Most active commenters
  • freediver(5)
  • Etheryte(3)

←back to thread

270 points lkellar | 18 comments | | HN request time: 1.014s | source | bottom
1. Etheryte ◴[] No.41873306[source]
This is a solid idea, but I wish they addressed the elephant in the room — image search is by far the weakest part of Kagi. For a considerable portion of queries, a large part of the results isn't relevant. If you use filters, they're often ignored or don't apply correctly. Many images are tangentially related at best. The list goes on. I've been paying for Kagi for a long while, yet I've seen nearly no improvements on this front. Image search is one part of their product where I often go to Google or other options because what Kagi does there just doesn't cut it.
replies(5): >>41873394 #>>41873578 #>>41873589 #>>41873830 #>>41873847 #
2. yzydserd ◴[] No.41873394[source]
I agree. I’ve often wondered, what could be the possible reason given Kagi is using google APIs behind the scenes?
replies(1): >>41873783 #
3. freediver ◴[] No.41873578[source]
Would you mind being specific and posting a search quality issue with a concrete example(s) to kagifeedback.org? We are keen to address the issues you are seeing.
replies(2): >>41873618 #>>41873809 #
4. huesatbri ◴[] No.41873589[source]
Yandex image search is really good.
replies(1): >>41873703 #
5. Etheryte ◴[] No.41873618[source]
I actually did post on your feedback site nearly a year ago, still no feedback or response on it so far [0]. It's been marked as under review, so it must be one thorough review process.

[0] https://kagifeedback.org/d/2565-image-search-doesnt-respect-...

replies(1): >>41873754 #
6. 42lux ◴[] No.41873703[source]
Probably the best tbh
7. freediver ◴[] No.41873754{3}[source]
Thanks, replied in the thread.
replies(1): >>41874141 #
8. freediver ◴[] No.41873783[source]
Do you have any example to share? (we do not hear this feedback frequently so want to make sure we address it, thanks!)
replies(1): >>41873908 #
9. Topgamer7 ◴[] No.41873809[source]
I'm a subscriber. But I don't feel like creating a signup for kagifeedback.

https://kagi.com/images?q=https%3A%2F%2Fs3.amazonaws.com%2Fp...

Lets use this as an example. I would personally like to have a list of exact results. Separately - a list of similar images would maybe be nice. But tbh, 99.99% of the time with reverse image search, I am trying to play detective, not find similar images. I am usually looking to see the first, original source of something. Or maybe other places I can find this image.

I will point out that detective stuff like this is crucial to try to prevent being catfished, or phished. I am not ignorant that it is also a concern for those who don't want to be found, for privacy or safety reasons; however some threat actor could just find a less public reverse image search I'm sure.

replies(1): >>41873891 #
10. dingnuts ◴[] No.41873830[source]
Well, your anecdote is completely contrary to mine. Image search has always worked great for me and it's easier to save the image because Kagi doesn't play games with the source of the image like Google Images does.

It's Google Images that I find unusable.

11. dmonitor ◴[] No.41873847[source]
Image search is the one area in Kagi that I've seen the most improvement over the past year. When I first subscribed, I'd often switch to Google to find what I was looking for. Nowadays it works exactly as I intend.
12. yjftsjthsd-h ◴[] No.41873891{3}[source]
> I'm a subscriber. But I don't feel like creating a signup for kagifeedback.

It's a separate login? That's really weird, isn't it?

replies(1): >>41874172 #
13. AndroTux ◴[] No.41873908{3}[source]
It’s hard to put it into solid feedback and examples, because it is highly subjective. But I also find the image results from Kagi lacking, while I really enjoy the text based results. Especially for more specific queries, the image search just doesn’t hold up.

I just tried generating an example. Take the query “screenshot nero burning rom windows xp” - of the first 10 images, only 6 are screenshots of the program on Kagi. On Google, it’s a solid 10/10.

Of course it’s hard to take just one example, but it reflects the general feeling I have when using the image search quite well. The results aren’t necessarily terrible, it’s just that they aren’t as relevant as Google’s.

replies(1): >>41874130 #
14. freediver ◴[] No.41874130{4}[source]
I see both Kagi and Google have 9/10 screenshots of Nero.

To make it easier for you to report any discrepancies I created a bug report with screenshots of what I see.

https://kagifeedback.org/d/5073-investigate-image-search-res...

15. Etheryte ◴[] No.41874141{4}[source]
That's kind of you, but if you have to resort to HN comments as your feedback system, then your existing feedback system doesn't really work, does it?
replies(1): >>41874257 #
16. tensor ◴[] No.41874172{4}[source]
Not really. It's not that uncommon for a support/forum/feedback site to use a separate account. It just means that they didn't have time, didn't want to spend the money, or couldn't link the support software to their main user account system.
replies(1): >>41874368 #
17. digging ◴[] No.41874257{5}[source]
I do not think this is a case of that happening like with other big tech names treating customers like shit until they make it onto HN. Vlad's super active (like, is he ok?) on the Orion browser bugs/feedback forums so I have to assume the same is true of Kagi, their actual money maker. Sometimes things fall through the cracks. One data point isn't a trend.
18. freediver ◴[] No.41874368{5}[source]
> It just means that they didn't have time,

That's it!