembering the options, reading manual or digging through the shell history, takes some amount of mental effort, your brain will happily stick "prettypath" into memory as long as doing so stays less mentally taxing than doing original task from scratch.
Personally I only ever needed it once. I was re-implementing javascript function doing some strange string processing by using characters in the input string to calculate indexes of alphabet array to replace them with. Since I was using Python I just imported string.ascii_lowercase instead of manually typing the sequence, and when I showed the code to someone more experienced than me, I was told it's base64, so all my efforts were replaced with a single base64.b64_decode() call.
If I were to choose between bearing the fruits of 20 years of practicing music and equal amount of time spent playing this kind of games, it would certainly be the latter, and I did try both.
I don't really get the point of disabling Youtube history just to never visit the main page again anyway. That said, recommendation algorithm is actually a valuable asset and it doesn't seem wise to discard it entirely instead of learning how to use it.
Imagine for a second, that you only care about watching cat videos, and arrange your viewing behavior accordingly. The algorithm will be quick to realize your preferences, and your entire home feed will be filled with nothing but kittens. Occasionally it will try to feed you some dogs and such, but Youtube is generous enough to provide "Not interested" and "Don't recommend channel" buttons on every video in recommendations, which are a great tool in shaping the home feed, or forming a nice little information bubble if you will. As for all the horrors of mainstream recommendations described in the article, they would look so out of place in your feed it won't take much mental effort to ignore them (or get rid of with aforementioned buttons).
Occasionally, you might want to watch something other than cat videos, or get sent a link in a personal message. For everything you don't want the algorithm to get the wrong idea about, use separate browser context or private mode.
Now, let's say you're got diverse tastes and like both cats and heavy construction videos. With two topics the Youtube algorithm might still manage to stay on track, but if you try to add more, the feed will quickly degrade to a mess. There is, however, a way around it: don't mix different topics in a single Youtube account. Youtube allows making multiple users on a single Google account, and with container tabs it doesn't even require switching between them.
Of course, and this is one part of the article I can agree with, use the subscribe feed if it fits your workflow, but certain kinds of content are made mostly by small channels on irregular basis, and the algorithm is a valuable tool to discover new ones.
Was able to correctly guess it's about Therac-25 from title alone.
I can't help but think this specific story is being overtold, while other software-related accidents doesn't get enough attention.
Software controlling a safety-critical equipment doesn't even need a bug to cause deaths, bad user interface can do it just fine. "Behind Human Error" gives an example of device used to assist a surgery, that required complicated procedure to set up, which had a step unrelated to device functions in context and was easy to forget (rendering the device non-functioning) and would not indicate in any way it wasn't working when it was expected to. As for a more modern example, many of recent aviation accidents involve pilots failing to correctly interact with cockpit systems.
In the Therac-25 case the core issue (leaving aside procedures put in place to prevent the exact thing) was an error in the code, that could eventually be found and corrected, but software with badly designed interface can kill people while working exactly per specification, shifting the blame on the operator.
If you are lucky you might get an image that hasn't yet been shown to enough people for the system to learn the "correct" answer, in which case you get a free pass submitting even blatantly incorrect solution.
The scraping and reuploading issue could be solved by some kind of universal global content identification system, integrated into the micropayments system, making sure no matter where certain piece of content is uploaded, the fee would still go to the copyright owner, perhaps with some small percent given to the hosting website. Not saying it would certainly work, but there is a few technologies probably everyone here have heard about that seems like a very good fit for the task.
Playing video on Youtube causes much more cpu load than playing the video with the same av1 (or even more computationally intensive vp9) codec in VLC.