Option to disable Quick Answers

bv

I also signed up just to give feedback on Quick Answers. They make Kagi unusable and infuriating, and if they are not removed, I cannot be a Kagi customer. I expect Kagi to surface the best of the web, not shovel out AI-generated garbage.

j-james

Vlad I think the point is that there is a huge subset of users - myself included - who do not want ChatGPT responses in their results at all, no matter if the scope was limited. It might be useful to hold off on asking for feedback for auto activation scope until the option to disable it has landed (otherwise you're going to get a whole lot of "never").

Vlad

j-james I see that now. Going into this we had a very strong signal that users wanted this, there is a very highly upvoted feature to even include it for every single query https://kagifeedback.org/d/1892-option-to-have-quick-answers-always-on

Sort of shows perils of user driven feedback development and you can not make every PM decision right.

We enabled this for probably less than 0.01% of all queries but this seems to have provoked strong reaction against it that we previously did not hear from users (I would venture to guess that inclusion of 'how' queries caused most of the dissatisfaction).

No worries we will fix this, just need a day or two.

Browsing6853

Vlad

Vlad Sort of shows perils of user driven feedback development and you cna not make every PM decision right.

I think it just wasn't implemented correctly. IIRC the "Quick Peeks" feature was implemented and disabled by default because it was a significant addition to the results page. The automatic trigger of quick answer should've been implemented in same way - first add the toggle button, them enable/disable it for everyone,

j-james

Vlad Yeah - I'd say part of it too is that users actively engaging with Kagi are interested in what Kagi is working on, which right now, has been LLM stuff for the better part of a year. I was way more active on here back in the beta, but with ranking and lenses and bangs Kagi is pretty much feature complete for me, sans a https://web.archive.org/web/20170404173014/https://duckduckhack.com/ kind of system.

CrunchyFritos

Vlad Do you guys have a beta-testing program? I would happily sign up to test out new features before they are rolled out to check for bugs / give basic feedback. I know that in theory Assistant is in "beta" for Ultimate users, but maybe a separate opt-in program for users willing to test out new features that might disrupt existing workflows? Maybe some of the very strong responses here might have gotten caught, though maybe the extra admin overhead isn't worth it. Just a thought.

Vlad

Browsing6853 Just fyi Quick Peeks was enabled for everyone by default too.

wrkyle

At the risk of "piling on", I hate this and one of the reasons I pay for Kagi is to have a search engine that isn't employing LLM-generated, zero-provenance search. Not being able to turn it off was also an unwelcome design choice. I don't need or want search tools that train my brain to seek out and uncritically accept unsourced answers to questions. Principles aside, what does this add to the functionality of search? Is LLM-generated text better than sourced answers? Why add this to your product rather than finding ways to surface more authoritative, accurate pages and highlight those answers? Does this make your product better or is it about cultural currency and "keeping up with the big players" even if they're making bad decisions to follow the hype train.

Vlad

wrkyle

Yes users want this. See https://kagifeedback.org/d/1892-option-to-have-quick-answers-always-on

Also many shared they love it today

The challange is that not all users do. This is why we are very careful when we roll things like this out, and less than 0.01% of queries (my guess) were affected by this.

The proper way to impact product roadmap is to voice your opinion like you just did. This sounded like a good idea but was probably overexposed. We are not chasing any hype, we are very user-centric, and sometimes things need a bit of time for equilibrium.

petterroea

@Vlad Thanks for the merge. I want to clarify that it is the post-load layout change that causes 90% of the irritation(for me). I have yet to get any use for GPT other than "give me the word for this special context", but please keep experimenting. Just please don't be one of the actors that thinks breaking layout post-load is ever acceptable, it is far to common nowadays.

Vlad

petterroea I agree, we should have fixed height.

ChristenGottschlich

petterroea

Anything that interferes with muscle memory/routine is super painful.

Oh my god its so nice to hear another human being say this.
Sometimes it feels like im taking crazy pills with all this love lately for ~ dynamic ~ websites and ~ adaptive ~ tools.
Starting to feel less like im using a tool and more like i am the tool.

steog32

Vlad Really appreciate the quick response and turnaround time to getting a fix out. I understand it's an easy mistake to make with a small team- product decisions are hard and I can see how this could happen given how noisy the LLM people are (seemingly crypto-bros in 2018/2021 levels these days, sheesh).

Anyways, so long as there's an option to keep all of the LLM features turned off for my account I'll remain a happy Kagi customer. Looking forward to the patch to be out tomorrow.

Clank

Wrote a rant, counted to ten, and then deleted it as it wasn't terribly constructive and there's no need for a pileon. Instead I'll just say that the reason I pay for Kagi is exactly so I can trust my data won't be used without my consent for AI or any other adventurous side hustle, and if the trust goes so does my subscription. Looking forward to the feature to disable this entirely arriving ASAP.

(ETA: And for clarity, I have absolutely nothing against generative AI in principle, hell I'm doing my own experiements to see what it's capable of and where it can be useful to me. But there's a huge difference between doing your own experiments, and enrolling me and my data in your experiments without my consent, that's the matter of principle.)

snowytrees

Some constructive feedback after we get a quick fix for disabling my painpoint with it is this from the changelog:
now simply starting your query with an interrogative word (what, where, who, which, when, how). Many of my queries use an interrogative word but I’ll switch to assistant it if I want AI to answer. Let me pick the trigger start/end word (such as ?) and i’ll be happy.

Also the lazy loading is a big issue causing content to move. A selling point of kagi from the early days was no JS compatible/basic html web page. The lazy loading UI is the antithesis of that sadly.

Talran

Would it be possible to have two settings for this feature?

I generally don't want the instant quick answer and I frequently want to go fishing for guides with the interrogative words in my query. So I would like to disable those triggering the quick answer feature.

But I could see trying out the use of ? as a shortcut to trigger it when I am intrigued to see what it comes up with before looking at the results. It's much easier to skip using that punctuation without having to think too hard about my query when I know I don't want to trigger it.

Peter_S

My first post here - just to say I signed up for Kagi recently, and a large part of that was because of the quick answer feature. It's great to have the option to have it triggered automatically, saves me having to click for it. Results shifting unpredictably down the page is the only issue (and fairly major one) I have with it, but seems like that will be fixed soon. As long as it's an option to toggle on/off then hopefully everyone can be happy 🙂

CrunchyFritos

Peter_S Yeah, I also really like the feature and am glad it was implemented. I agree the post-load layout change is annoying, but I find the feature in general pretty helpful.

Igak

I understand the need of AI being rammed into everything and everyone down to the point of choking, for VC funding purposes if not anything else, but please add a toggle to disable. You won't have to tell the VC folks.

migelius

I appreciate the Quick Answers feature and am glad it's been added. I would like the presentation to be improved so that it does not push down links as it is creating its response. This would be preferable to disabling it entirely.

« Previous Page Next Page »