From the github repo it looks like it is using directly "google custom search engine", to index a limited set of websites https://cse.google.com (recognizing the css classes here https://github.com/txthinking/soso/blob/master/main.js#L41)

is this a google proxy?

My hunch is that it is a bit of this and a bit of that.

It does index some stuff that appear to indicate it's backed by google, you're getting Google Books results for example. They're also getting GitHub results, which you'll never get if you crawl on your own.

On the other hand there's a heavy skew toward publicly available datasets (wikipedia, stack overflow, reddit), which might indicate they're mixing and matching a bit with publicly available data. Could be wrong though.

Any new takes on search is ultimately a neat thing. I do think it's sadly under-explored design-wise. Google has been so dominant that nobody seems to think search can be anything other than a white page with a plain text input.

I've Google Translated the readme file in the GitHub repo and it says:

> At present, the content farm is flooded, and only sites with good experience are included in the whitelist.

https://github.com/txthinking/soso/

Based on that, my interpretation would be that it's an opinionated search engine that indexes a limited set of sites that it considers to be "good", ignoring everything else.

Looking at the javascript on the site or on their github[0], it indeed looks like they use Google programmable search engine for their results.

[0] https://github.com/txthinking/soso/blob/master/main.js

It looks more like Chinese government proxy.

Can't say I can read either the ReadMe or the code as it's not in English but it can be found here: https://github.com/txthinking/soso

every query just lead to social networks. kinda useless.

Well this seems pretty useless

the HN hug of death, site is down