I had a similar idea: build a kind of "stackexchange" of search engines. You start with a "generic" search engine infrastructure, and each community runs its own instance, tweaked accordingly. This would mean:

* The community chooses what goes in the search database, the rules for the crawler etc.

* People in the community can vote on stuff, etc.

* The engine can be customized to have some "semantic" understanding of what is scraped (i.e. on a math-oriented instance, it would understand latex, in a cooking one, it would be able to parse recipes if they respect schema.org).

I really believe in this kind of concept of "user curated, community oriented" search engine, since this means taking pretty much the opposite approach to what google does, thus:

* this wouldn't compete on google's own ground ==> higher chances of success.

* you could keep more control on the data

* the engine wouldn't pull any tricks on you by trying to overfit what you meant

* lots of customization options, etc.

But I never had the energy to try and start something with the idea…