> But maybe our use cases are the outliers, and the new search is what works for most people.

I think it's a probability problem more often expressed in terms of medical tests / false positives. If a million people search for <word> but 1% of those make a typo and search for <wrod> instead, whereas a thousand people deliberately search for <wrod>, Google's observation of a search for <wrod> is 90% likely to be a typo.

However well-intentioned, this can undermine one of the core, original features of a good search engine like Google: deriving value from the "long tail" of esoteric (and often very technical) quests for information.