Disclaimer: I work at Google so I'm 100% biased. All opinions are my own.
Edit: formatting
>search engine is a cloud service, which is not controlled by user of that engine
So is hacker news.
>search gives final results in milliseconds (why? because google cannot spend many seconds/minutes for you, that's why)
Do you want latency to be larger??
>open source, because user should trust the code
That would make it easier for SEOs to game the system.
Or perhaps it would put everyone on a level playing field. I'm not sure.
>self-hosted (easily deployed in a click), not a cloud SaaS.
Do you mean users should keep their own index of the whole web by themselves?
>more useful than google because of accumulated data about you that are processed by computational knowledge engine
First I think you greatly overestimate how useful information about you can be.
Second if that worked then it would make they problem you mentioned before (information bubble) worse...
Instant results are good for sure, however very often few more seconds - in addition to instant results! - is not a problem if late, more carefully processed results can save minutes of my time - for now I spend it for opening the links and scanning the content with my eyes.
The same is about not very often but important searches that may be described as 'research about something', in this case I'm ready to make complex, well detailed query and wait even hours - then back and get well organized and intelligent results.
> Do you mean users should keep their own index of the whole web by themselves?
oh no! Users should keep only their personal data - in wide meaning, this includes all history of searches, search results, refinements, anything that ML currently uses to bring personal search experience. In addition to that, relatively small index of important content may be saved. For internet search this 'personal search' will use API of anything that can be used manually for now - google, bing, consume direct API of Twitter/FB/Medium/WolframAlpha and hundreds of connectors to other cloud services. It is important to say, that this 'delegated search calls' may be anonymized.
It will be important that search results are not limited only by what google decided to be 'top results for this user'. At this moment I can do all this manually - open N tabs, query many services, compare results, open most 'relevant' (from my human point of view, not google) links and scan them for most interesting information. I believe that all this can be automated.
> First I think you greatly overestimate how useful information about you can be. Second if that worked then it would make they problem you mentioned before (information bubble) worse...
As for now, all this just thoughts. I'm a programmer with almost 20YOE; I have understanding about how google works in general, how lucene works, how WolframAlpha works, modern approaches to NLP and search-driven queries processing, and I think - without a MVP that works, this is more belief, of course - that value of this 'personal computation engine' combined with modern ML approaches might be ultimate. Challenge, but nothing impossible!
>search engine is a cloud service, which is not controlled by user of that engine
So is hacker news.
>search gives final results in milliseconds (why? because google cannot spend many seconds/minutes for you, that's why)
Do you want latency to be larger??
>open source, because user should trust the code
That would make it easier for SEOs to game the system. Or perhaps it would put everyone on a level playing field. I'm not sure.
>self-hosted (easily deployed in a click), not a cloud SaaS.
Do you mean users should keep their own index of the whole web by themselves?
>more useful than google because of accumulated data about you that are processed by computational knowledge engine
First I think you greatly overestimate how useful information about you can be. Second if that worked then it would make they problem you mentioned before (information bubble) worse...