Article resonates with me but I think it's worth mentioning that it's not an either/or situation.
Nothing wrong with using XGBoost, CatBoost et al to quickly identify features that might be worth digging into even if your actual goal is to come up with a statistical model based on understanding the domain.
I had to switch to light mode at the OS level to be able to read the image text in the included proof. I wish Mataroa would let you swap that on the site and not determine it based solely on your OS settings.