3 ms·

It's not just hacker news! They were up in arms about some actual analysis of search quality here: https://danluu.com/seo-spam/ https://danluu.com/seo-spam/

It's not just hacker news! They were up in arms about some actual analysis of search quality here: https://danluu.com/seo-spam/ https://danluu.com/seo-spam/ Another common response is >Your table is wrong. I tried these queries on Kagi and got Good results for the queries [but phrase much more strongly] I'm not sure why people feel so strongly about Kagi but, all of these kinds of responses so far have come from Kagi users. No one has gotten good results for the tire, transistor, or snow queries (note, again, that this is not a query looking for a daily forecast, as clearly implied by the "winter 2023" in the query), nor are the results for the other queries very good if you don't have an ad blocker. I suppose it's possible that the next person who tells me this actually has good results, but that seems fairly unlikely given the zero percent correctness rate so far. For example, one user claimed that the results were all good, but they pinned GitHub results and only ran the queries for which you'd get a good result on GitHub. This is actually worse than you get if you use Google or Bing and write good queries since you'll get noise in your results when GitHub is the wrong place to search. Of course you make a similar claim that Bing is amazing is you write non-naive queries, so it's curious that so many Kagi users are angrily writing me about this and no Google or Bing users. Kagi appears to have tapped into the same vein that Tesla and Apple have managed to tap into, where users become incensed that someone is criticizing something they love and then write nonsensical defenses of their favorite product, which bodes well for Kagi. I've gotten comments like this from not just one Kagi user, but many.

dotnet00 3y ago
I'm not really a fan of this person's dismissive attitude, they're trying to avoid dealing with domain rank adjustments, understandably so the results are comparable, but the adjustments are part of the value add. The testing methodology is also odd, looking specifically for yt-dlp when asking for "download youtube video", when the web based downloaders that most of the big engines turn up on top make more sense for the average user, both due to not having to install a command line tool and due to not necessarily being on a device the tool can be installed on. I've often used the top Kagi result, and it has worked fine, yet has been labeled as scammy by the evaluator. IIRC that site also ranks high in Google's results. For adblock they expect specifically ublock origin, and consider ABP to not count despite it also being a very popular blocker. Them complaining about why it's only Kagi users reaching out to him also seems unnecessarily aggressive, eg I've found their article through you, in the context of Kagi, so if I were to comment, it'd obviously be in that context. However, how likely is this article to be brought up in the context of Bing, Google, ChatGPT or DDG, where there are many more carefully thought out and less dismissive comparisons and discussions out there?
- itsdrewmiller 3y ago
  fwiw I agree with you that the evaluation criteria here are not ones that I particularly care about - I'm generally pretty happy with AI-first and then expanding to search engines if I can't get what I want there. My main intent was to note that Kagi users are just genuinely that enthusiastic, since GP at least lightly implied there might be some astroturfing going on.

dotnet00 3y ago

I'm not really a fan of this person's dismissive attitude, they're trying to avoid dealing with domain rank adjustments, understandably so the results are comparable, but the adjustments are part of the value add. The testing methodology is also odd, looking specifically for yt-dlp when asking for "download youtube video", when the web based downloaders that most of the big engines turn up on top make more sense for the average user, both due to not having to install a command line tool and due to not necessarily being on a device the tool can be installed on. I've often used the top Kagi result, and it has worked fine, yet has been labeled as scammy by the evaluator. IIRC that site also ranks high in Google's results. For adblock they expect specifically ublock origin, and consider ABP to not count despite it also being a very popular blocker. Them complaining about why it's only Kagi users reaching out to him also seems unnecessarily aggressive, eg I've found their article through you, in the context of Kagi, so if I were to comment, it'd obviously be in that context. However, how likely is this article to be brought up in the context of Bing, Google, ChatGPT or DDG, where there are many more carefully thought out and less dismissive comparisons and discussions out there?

itsdrewmiller 3y ago
fwiw I agree with you that the evaluation criteria here are not ones that I particularly care about - I'm generally pretty happy with AI-first and then expanding to search engines if I can't get what I want there. My main intent was to note that Kagi users are just genuinely that enthusiastic, since GP at least lightly implied there might be some astroturfing going on.