News for hackers. Stuff that matters.
Re:Google Search document leak reveals inner workings of ran... by eitland on Wednesday May 29, 2024 @09:35AM (#40511843)
Well, don't use it then.

I am happy for alternatives, otherwise I guess Kagi wouldn't improve so fast in areas I care about.

Re:Google Search document leak reveals inner workings of ran... by p3rls on Wednesday May 29, 2024 @09:59AM (#40512102)
I won't use it. I am pushing back on your belief that it has figured something out about anything, because to me it looks like Kagi takes Google's results and then makes them worse. I have total shit ranking #1 on Kagi and Google. On other pages I am struggling to break into the top 10 where I should be right under wikipedia in a just world.

I think many people whose lives don't revolve around these things (as in your finances/mortgage are not dependent on google search) get weird magical views about search engines where it's sorta like a modern delphic oracle where the magic works as long as you believe in it. Of course I have the opposite problem where Google is angry chthonic god to be supplicated/scorned.

Oh well, I do know I will not be worshipping a new god that cannot even destroy the abominations sitting at the top of the rankings.

Re:Google Search document leak reveals inner workings of ran... by bitcharmer on Wednesday May 29, 2024 @11:06AM (#40512722)
> Kagi takes Google's results and then makes them worse

And this is how I know you haven't used Kagi at all.

Re:Google Search document leak reveals inner workings of ran... by p3rls on Wednesday May 29, 2024 @12:19PM (#40513636)
> And this is how I know you haven't used Kagi at all.

I guess the identical search results in my niche are an example of convergent enshittification of both Kagi and Google then

Re:Google Search document leak reveals inner workings of ran... by bitcharmer on Wednesday May 29, 2024 @02:11PM (#40514958)
Ok, if that's the case then I give up. Totally possible for different people to have wildly different experience.
Re:Google Search document leak reveals inner workings of ran... by oooyay on Wednesday May 29, 2024 @05:29PM (#40517244)
> I won't use it. I am pushing back on your belief that it has figured something out about anything, because to me it looks like Kagi takes Google's results and then makes them worse.

Kagi is pretty clear about how they do what they do: https://help.kagi.com/kagi/search-details/search-sources.htm.... You're half right, except that Google is not a vertical that they leverage for their index. Their solution works for me, but I'm also a techy. Anecdotally, my partner uses it, and she is not a techy.

> Oh well, I do know I will not be worshipping a new god that cannot even destroy the abominations sitting at the top of the rankings.

The only way this statement applies (to me) is if you can optimize for Kagi. The only way I can think of to do that is organically, but admittedly I haven't put much effort into it because the magic of Kagi's flavor of search is that it puts me in near proximity to information density and sources that I already trust. The real proving grounds for Kagi are a decade out.

Re:Google Search document leak reveals inner workings of ran... by p3rls on Wednesday May 29, 2024 @06:38PM (#40517913)
Lol, it's just complete bullshit marketing on that page. But I do find it fascinating how they obfuscate what is actually happening behind the scenes and cannot come out and clearly say how their technology works, or that they are using google's index.

>Kagi is known for delivering a unique flavor of high-quality search results, sourced from our own web index (internally named "Teclis") and news index (internally named "TinyGem").

The marketing at least is pretty good -- I haven't seen this many committed devotees since the iPhone was announced.

But here on the ground, I can type a few queries and reveal that Kagi is the same shit as Google. Perhaps you are seeing some manually adjusted queries?

P.S. there are lots of other concerns about Kagi too, I'm just talking about the search https://hackers.town/@lori/112255132348604770

Re:Google Search document leak reveals inner workings of ran... by eitland on Thursday May 30, 2024 @05:22AM (#40521673)
> But here on the ground, I can type a few queries and reveal that Kagi is the same shit as Google.

Here on the ground in Norway I just say "lucky you" if Google serves the same results as Kagi.

My experience is Google expands my search terms beyond recognition and I am absolutely and utterly fed up with starting to read an article about a problem I searched for, getting a result from a trustworthy site, starting to read only to realize Google secretly behind my back expanded my search from "unpopular-ancient-javascript-framework-from-2021" to "currently-popular-js-framework-with-kinda-similar-name".

My time is valuable. Not having to babysit my search engine is something I happily pay for.

Could probably get my employer to pay for it too.

If Google by some miracle has you in an experimental group where they don't mess with your queries, or if you have time to babysit it - more power to you.

Re:Google Search document leak reveals inner workings of ran... by p3rls on Thursday May 30, 2024 @10:30AM (#40524197)
Why do Kagi supporters keep insisting I am supporting Google? I save all my best vituperative comments for how Google has destroyed the internet.

I am only saying Kagi is just as shitty as Google for my queries and certainly is using their index based off the 30 or so searches I made. I am not in a technical niche however, so maybe they're using something different for that or you saw a manually-adjusted query?

But for the regular, everyday people queries, yeah 100% it's just google bro.

Re:Google Search document leak reveals inner workings of ran... by eitland on Friday May 31, 2024 @12:45AM (#40531594)
Ok, thanks for the clarification.

Let me also clarify: I am absolutely not denying that Kagi uses Googles index.

I am just pointing out that for some reason when I search with Google I get lots and lots of results for things I didn't search for. For a while I kept them away by sprinkling doublequotes everywhere but the last few years, no amount of double quotes or even the verbatim operator works anymore.

With Kagi however not only does doublequotes work (for now at least), - I usually don't even need to apply them, my queries work just like they used to do in "old" Google back in 2009.

That is what I pay Kagi for.

I can be totally open that for what I know the reason might be as simple as the fact that Kagis API access gets results directly from Googles search API without passing through what I think of as a "search broadening" function and without applying A/B testing experiments to the results that come back.

I don't care. I get my results.

And if someone come along and present a better alternative that uses only their own independent index, I'll be inclined to pay even more just because I'd love to see it. (I was a marginalia supporter for months and might become again in the future.)