The article make notable mistakes:
1) Yes, you have more/big data, but what you need is labeled data which is different. I.e. it does not matter if I have 100 picture of cats or 100000 picture of cats, if this is the same cat.
2) Feature store are not for feature engineering, more for feature discovery and sharing, feature engineering is a different area of concern.
3) Not sure about the fact that data changes so much. Yes some domain it might , but some it does not.
