seen a lot of these articles, what I'd be interested in is some proof that any of this works better than vanilla prompting -- repos would be a good indication of the code quality, as well as some professional postmortems.
the thing is, to get the proof you'll need to evaluate it empirically, in other words, practicing good context engineering ;)