The thing that killed Google for me isn't the AI, but seeing issues that clearly indicate bad search engineering versus simply search engineering designed to ma

The thing that killed Google for me isn't the AI, but seeing issues that clearly indicate bad search engineering versus simply search engineering designed to make Google more money. The most obvious one was personalization. The pitfall of personalization is that a search system clings to a single past query or event despite all other data indicating it's an outlier. I've seen this on a lot of other search systems I worked on personally but I have not seen it on Google before. There's a lot of ways to deal with it if you have enough people and data which google definitely does. edit: The reason this is so indicative of a declining engineering culture to me is because over-personalization looks amazing on offline metrics. So if you have PMs driving focus decisions based on offline metrics they don't fully understand you get over-personalization. If you do it slowly enough then users simply deal with it so online metrics looks slightly positive. Then one day you realize you're now losing users and your search is useless. But every AB test was neutral or positive. And due to politics you can't roll anything back or even test rolling it back.