I gave it a shot. It's interesting for sure, and I appreciate just how many more languages there are to choose from than DeepL. However, as they say in the blog post they use LLMs for translation....
I gave it a shot. It's interesting for sure, and I appreciate just how many more languages there are to choose from than DeepL. However, as they say in the blog post they use LLMs for translation. Which would be fine on the face of it, but I think because the AI has restrictions on curses and swears the translation is Ned Flander-ized sometimes.
In some cases the translation is just as vulgar as the input, in some cases it's censored with asterisks and in others it's just mistranslated to be less vulgar, "damn" instead of "f**k" etc. It's not very consistent.
Considering that it happily spits out many standard curses, I'm guessing that it's not necessarily restrictions like swear filters or instructions in the system prompt, but rather that its...
Considering that it happily spits out many standard curses, I'm guessing that it's not necessarily restrictions like swear filters or instructions in the system prompt, but rather that its training data contained censored words as well as uncensored ones.
I personally think that translation is one of the few suitable use-cases for LLMs (they do it pretty well and support an incredible amount of languages), so I'm glad to see services like this being implemented instead of shoving AI in random places.
I'd be very hesitant to use it for smaller languages that aren't supported by other services. There's a reason DeepL started with a small number of languages and then expanded, and it's extremely...
I'd be very hesitant to use it for smaller languages that aren't supported by other services. There's a reason DeepL started with a small number of languages and then expanded, and it's extremely well-known that machine translation (whether LLM-based or not) tends to get way worse in quality with languages that don't have a lot of high-quality training data. All machine translation should be used mindfully, but it's extra worth paying attention to quality vs quantity when it comes to which languages you support.
That said, I'm happy to see more competition in the space. Machine translation can be super valuable for people and has saved my bacon plenty of times in the past.
This feels like another instance of Kagi actually wanting to be an AI company... I recently came across this blog post and this announcement reminded me about it again.
This feels like another instance of Kagi actually wanting to be an AI company...
I recently came across this blog post and this announcement reminded me about it again.
They always have been Source: History | Kagi's Docs
They always have been
In the first couple of years (2018 - 2019) Kagi was focused on AI technology, driven by the notion that any future search engine will need to have a strong instant question-answering component.
As someone that works a bit more with GDPR data requests etc, that blog post seems pretty far off the mark. I don't think the author is speaking from a place of authority. Kagi's approach to...
As someone that works a bit more with GDPR data requests etc, that blog post seems pretty far off the mark. I don't think the author is speaking from a place of authority. Kagi's approach to privacy of "collect and store as little information as possible" is pretty much best practice. It's possible Kagi should provide a checkout page to let you download a CSV with all user configured settings, but I'm not sure it's legally necessary.
The t-shirt thing does seem silly but generally overblown. I've seen far more egregious wastes of money than $200k on free t-shirt fulfillment for early supporters.
I mean, the author is pretty clear that they are just writing down their personal experience and are not a journalist. Having said that, it seems you have glossed over things like the "email is...
I mean, the author is pretty clear that they are just writing down their personal experience and are not a journalist. Having said that, it seems you have glossed over things like the "email is not pi data" bit, which in the context of GDPR very much is.
Finally, it isn't one thing specifically in that blog post that made me take notice. It is the accumulation of various things that does it for me.
I was taken in by what the author was saying here and close to cancelling my subscription — but I read his email exchange with the CEO and it didn’t come across nearly as confrontational or...
I was taken in by what the author was saying here and close to cancelling my subscription — but I read his email exchange with the CEO and it didn’t come across nearly as confrontational or deluded as the author made it out to be. I thought the point by point rebuttal was straightforward and persuasive and matched up with the stances on these things I’ve seen the company take publicly in the past.
I mean, I'm not here to shill for the CEO, but the blog post clearly gained some traction so I could see why he would reach out. I do think responding after the author specifically said "I am not...
I mean, I'm not here to shill for the CEO, but the blog post clearly gained some traction so I could see why he would reach out. I do think responding after the author specifically said "I am not interested in getting more replies from you on this subject. Declining a call does not mean I want you to argue with me about Kagi in email either" is iffy at best and disrespectful at worst, but I don't think the tone or contents of what the CEO was saying specifically were disrespectful, and I found what he was saying to be a pretty credible response to what was said in the post.
I read the email exchange (thats the author himself posted) after I read the blog post, and my conclusion was that the blog author is an asshole. The conversation pretty much went: Author: Kagi...
I read the email exchange (thats the author himself posted) after I read the blog post, and my conclusion was that the blog author is an asshole. The conversation pretty much went:
Author: Kagi sucks
Vlad (CEO): Hey can we talk about it?
Author: NO FUCK YOU FUCK OFF
So it's less about the specific "don't messagee anymore" and more about the rancid attitude. Also if the author didn't really want Vlad's messages, he could've blocked his email, simple as that. The mere fact he didn't points to Author being a drama queen.
I find your take oddly out of context if I am being honest. I am not sure how you make this: Out of what was actually said, which is: I will give you that the last bit sentence is a bit snippy,...
I find your take oddly out of context if I am being honest.
I am not sure how you make this:
Author: NO FUCK YOU FUCK OFF
Out of what was actually said, which is:
Thanks for reaching out, but no, I would not. I am not interested in being cornered into a call by the owner of a business because I made a blog
post about it.
I will give you that the last bit sentence is a bit snippy, but it is far cry from your capital case "fuck off" take. It also is not an out of the blue response, as the author of the post has interacted with the CEO before on discord and has seen how they conduct themselves. And again, by their own admission they are not a journalist, the blog can barely be called a blog and they only made it because some people kept asking about why they stopped using Kagi.
Context is indeed important, and I definitely do not claim the author literally said what I implied. I only exaggerated how his words felt to me at the time, that is, intentionally obtuse and...
Context is indeed important, and I definitely do not claim the author literally said what I implied. I only exaggerated how his words felt to me at the time, that is, intentionally obtuse and provocative. In that since, context explainsy choice of words summarizing the exchange.
Again, the author could've simply blocked the CEO, or ignored him, or anything, but the fact he didn't should imply something.
Personally, I find it generally polite to respond to people, even if it is to say “no” and even if it is online. I find it even more polite to take “no” for an answer, regardless of context.
Personally, I find it generally polite to respond to people, even if it is to say “no” and even if it is online. I find it even more polite to take “no” for an answer, regardless of context.
If it needs to imply anything I would say that the author has been part of the discord, has had conversations with the CEO and effectively gives one big massive hint towards the CEO saying he...
but the fact he didn't should imply something.
If it needs to imply anything I would say that the author has been part of the discord, has had conversations with the CEO and effectively gives one big massive hint towards the CEO saying he should do some reflection before attempting to debate it out.
Should they have replied? Maybe not, but other than that it becomes very clear the CEO doesn't take no for an answer, I am not sure if it implies much else.
I’m curious, what AI companies are you comparing to Kagi? An “AI Company” rightfully has a negative connotation, though I disagree that Kagi is an AI Company. Their product is their search engine,...
I’m curious, what AI companies are you comparing to Kagi? An “AI Company” rightfully has a negative connotation, though I disagree that Kagi is an AI Company. Their product is their search engine, with AI tools offered, not forced. It may feel like I’m splitting hairs, in this case I think the words and connotation matter.
I think the risk of paying for Kagi is minimal. If Kagi starts to suck, I’ll stop using it. Kagi Search is profitable. There is no other significant revenue stream. The privacy policy
makes clear there is no logged searches, no data to profit from.
I am not, it just seems to me with all the AI features they are focussing heavily on that this seems more of the thing they want to focus on than just plain old searching. I think it is quite...
I’m curious, what AI companies are you comparing to Kagi?
I am not, it just seems to me with all the AI features they are focussing heavily on that this seems more of the thing they want to focus on than just plain old searching.
It is a bit melodramatic.
I think it is quite reasonable and far from really dramatic. But it is very clear people read these sorts of things differently.
I was never entirely convinced why I should trust this "trust me bro" assertion over any other. Other than vague hand waving along the lines of /why would they though?/. Sometimes the so-called...
The privacy policy
makes clear there is no logged searches, no data to profit from.
I was never entirely convinced why I should trust this "trust me bro" assertion over any other. Other than vague hand waving along the lines of /why would they though?/.
Sometimes the so-called independent audit that they paid for is cited. But that was a long time ago now.
Privacy policies are legally binding in the US. But yeah, it really is just putting trust in the company’s word, its employees, their security, and the consumer protection institutions that...
I was never entirely convinced why I should trust this "trust me bro" assertion over any other
Privacy policies are legally binding in the US. But yeah, it really is just putting trust in the company’s word, its employees, their security, and the consumer protection institutions that enforce privacy policies. Good enough for me, personally. I understand the skepticism though.
Edit: I read this comment over again and I think it comes across a bit sarcastic but I meant it in earnest
It makes no business sense for Kagi to lie about this. They have a stable income from subscribers, which will dissappear immediately if it was leaked they were lying. I seriously doubt any data...
It makes no business sense for Kagi to lie about this. They have a stable income from subscribers, which will dissappear immediately if it was leaked they were lying. I seriously doubt any data they could collect would be worth so much money for that risk, both financially and legally, to be worth it.
That aside, Google and Bing are with 100% certainty tracking you, so I would rather pick one that doesn't have a financial incentive to track or sell data.
I must say, I use the Quick Answer feature a lot. I find it super handy for a lot of things. But yeah, they are definitely leaning into AI. edit: I've read that blog post now. I had some...
I must say, I use the Quick Answer feature a lot. I find it super handy for a lot of things. But yeah, they are definitely leaning into AI.
edit: I've read that blog post now.
I had some interactions with Vlad and the team in the early days exactly about their privacy statements, and gdpr, and was met with much the same response, and it took the (perceived, to me) slide in ddg results quality over 18 months or so for me to decide to give them a try.
I have been happy enough with the results, which I guess is /fine/. And as I say, I find Quick Answer useful. Didn't even know about the other services offered and proposed.
I've kind of always had a spidey sense tingle about Kagi, but never could put my finger on it, but that blog post nails it.
As one of the mentioned, ahem, dedicated types who (very occasionally) “hangs out” in their Discord – Vlad has mentioned in the past that they have a more or less neutral stance towards AI...
I must say, I use the Quick Answer feature a lot. I find it super handy for a lot of things. But yeah, they are definitely leaning into AI.
As one of the mentioned, ahem, dedicated types who (very occasionally) “hangs out” in their Discord – Vlad has mentioned in the past that they have a more or less neutral stance towards AI features/LLMs specifically; but it’s (a subset of) the users requesting them. They need this stuff implemented to be unique, or at the minimum competitive, in the search business. Since the “AI” portion of features can pretty much all be ignored without effort or downside and aren’t really shoved in your faces anywhere in the core product, I don’t mind that logic, in fact, I think it’s a pretty reasonable approach.
I had some interactions with Vlad and the team in the early days exactly about their privacy statements, and gdpr, and was met with much the same response, and it took the (perceived, to me) slide in ddg results quality over 18 months or so for me to decide to give them a try.
That said, it does feel like they’re throwing a lot of experimental features at the eager user wall to see and figure out what might stick. But: I don’t think that’s a bad thing! Their business model will never scale as much as e.g. Google’s, since the overwhelming majority of people will never become paying users. So why not try to capture them with a browser/summarizer/translation/email/whatever other feature offered and try to convert and or upgrade from there? Hell, that’s my story too, and I was convinced and ready to subscribe before using up the 100 free searches you get.
I have been happy enough with the results, which I guess is /fine/. And as I say, I find Quick Answer useful. Didn't even know about the other services offered and proposed.
Thinking about it for a bit, I think I stay subscribed because of the “clean” results. No SEO spam, surprisingly few LLM slop “articles” and help pages/fake guides, and no ads. Zero. Not a single sponsored search result, no flashy unrelated product imagery, just peace of mind when browsing.
I've kind of always had a spidey sense tingle about Kagi, but never could put my finger on it, but that blog post nails it.
Opposite for me (and to be clear, that’s completely fine!). Kagi actually became a Public Benefit Corporation recently. If it weren’t for that, I could see how superficially you could get the “tech bro” impression from them. But to me, this sealed the deal. Zero marketing expenses/only word-of-mouth growth and a nice product and all these additional features are cool, sure, but what tech company in this day and age actually dedicates themselves to a goal that’s not “grab as much VC cash as you can, then burn all users” (or their experience)?
You know how the shirt has the little stitched on kagi tag on the hem, both mine just fell off, not properly stitched on. Both on the yellow, and black ones. And yeah, I'm a sucker for a t shirt :-/
Opposite experience for me. :-)
You know how the shirt has the little stitched on kagi tag on the hem, both mine just fell off, not properly stitched on. Both on the yellow, and black ones.
Same here, and I was very sceptical about it when they announced it. In general, I really don't want AI in my search. I use search engines to look for information, not answers. But Kagi's...
I must say, I use the Quick Answer feature a lot.
Same here, and I was very sceptical about it when they announced it. In general, I really don't want AI in my search. I use search engines to look for information, not answers. But Kagi's implementation kind of works for me.
I notice that this has actually been quite typical for me with whatever Kagi announces. My first reaction often is negative, but when I look into the new feature a little, I can usually see why they are doing it. I still don't really use any features other than the core search, but I don't mind that they are experimenting, if it makes sense from their point of view.
It's actually a bit different with this translator thing, which I am very interested it. Unfortunately, at least based on my first tests, the results and the user experience aren't quite there yet. Maybe it will improve.
Oh, and their T Shirts are poor quality.
I got the yellow basic doggy and I thought it was fine. Not that I'm a big t-shirt person.
The typo in my shirt also made me smile. As Kagi is taking on Google and Bing and such, I keep thinking of them as a big company, but things like these remind me that it's just a bunch of people, and they make mistakes. Which I kind of like.
Even so, I too am a bit uncomfortable with their privacy philosophy, since it's so much based on trust. But somehow I find it easier to trust a bunch of people, rather than a bigger corporate entity. Even if, in reality, my rational reaction should likely be the exact opposite.
I gave it a shot. It's interesting for sure, and I appreciate just how many more languages there are to choose from than DeepL. However, as they say in the blog post they use LLMs for translation. Which would be fine on the face of it, but I think because the AI has restrictions on curses and swears the translation is Ned Flander-ized sometimes.
In some cases the translation is just as vulgar as the input, in some cases it's censored with asterisks and in others it's just mistranslated to be less vulgar, "damn" instead of "f**k" etc. It's not very consistent.
Considering that it happily spits out many standard curses, I'm guessing that it's not necessarily restrictions like swear filters or instructions in the system prompt, but rather that its training data contained censored words as well as uncensored ones.
I personally think that translation is one of the few suitable use-cases for LLMs (they do it pretty well and support an incredible amount of languages), so I'm glad to see services like this being implemented instead of shoving AI in random places.
I'd be very hesitant to use it for smaller languages that aren't supported by other services. There's a reason DeepL started with a small number of languages and then expanded, and it's extremely well-known that machine translation (whether LLM-based or not) tends to get way worse in quality with languages that don't have a lot of high-quality training data. All machine translation should be used mindfully, but it's extra worth paying attention to quality vs quantity when it comes to which languages you support.
That said, I'm happy to see more competition in the space. Machine translation can be super valuable for people and has saved my bacon plenty of times in the past.
This feels like another instance of Kagi actually wanting to be an AI company...
I recently came across this blog post and this announcement reminded me about it again.
They always have been
Source: History | Kagi's Docs
Well yeah, that is also mentioned in the blog post.
As someone that works a bit more with GDPR data requests etc, that blog post seems pretty far off the mark. I don't think the author is speaking from a place of authority. Kagi's approach to privacy of "collect and store as little information as possible" is pretty much best practice. It's possible Kagi should provide a checkout page to let you download a CSV with all user configured settings, but I'm not sure it's legally necessary.
The t-shirt thing does seem silly but generally overblown. I've seen far more egregious wastes of money than $200k on free t-shirt fulfillment for early supporters.
I mean, the author is pretty clear that they are just writing down their personal experience and are not a journalist. Having said that, it seems you have glossed over things like the "email is not pi data" bit, which in the context of GDPR very much is.
Finally, it isn't one thing specifically in that blog post that made me take notice. It is the accumulation of various things that does it for me.
I was taken in by what the author was saying here and close to cancelling my subscription — but I read his email exchange with the CEO and it didn’t come across nearly as confrontational or deluded as the author made it out to be. I thought the point by point rebuttal was straightforward and persuasive and matched up with the stances on these things I’ve seen the company take publicly in the past.
You don't think it is odd in the slightest that a CEO of a company insists on messaging a no name blog and doesn't take no for an answer?
I mean, I'm not here to shill for the CEO, but the blog post clearly gained some traction so I could see why he would reach out. I do think responding after the author specifically said "I am not interested in getting more replies from you on this subject. Declining a call does not mean I want you to argue with me about Kagi in email either" is iffy at best and disrespectful at worst, but I don't think the tone or contents of what the CEO was saying specifically were disrespectful, and I found what he was saying to be a pretty credible response to what was said in the post.
I read the email exchange (thats the author himself posted) after I read the blog post, and my conclusion was that the blog author is an asshole. The conversation pretty much went:
Author: Kagi sucks
Vlad (CEO): Hey can we talk about it?
Author: NO FUCK YOU FUCK OFF
So it's less about the specific "don't messagee anymore" and more about the rancid attitude. Also if the author didn't really want Vlad's messages, he could've blocked his email, simple as that. The mere fact he didn't points to Author being a drama queen.
I find your take oddly out of context if I am being honest.
I am not sure how you make this:
Out of what was actually said, which is:
I will give you that the last bit sentence is a bit snippy, but it is far cry from your capital case "fuck off" take. It also is not an out of the blue response, as the author of the post has interacted with the CEO before on discord and has seen how they conduct themselves. And again, by their own admission they are not a journalist, the blog can barely be called a blog and they only made it because some people kept asking about why they stopped using Kagi.
Context is important.
Context is indeed important, and I definitely do not claim the author literally said what I implied. I only exaggerated how his words felt to me at the time, that is, intentionally obtuse and provocative. In that since, context explainsy choice of words summarizing the exchange.
Again, the author could've simply blocked the CEO, or ignored him, or anything, but the fact he didn't should imply something.
Personally, I find it generally polite to respond to people, even if it is to say “no” and even if it is online. I find it even more polite to take “no” for an answer, regardless of context.
If it needs to imply anything I would say that the author has been part of the discord, has had conversations with the CEO and effectively gives one big massive hint towards the CEO saying he should do some reflection before attempting to debate it out.
Should they have replied? Maybe not, but other than that it becomes very clear the CEO doesn't take no for an answer, I am not sure if it implies much else.
I’m curious, what AI companies are you comparing to Kagi? An “AI Company” rightfully has a negative connotation, though I disagree that Kagi is an AI Company. Their product is their search engine, with AI tools offered, not forced. It may feel like I’m splitting hairs, in this case I think the words and connotation matter.
I think the risk of paying for Kagi is minimal. If Kagi starts to suck, I’ll stop using it. Kagi Search is profitable. There is no other significant revenue stream. The privacy policy
makes clear there is no logged searches, no data to profit from.
Kagi’s CEO actively participates in online discussion surrounding Kagi, search engines, and AI. His values mostly align with my own.
I see a stark contrast between Kagi and the typical AI Company. Maybe because I want to see it that way.
Also, I’ve read that blog before. It is a bit melodramatic.
I am not, it just seems to me with all the AI features they are focussing heavily on that this seems more of the thing they want to focus on than just plain old searching.
I think it is quite reasonable and far from really dramatic. But it is very clear people read these sorts of things differently.
I was never entirely convinced why I should trust this "trust me bro" assertion over any other. Other than vague hand waving along the lines of /why would they though?/.
Sometimes the so-called independent audit that they paid for is cited. But that was a long time ago now.
Privacy policies are legally binding in the US. But yeah, it really is just putting trust in the company’s word, its employees, their security, and the consumer protection institutions that enforce privacy policies. Good enough for me, personally. I understand the skepticism though.
Edit: I read this comment over again and I think it comes across a bit sarcastic but I meant it in earnest
It makes no business sense for Kagi to lie about this. They have a stable income from subscribers, which will dissappear immediately if it was leaked they were lying. I seriously doubt any data they could collect would be worth so much money for that risk, both financially and legally, to be worth it.
That aside, Google and Bing are with 100% certainty tracking you, so I would rather pick one that doesn't have a financial incentive to track or sell data.
I must say, I use the Quick Answer feature a lot. I find it super handy for a lot of things. But yeah, they are definitely leaning into AI.
edit: I've read that blog post now.
I had some interactions with Vlad and the team in the early days exactly about their privacy statements, and gdpr, and was met with much the same response, and it took the (perceived, to me) slide in ddg results quality over 18 months or so for me to decide to give them a try.
I have been happy enough with the results, which I guess is /fine/. And as I say, I find Quick Answer useful. Didn't even know about the other services offered and proposed.
I've kind of always had a spidey sense tingle about Kagi, but never could put my finger on it, but that blog post nails it.
Oh, and their T Shirts are poor quality.
As one of the mentioned, ahem, dedicated types who (very occasionally) “hangs out” in their Discord – Vlad has mentioned in the past that they have a more or less neutral stance towards AI features/LLMs specifically; but it’s (a subset of) the users requesting them. They need this stuff implemented to be unique, or at the minimum competitive, in the search business. Since the “AI” portion of features can pretty much all be ignored without effort or downside and aren’t really shoved in your faces anywhere in the core product, I don’t mind that logic, in fact, I think it’s a pretty reasonable approach.
That said, it does feel like they’re throwing a lot of experimental features at the eager user wall to see and figure out what might stick. But: I don’t think that’s a bad thing! Their business model will never scale as much as e.g. Google’s, since the overwhelming majority of people will never become paying users. So why not try to capture them with a browser/summarizer/translation/email/whatever other feature offered and try to convert and or upgrade from there? Hell, that’s my story too, and I was convinced and ready to subscribe before using up the 100 free searches you get.
Thinking about it for a bit, I think I stay subscribed because of the “clean” results. No SEO spam, surprisingly few LLM slop “articles” and help pages/fake guides, and no ads. Zero. Not a single sponsored search result, no flashy unrelated product imagery, just peace of mind when browsing.
Opposite for me (and to be clear, that’s completely fine!). Kagi actually became a Public Benefit Corporation recently. If it weren’t for that, I could see how superficially you could get the “tech bro” impression from them. But to me, this sealed the deal. Zero marketing expenses/only word-of-mouth growth and a nice product and all these additional features are cool, sure, but what tech company in this day and age actually dedicates themselves to a goal that’s not “grab as much VC cash as you can, then burn all users” (or their experience)?
Opposite experience for me. :-)
You know how the shirt has the little stitched on kagi tag on the hem, both mine just fell off, not properly stitched on. Both on the yellow, and black ones.
And yeah, I'm a sucker for a t shirt :-/
Same here, and I was very sceptical about it when they announced it. In general, I really don't want AI in my search. I use search engines to look for information, not answers. But Kagi's implementation kind of works for me.
I notice that this has actually been quite typical for me with whatever Kagi announces. My first reaction often is negative, but when I look into the new feature a little, I can usually see why they are doing it. I still don't really use any features other than the core search, but I don't mind that they are experimenting, if it makes sense from their point of view.
It's actually a bit different with this translator thing, which I am very interested it. Unfortunately, at least based on my first tests, the results and the user experience aren't quite there yet. Maybe it will improve.
I got the yellow basic doggy and I thought it was fine. Not that I'm a big t-shirt person.
The typo in my shirt also made me smile. As Kagi is taking on Google and Bing and such, I keep thinking of them as a big company, but things like these remind me that it's just a bunch of people, and they make mistakes. Which I kind of like.
Even so, I too am a bit uncomfortable with their privacy philosophy, since it's so much based on trust. But somehow I find it easier to trust a bunch of people, rather than a bigger corporate entity. Even if, in reality, my rational reaction should likely be the exact opposite.
My t-shirt looks so unflattering on me. Unisex cut, and I forgot that American small is massive :(((
Hey, I guess it'll contribute to the untidy aesthetic! But it seems you're not so tall after all :)