>Speaking as a person who spent some years if his career working on full-text search, I doubt that there’s much left in the way of low-hanging fruit.

Maybe so, but judging by my search results most of the fruit is still on the tree. Natural language understanding is an active field of research.

Turning Google into a search utility is a bad idea. Splitting off the crawling part and turning that into a utility would be a lot more interesting. It would allow search engines to compete on search quality while sharing the capex required for crawling.

> most of the fruit is still on the tree

Case in point: image search for "shirt without stripes":

https://github.com/elsamuko/Shirt-without-Stripes

HN comments: https://news.ycombinator.com/item?id=22925087