It was first announced at a TED talk in 2018. 5 years later,
the output is just "close approximations to the words you use
in your query", and based on 120,000 books scanned into
google.books. The results are pretty fast, although it only
matches 3 or 4 books at a time.