Good. I'm a paying user from day one, and I never used, nor want to see any AI features on my search engine. I'd use Google if I cared about any of that stuff.
Honestly, I just want my money to be used to improve good old match-keywords-against-index search.
Unless you append a question mark to your queries you're not going to interact with AI. Just don't use it. They're not forcing it down your throat. Cheering for people losing access to the service while paying the same price just seems wrong.
And on the other hand, I have the "end with question mark? That means you want AI" thing disabled. I am subsidizing the tokens other paying users are using far more than I am, which is approximately zero. Why should I pay for features I don't use?
> A common misconception is that AI is more expensive than search. Opposite is true - performing a search is 100x-1000x more expensive than doing an interaction with AI.
> A single search is about ~ 1.5 cents (probably bit less these days, but general ballpark)
> A small AI model can generate a lot of tokens for the same price
> To put this in perspective, 1.5 cents of Gemini Flash usage:
> 50,000 input tokens ≈ 37,500 words (roughly 75-150 pages of text)
or
6,000 output tokens ≈ 4,500 words (roughly 9-18 pages of text)
> That is A LOT of information for the cost of just 1 search.
> It may be counter-intuitive but that is how it is. If someone is using AI to answer common questions, learn about a concept, do a quick brainstorm or a translation - instead of searching - our cost is (much) lower.
> So if anything, having AI lowers our cost, not increases it. This is why when we added access to AI models to Pro tier we didn't increase the price.
Doesn't make sense because AI responses are not grounded, still for AI to make sense in this context, and have any relation with Kagi purpose, you need to have the search still, and then the AI process the search results.
Wasn’t it also a free upgrade when it launched? They didn’t bump the price of their search plans when they added Assistant.
I wish they’d separated them from the start because I knew immediately that my subscription (where I don’t use Assistant) was going to subsidise subscriptions that do. Now I don’t know what the right thing for them to do is, given they’ve been marketing Assistant as a feature.
Same here. I'd definitely drop the subscription immediately, unless the assistant was super cheap, like $1 or $2 more which is the most I'd pay for kagi personally.
Is it any better than Google's or the one in DDG? Those things are wrong most of the time for questions that couldn't have been answered by looking at a few words in the summary of the very first search result anyway. I find they only save me time if what I was searching was so trivial that it doesn't matter at all if it's wrong ("How old is such-and-such actor?") and I didn't really even need to search it in the first place. Otherwise, if it actually matters, I still have to check the sources, because no matter how confident the answer looks, it's wrong way too often to rely on it.
(Gather 'round, kids: used to be, questions of utter trivia like the exact ages of actors or whatever didn't even take root in our minds in the first place, we didn't even feel the itch to answer them to begin with, because it was far more effort to answer them than it was worth, unless it was a topic you cared enough about to have a bunch of books about it on-hand, so such questions would usually fall out of our brains before we even consciously engaged with them. Our heads were a lot less noisy then. It was very nice.)
I've never user Google or DDG assistants, tbh. My use for Kagi's assistant is for stuff related to work. I'm a university prof., so it "saves me a bit of time" while checking for errors in assignments, clarify concepts to explain to students and so on. I almost never use it to search for stuff.
You still have to live in a world where LLMs exist and are based on the stolen labor of millions of people and are actively destroying the livelihood of people, our environment and democracy.
I happened to be checking out kagi just yesterday (as a first time costumer) and I can say the AI focus put me off it.
I was running away from the constant shoving of AI features by google, so most of their landing page advertising AI related stuff wasn't really selling it for me.
Totally anecdotal evidence, of course, but I'm sure I'm not the only one dropping off the funnel for that reason.
My comment sounds negative, but for what it's worth, I still am a very happy user, and you can disable assistant/quick answer so it's never pushed on you if you don't care about it.
I just don't want Kagi to get enshittified. If everybody and their dog are focusing on AI, good-old search becomes a vanishing product space which definitely still has its place in the AI era, when you want determinism from your search.
I’m also fine with the change. Claude is my go-to for research as the standalone app and shortcut are very convenient.
The Kagi team should focus on the core product; that’s what I’m paying for. I stopped being a ProtonMail customer once they began chasing side projects instead of polishing their actual offering. I hope Kagi doesn't go down that same route.
That's fine for me - as long as the new subscription (Pro+AI) is the same price as it is now.
If they increase the prices, well then my Kagi subscription is gone and I will move on. I'm a happy user for noe but I think search wuality has gotten worse lately and I'm more often using the AI instead of search because search just does not bring any good results anymore.
Exactly. Same as you I am just paying for search. I never used the assistant, and never will. Right now Kagi is good enough at search that it would be annoying to lose. But if I was forced to go back to Google I could survive by using adblock. I really wish Kagi would just put all their engineering efforts on search to make it so good that I couldn’t possibly live without it.
I don’t need a new browser. I don’t need a replacement for Google Maps, since Google Maps is actually good and Kagi will never even catch up to Apple Maps. I don’t need any AI trash.
Just have everybody work on the search engine to make it is faster, more reliable, and free of content farms or slop. That is the only reason I’m paying for Kagi.
They could use AI is to remove web pages with AI generated spam content in search results. I don't know how accurate that could work. For technical blogs there should be a lens that ranks GPG signed articles higher, is that even possible?
He's say splitting the pot for AI features reduces the value of improved search quality.
I'd rather that too. Kagi should try to curate like a Librarian and not some kind of oracle. They're not going to capture people with offering AI assistants. They will capture people by filtering out all the garbage places like Google throw into their results to get ad dollars.
I think is because, whether intentional or not, Kagi wound up being aimed at (or sought by, either or) people who don't see these things are related.
I don't know if they're arrogant to consider themselves the main customer base, but I don't think it's an unfair assumption, either.
Kagi was almost advertised to me as what Google used to be. Again, I don't know if that's ever how Kagi put it themselves, but that's the impression that was given.
I agree with you for a general assistant but even if I'm also not interested to pay for an assistant there are 2 features that I like and bring a lot of value by default in my opinion:
- if you put an interrogation point at the end of the query you have an AI reply based on the search query.
- you can ask a question about that to investigate more.
Strangely: "Searches are, unexpectedly, more expensive for us to serve than AI. A user that does only traditional searches (not using Assistant) will cost more on our end than someone using AI + search."
> A user that does only traditional searches (not using Assistant) will cost more on our end than someone using AI + search.
Unless I am missing something, that appears to be a mathematically impossible claim. They are saying A > A+B, where both A and B are positive values. I suppose it could be that non-AI users do more total searches, and the extras add up. Hmm.
How can this be true? Should I not worry about margin compression at Google due to heavy cap-ex requirements anymore? Is this temporary because Kagi got a good deal for tokens? Are they paying through their noses for the Google search API calls? I don't get it.
Because they pay API costs to send the search to SerpApi. I forget exactly what the cost was for them per-search and I'm having little luck finding it, but I know they've published that cost before and I know it's more than a whole cent. By comparison, running a good but not top-tier model to answer the same question might run a small fraction of a cent. Cheaper than a follow up query by the user.
I use Kagi search all the time. I often use the Quick Answer feature (append "?" at the end of the query, gives an AI overview/answer, then standard search results below). I never use the full Assistant - I have ChatGPT and Claude for chatbot use cases.
I like Quick Answer because of its ligtweight UX. Ctrl-T for a new tab in any browser window, type the question, get the answer. This is faster and mentally lighter than switching to a chatbot, typing the question there and answering. If I'm to use the chatbot, I don't see a special need to use Kagi's.
This is generally how I use it too. Every now and then I click through to the assistant for follow-up questions. I appreciate that it’s there, even if I use it infrequently. The default kimi model gives surprisingly good answers too
> Kagi Assistant's web tool uses Kagi Search, and that has nothing to do with this subscription plans discussion, we're not changing anything there. The same applies to LLM-powered features in Kagi Search, like Quick Answer.
I'm a happy subscriber, and it's certainly a big improvement over Google search. But the internet just isn't the same place it was five years ago. And as search results (for non-navigational queries) are becoming less useful by the day, I find myself asking AI to do it for me more.
There's a lot to like about Kagi, but they'll probably have to reinvent themselves if they want to grow beyond the niche that high level internet search will probably become.
Why though. Why does it need to grow beyond a niche premium option? As long as they’re paying the bills and everyone is happy why not just let a good thing keep chugging along.
Agreed. I LOVE Kagi as a search engine - so long as it answers queries in under 2s with no ads, I'm a very happy customer. I don't mind if they flirt with LLMs, but if the LLM work detracts from the search work they will lose me as a paying customer. If the LLM work slows down search results I also lose the only thing I pay for - search result response time and correctness.
I've been a paying customer since 04/2022, and have the early adopter badge. I was easily doing 600-800 searches/month, and now I do 400-300 searches. I think that's the reality. More and more people are asking ChatGPT or whatever for search.
I didn't renew my Kagi subscription, as I am now mostly using AI based search (google, chat.bing.com, perplexity). Search engine wise, Kagi was superior but it is just that traditional search engines are less and less needed with the rise of AI.
BUT Kagi is in a good spot, as they have their user data (and the feedback/upvote/downvote/blacklist feature) to train their own models on. Maybe their AI will one day be a superior search. Especially when the big ones like Google will start to enshittify the free AI tier with ads, or SEO-like AI manipulation on Google will take off.
I’m a happy user of both Kagi Search and Assistant. I would totally support a “even more than Pro” subscription for Assistant, but then it should also have its own proper API if I pay for it like a proper product.
That said, I wonder how things would work if you subscribe to the assistant but not to search? Because to me, the deep search integration is precisely what makes this specific AI assistant better than, for example, ChatGPT.
"As we prepare to launch Kagi Assistant as a standalone subscription, we're considering changes to how it will be available on the Pro and Starter plans (Search subscriptions).
Currently these plans let you use the Kagi Assistant each month up to the AI cost value included in your subscription, similar to how it works on Ultimate. The change we're consideringn would turn this into more of a trial where you'd still have access to the Assistant, but only up to a fixed usage cap. Once you hit that cap, you'd need to subscribe to either the Ultimate plan or the Kagi Assistant's own subscription to keep using it (same mechanism we already have for Search)."
Ah, but only a truly great writer could have come up with:
> Start reading books or you're going to look stupid to the people around you
Wherein the prose wasn’t at all sloppy, the tautology was certainly intentional; the implied audience of “look stupid” could be to people entirely absent of the vicinity!
Make grand assumptions about other HN users without knowing anything else but their HN comments, I betcha you won't find a better past-time for internet strangers :)
If they remove a feature (even if I am not using it) while keeping the price the same - why should I stay a customer? I have a contract. And at least in Germany one party to a contract can not just alter the terms of the contract.
They need to cancel the existing contract and offer a new one. And I would not need to accept the new one. Yes, I would not be using Kagi anymore. But why should I have a business relationship with a company that does not honor the contractual obligations it entered into?
Well, no... I use maybe 1/5 to 1/3 of may month allowance on Assistant. It's nice to have. And I subscribed to Kagi also because of it. So, the Assistant going away would mean my expensive Duo subscription going away as well.
It is not worse I would say. It uses neutral system prompt by default, whereas Gemini and ChatGPT will please you too much to mislead you badly. Also the base search is much batter. You can control the search while with Gemini, for example, you can't.
Keep in mind that kagi offers a wide range of models, not just one. I wouldn't want to have multiple subscriptions (for chatgpt, anthropic, gemini etc.)
This is also true if you use something like OpenRouter, and it will almost always cheaper or betterter (excluding Kagi Search).
I love Kagi-I can't imagine going back to any other search engine-but it isn’t competitive when it comes to LLMs. In its defense, that’s largely because others are bleeding money.
In my experience that is true, but I get the answers I need much faster than any kagi search with or without their ai integration from Gemini, ChatGPT or Claude. I was rooting for them, but they just seemed to far behind the leading llms for search. Same with perplexity. Could never figure out why it needs to exist. If I want citations, I’ll just ask one of the LLMs I mentioned to provide them.
The privacy buffer between the LLM providers and the user is a large part of the appeal. Having it hooked up to a search engine outside of the data broker space makes it uniquely attractive.
Given that window dressing, having toasted bread in a kitchen that isn't selling my data is something I want.
Because at least for my own usage of Google the LLM started out as an interactive search with substantially better context filtering that could tune the results to my desired technical level. However I promptly started just having it explain the subject matter to me rather than spending 30+ minutes consulting various docs and forum posts because it makes for an excellent secretary/tutor combo provided you vigilantly watch for misinformation.
So to answer your question, while charts might not be particularly useful for a search engine a tutor certainly benefits from them.
I see the slowness and now the news that they might remove the assistant for the $10 plans as evidence of how costly it is to run LLMs and by extension how unsustainable it must be for OpenAi, anthropic, Microsoft etc to be offering such performance for free or very low prices. Surely something has to give soon.
I didn't consider that bundling Search & Assistant maybe puts them in a tricky spot among some users who revile LLM features, and others who will utilize them to the cap. To the degree that the former is subsidizing the latter, or costing them customers (probably not a ton): I can see why separating the two offerings makes sense.
Though I'm sympathetic to the users for whom this would basically be a strict downgrade in featureset.
The big advantage of this is that instead of getting commentariat feedback, you will see people’s wallet votes on whether they value kagi for being the best of like 6 options for search, or for being a distinctly middle of the pack option in a field of millions of thin llm wrappers.
I’ve used it quite a bit, but just as a natural language frontend to their search engine. I don’t see it as a true alternative to generic chat bots like Claude or ChatGPT, and certainly not a coding tool. I would likely not subscribe to Assistant.
You can look at it other way around - why to have ChatGPT when you can use chatgpt through kagi. One of the advantages being that they claim to be "privacy" layer that doesn't give other companies metadata about you.
For a beautiful 6 months before the frontier labs were able to search as part of their responses, Kagi had an edge here. This was when the “AIs hallucinate!” meme was thriving and my response was “Wait, your AI AGENT doesn’t provide references?” and I felt very cutting edge then.
Now that frontier labs have search, generalised tool use, memory, I no longer use The Agent. Haven’t done so for months. I still find great value in Kagi’s FastGPT even though this overlaps with many search engines today. It works well and I don’t pay with my privacy. But there’s no Agent edge any more, Claude wins.
Ultimate user here. I assume this doesn't kill "quick answers" for pro users - which I use frequently when I need a quick summarization. For assistant use directly, I've been thinking about stepping back from ultimate, as I use claude for AI rubber ducking; which works better than all of the LLMs available on assistant.
> Kagi Assistant's web tool uses Kagi Search, and that has nothing to do with this subscription plans discussion, we're not changing anything there. The same applies to LLM-powered features in Kagi Search, like Quick Answer.
I have maxed out my ultimate usage before and when that happens the quick summarization tools did not function indicating I hit my limit. I assume it would affect those, but that might be part of how they break it up.
To be honest, while it's great that it uses Kagi as its search engine and that improves the results, I find myself more and more using Claude and Gemini over Assistant while still using Kagi as my main manual search engine. They should probably focus more on their main product instead of LLMs.
I like the occasional feature of appending a question mark to your query to get a nice summarizer to comb through the internet so I don't, but I only use it a few times a month.
Why is this on a whole separate domain and why can't I use my Kagi account to comment? Why do I have to sign up here? Why haven't I seen any messaging from Kagi directly about this? This seems really weird.
Those domains have been around since forever (also https://orionfeedback.org). I guess it was easier to just set up a new platform instead of integrating it, or building something from scratch.
I haven't checked in a while, but I'm sure there's been conversations about this on discord as well (https://help.kagi.com/kagi/support-and-community/discord-ser...). I'm too busy to read up on every little thing, so I'm glad this happens elsewhere, off-site, and I just get the big changes through my RSS feed.
I know I'll sound like a Google shill, but I really don't know how this could compete against what Google AI Studio offers for now. Studio even has free quota (while being completely opaque).
I once feel Kagi is some fresh air that we need for search engine. But now it seems to be more and more directly competing against the other LLM web summery products and imo Google/Gemini is light-seconds ahead in that space. The AI Overview is an ultra water downed version of it and it's still somewhat usable.
Perhaps this move is signaling that they'll be back focusing on search itself more though (I hope so).
Please keep Assistant, it's an easy to use alternative to OpenAI, and is not intrusive. I will say that I find the GUI infuriating though...
It's never clear where the cursor is when you start a new chat. Don't understand the low contrast layout with too much whitespace.
I’m a pro subscriber. I like Kagi, but I think it is getting lost. Instead of focusing on being the best search engine in a world where other search engines have become deeply enshitified, they are joining the enshitification train.
Start by removing features and adding a different subscription. Next add one ad from curated sources. Then add ads everywhere.
I will probably cancel my subscription soon and spare myself the disappointment.
I've been a Kagi Pro user for several years now, and to be honest, the models from the pro tier aren't that useful compared to the free versions of Gemini & ChatGPT.
For me, I pay for Kagi pro for search without Google/Bing enshittification, their Translate (which I use quite often while I'm learning German/working in German - better for me than Google Translate), and their Summarizer. I pay for Claude, and also occasionally use OpenRouter for my AI needs.
I honestly think that AI goes against Kagis's main purpose. I see it as a tool to search the Internet. Carefully craft what you think is relevant, try to remove as much AI generated crap from your results, etc.
By using AI you are doing the opposite. You are letting some random AI get the results for you.
That's assuming Kagi assistant is the AI that searches the Internet. I don't know, I have never used it. I use Kagi search every day though
I use the assistant daily, often instead of search. The search is good, but the sad fact is searching www has been getting less interesting fast and there simply isn't much of value to find (maybe if they added an option to search in the Wayback Machine? I would pay extra for that!). Adding a LLM on top of search results and to fire off multiple searches makes it possible to sometimes squeeze a little bit of actually useful information out of the cesspool that is modern web.
Without the Assistant I would probably go back down to the lowest kagi tier.
While the assistant is a decent add on, it fails as a standalone product. It just isn't in the same league as Claude, ChatGPT, Gemini, or even LeChat, and I’d hate to see them waste too much development time on it.
I recommend Kagi focus on being a sidecar to AIs rather than try to offer it in any way; I'm hoping the API comes out and I can get proper search results for my local LLM.
Unless they're building their own LLMs, best not to annoy people who already see random LLMs everywhere.
Good. I'm a paying user from day one, and I never used, nor want to see any AI features on my search engine. I'd use Google if I cared about any of that stuff.
Honestly, I just want my money to be used to improve good old match-keywords-against-index search.
Unless you append a question mark to your queries you're not going to interact with AI. Just don't use it. They're not forcing it down your throat. Cheering for people losing access to the service while paying the same price just seems wrong.
And on the other hand, I have the "end with question mark? That means you want AI" thing disabled. I am subsidizing the tokens other paying users are using far more than I am, which is approximately zero. Why should I pay for features I don't use?
According to the Kagi team, customers who don’t use AI cost them more, so you’re not subsidizing anything.
I am interested in this claim. Do you have a link?
https://kagifeedback.org/d/1338-provide-a-plan-without-ai-fu...
> A common misconception is that AI is more expensive than search. Opposite is true - performing a search is 100x-1000x more expensive than doing an interaction with AI.
> A single search is about ~ 1.5 cents (probably bit less these days, but general ballpark)
> A small AI model can generate a lot of tokens for the same price
> To put this in perspective, 1.5 cents of Gemini Flash usage:
> 50,000 input tokens ≈ 37,500 words (roughly 75-150 pages of text) or 6,000 output tokens ≈ 4,500 words (roughly 9-18 pages of text)
> That is A LOT of information for the cost of just 1 search.
> It may be counter-intuitive but that is how it is. If someone is using AI to answer common questions, learn about a concept, do a quick brainstorm or a translation - instead of searching - our cost is (much) lower.
> So if anything, having AI lowers our cost, not increases it. This is why when we added access to AI models to Pro tier we didn't increase the price.
Doesn't make sense because AI responses are not grounded, still for AI to make sense in this context, and have any relation with Kagi purpose, you need to have the search still, and then the AI process the search results.
Common facts like “what is the capital of Hungary” are repeated so many times in the training data that the LLM knows them without a search.
Makes sense.
Thank you.
Well, good for you, I guess. But for the rest of us who use the Assistant this would be a net downgrade, AKA paying the same for less.
Wasn’t it also a free upgrade when it launched? They didn’t bump the price of their search plans when they added Assistant.
I wish they’d separated them from the start because I knew immediately that my subscription (where I don’t use Assistant) was going to subsidise subscriptions that do. Now I don’t know what the right thing for them to do is, given they’ve been marketing Assistant as a feature.
Same here. I'd definitely drop the subscription immediately, unless the assistant was super cheap, like $1 or $2 more which is the most I'd pay for kagi personally.
Is it any better than Google's or the one in DDG? Those things are wrong most of the time for questions that couldn't have been answered by looking at a few words in the summary of the very first search result anyway. I find they only save me time if what I was searching was so trivial that it doesn't matter at all if it's wrong ("How old is such-and-such actor?") and I didn't really even need to search it in the first place. Otherwise, if it actually matters, I still have to check the sources, because no matter how confident the answer looks, it's wrong way too often to rely on it.
(Gather 'round, kids: used to be, questions of utter trivia like the exact ages of actors or whatever didn't even take root in our minds in the first place, we didn't even feel the itch to answer them to begin with, because it was far more effort to answer them than it was worth, unless it was a topic you cared enough about to have a bunch of books about it on-hand, so such questions would usually fall out of our brains before we even consciously engaged with them. Our heads were a lot less noisy then. It was very nice.)
I've never user Google or DDG assistants, tbh. My use for Kagi's assistant is for stuff related to work. I'm a university prof., so it "saves me a bit of time" while checking for errors in assignments, clarify concepts to explain to students and so on. I almost never use it to search for stuff.
For the rest of us that deeply distrust and despise the anti-ethical and wasteful assistants is an upgrade.
Friendly reminder that you're not forced to use it. "I despise soccer, so soccer should disappear".
You still have to live in a world where LLMs exist and are based on the stolen labor of millions of people and are actively destroying the livelihood of people, our environment and democracy.
Again, right now the assistant is entirely optional. Oh, summarizer and translate are base on the same models, by the way.
Soccer is not destroying the planet and making basic computer hardware unaffordable.
You're looking at the finger and missing the Moon.
I think a more fair analogy is "I despise soccer, so the specific cable package I choose should not include it".
I am interested in your manifesto
You haven't thought too hard about it then; Kagi is (in part) looking to get better margins on its AI spend here.
I happened to be checking out kagi just yesterday (as a first time costumer) and I can say the AI focus put me off it.
I was running away from the constant shoving of AI features by google, so most of their landing page advertising AI related stuff wasn't really selling it for me.
Totally anecdotal evidence, of course, but I'm sure I'm not the only one dropping off the funnel for that reason.
My comment sounds negative, but for what it's worth, I still am a very happy user, and you can disable assistant/quick answer so it's never pushed on you if you don't care about it.
I just don't want Kagi to get enshittified. If everybody and their dog are focusing on AI, good-old search becomes a vanishing product space which definitely still has its place in the AI era, when you want determinism from your search.
So how much Kagi without AI costs? It should me much cheaper subscription, right?
I’m also fine with the change. Claude is my go-to for research as the standalone app and shortcut are very convenient.
The Kagi team should focus on the core product; that’s what I’m paying for. I stopped being a ProtonMail customer once they began chasing side projects instead of polishing their actual offering. I hope Kagi doesn't go down that same route.
That's fine for me - as long as the new subscription (Pro+AI) is the same price as it is now.
If they increase the prices, well then my Kagi subscription is gone and I will move on. I'm a happy user for noe but I think search wuality has gotten worse lately and I'm more often using the AI instead of search because search just does not bring any good results anymore.
Exactly. Same as you I am just paying for search. I never used the assistant, and never will. Right now Kagi is good enough at search that it would be annoying to lose. But if I was forced to go back to Google I could survive by using adblock. I really wish Kagi would just put all their engineering efforts on search to make it so good that I couldn’t possibly live without it.
I don’t need a new browser. I don’t need a replacement for Google Maps, since Google Maps is actually good and Kagi will never even catch up to Apple Maps. I don’t need any AI trash.
Just have everybody work on the search engine to make it is faster, more reliable, and free of content farms or slop. That is the only reason I’m paying for Kagi.
They could use AI is to remove web pages with AI generated spam content in search results. I don't know how accurate that could work. For technical blogs there should be a lens that ranks GPG signed articles higher, is that even possible?
They're working on it: https://help.kagi.com/kagi/features/slopstop.html
Discussed at: https://news.ycombinator.com/item?id=46716806
Can't AI sign their articles with GPG too though?
Wouldn't a a feature to turn all AI features work? You can turn of all AI from Brave search and duckduckgo and for Google just use startpage.
He's say splitting the pot for AI features reduces the value of improved search quality.
I'd rather that too. Kagi should try to curate like a Librarian and not some kind of oracle. They're not going to capture people with offering AI assistants. They will capture people by filtering out all the garbage places like Google throw into their results to get ad dollars.
Why are people pretending that search and AI are unrelated things that distract from each other? They’re obviously compliments
I think is because, whether intentional or not, Kagi wound up being aimed at (or sought by, either or) people who don't see these things are related.
I don't know if they're arrogant to consider themselves the main customer base, but I don't think it's an unfair assumption, either.
Kagi was almost advertised to me as what Google used to be. Again, I don't know if that's ever how Kagi put it themselves, but that's the impression that was given.
I agree with you for a general assistant but even if I'm also not interested to pay for an assistant there are 2 features that I like and bring a lot of value by default in my opinion: - if you put an interrogation point at the end of the query you have an AI reply based on the search query. - you can ask a question about that to investigate more.
Strangely: "Searches are, unexpectedly, more expensive for us to serve than AI. A user that does only traditional searches (not using Assistant) will cost more on our end than someone using AI + search."
https://kagifeedback.org/d/1338-provide-a-plan-without-ai-fu...
> A user that does only traditional searches (not using Assistant) will cost more on our end than someone using AI + search.
Unless I am missing something, that appears to be a mathematically impossible claim. They are saying A > A+B, where both A and B are positive values. I suppose it could be that non-AI users do more total searches, and the extras add up. Hmm.
I think their point is that an AI user makes fewer searches on average because AI helps ttem find what they’re looking for faster.
They're likely saying that at equal usage, the user with mixed usage will cost less because the cost of B is lower than A.
How can this be true? Should I not worry about margin compression at Google due to heavy cap-ex requirements anymore? Is this temporary because Kagi got a good deal for tokens? Are they paying through their noses for the Google search API calls? I don't get it.
Because they pay API costs to send the search to SerpApi. I forget exactly what the cost was for them per-search and I'm having little luck finding it, but I know they've published that cost before and I know it's more than a whole cent. By comparison, running a good but not top-tier model to answer the same question might run a small fraction of a cent. Cheaper than a follow up query by the user.
Yeah, that reply is a mildly infuriating, in view of what they're planning.
I use Kagi search all the time. I often use the Quick Answer feature (append "?" at the end of the query, gives an AI overview/answer, then standard search results below). I never use the full Assistant - I have ChatGPT and Claude for chatbot use cases.
I like Quick Answer because of its ligtweight UX. Ctrl-T for a new tab in any browser window, type the question, get the answer. This is faster and mentally lighter than switching to a chatbot, typing the question there and answering. If I'm to use the chatbot, I don't see a special need to use Kagi's.
(my €0.02 as a paid user)
This is generally how I use it too. Every now and then I click through to the assistant for follow-up questions. I appreciate that it’s there, even if I use it infrequently. The default kimi model gives surprisingly good answers too
Luckily they're not changing that part.
> Kagi Assistant's web tool uses Kagi Search, and that has nothing to do with this subscription plans discussion, we're not changing anything there. The same applies to LLM-powered features in Kagi Search, like Quick Answer.
You can add some "bang search" (like your ? for quick answer) to use the other models Kagi supports like Kimi.
The quick model is there to be mainly quick.
Kagi is unfortunately in a tough spot, imo.
I'm a happy subscriber, and it's certainly a big improvement over Google search. But the internet just isn't the same place it was five years ago. And as search results (for non-navigational queries) are becoming less useful by the day, I find myself asking AI to do it for me more.
There's a lot to like about Kagi, but they'll probably have to reinvent themselves if they want to grow beyond the niche that high level internet search will probably become.
Why though. Why does it need to grow beyond a niche premium option? As long as they’re paying the bills and everyone is happy why not just let a good thing keep chugging along.
Amen. "Growth" is literally product cancer
Investors want to see growth, chugging along is not the return they expect. It would be nice tho.
Sounds like a poisoned chalice.
As far as I recall, the founder and angels have put in almost as much money as VCs. Certainly an interesting funding story.
Agreed. I LOVE Kagi as a search engine - so long as it answers queries in under 2s with no ads, I'm a very happy customer. I don't mind if they flirt with LLMs, but if the LLM work detracts from the search work they will lose me as a paying customer. If the LLM work slows down search results I also lose the only thing I pay for - search result response time and correctness.
I always thought the browser, coworking space (https://hub.kagi.com/), and mail (https://kagimail.com/) are a distraction.
I've been a paying customer since 04/2022, and have the early adopter badge. I was easily doing 600-800 searches/month, and now I do 400-300 searches. I think that's the reality. More and more people are asking ChatGPT or whatever for search.
"Asking AI" is doing a lot of work there.
The people who pay for Kagi do so for very specific reasons, often because they know what "asking AI" really means for their privacy.
I didn't renew my Kagi subscription, as I am now mostly using AI based search (google, chat.bing.com, perplexity). Search engine wise, Kagi was superior but it is just that traditional search engines are less and less needed with the rise of AI.
BUT Kagi is in a good spot, as they have their user data (and the feedback/upvote/downvote/blacklist feature) to train their own models on. Maybe their AI will one day be a superior search. Especially when the big ones like Google will start to enshittify the free AI tier with ads, or SEO-like AI manipulation on Google will take off.
I’m a happy user of both Kagi Search and Assistant. I would totally support a “even more than Pro” subscription for Assistant, but then it should also have its own proper API if I pay for it like a proper product.
That said, I wonder how things would work if you subscribe to the assistant but not to search? Because to me, the deep search integration is precisely what makes this specific AI assistant better than, for example, ChatGPT.
From the thread on kagifeedback.org:
"As we prepare to launch Kagi Assistant as a standalone subscription, we're considering changes to how it will be available on the Pro and Starter plans (Search subscriptions).
Currently these plans let you use the Kagi Assistant each month up to the AI cost value included in your subscription, similar to how it works on Ultimate. The change we're consideringn would turn this into more of a trial where you'd still have access to the Assistant, but only up to a fixed usage cap. Once you hit that cap, you'd need to subscribe to either the Ultimate plan or the Kagi Assistant's own subscription to keep using it (same mechanism we already have for Search)."
[flagged]
Guess that serves as a signal that this move from Kagi isn't so bad, if typos is the biggest and greatetst criticism you can come up with as a reply.
[flagged]
Ah, but only a truly great writer could have come up with:
> Start reading books or you're going to look stupid to the people around you
Wherein the prose wasn’t at all sloppy, the tautology was certainly intentional; the implied audience of “look stupid” could be to people entirely absent of the vicinity!
Make grand assumptions about other HN users without knowing anything else but their HN comments, I betcha you won't find a better past-time for internet strangers :)
Even after you pointed out specific mistakes, I fail to understand how the writing is indicative of AI over-reliance?
If they remove a feature (even if I am not using it) while keeping the price the same - why should I stay a customer? I have a contract. And at least in Germany one party to a contract can not just alter the terms of the contract.
They need to cancel the existing contract and offer a new one. And I would not need to accept the new one. Yes, I would not be using Kagi anymore. But why should I have a business relationship with a company that does not honor the contractual obligations it entered into?
Presumably they would discontinue it at the end of a billing cycle? Or have you entered into a long term contract with them?
To me this seems fine. You'd run out of tokens pretty quick without ultimate anyway, so it is a small cross section of users this affects.
I am interested to learn if anything else is coming besides a billing change. Like will ultimate/assistant subscribers get access to MCP?
Well, no... I use maybe 1/5 to 1/3 of may month allowance on Assistant. It's nice to have. And I subscribed to Kagi also because of it. So, the Assistant going away would mean my expensive Duo subscription going away as well.
I admire and applaud Kagi for what they are trying to be.
But on the AI front, the Assistant is simply worse than using for example Gemini or ChatGPT directly. It is slower, it cannot generate images etc.
That's why I don't think anybody is going to pay for the Assistant alone.
The research assistant should be able to generate images
It is not worse I would say. It uses neutral system prompt by default, whereas Gemini and ChatGPT will please you too much to mislead you badly. Also the base search is much batter. You can control the search while with Gemini, for example, you can't.
Keep in mind that kagi offers a wide range of models, not just one. I wouldn't want to have multiple subscriptions (for chatgpt, anthropic, gemini etc.)
This is also true if you use something like OpenRouter, and it will almost always cheaper or betterter (excluding Kagi Search).
I love Kagi-I can't imagine going back to any other search engine-but it isn’t competitive when it comes to LLMs. In its defense, that’s largely because others are bleeding money.
In my experience that is true, but I get the answers I need much faster than any kagi search with or without their ai integration from Gemini, ChatGPT or Claude. I was rooting for them, but they just seemed to far behind the leading llms for search. Same with perplexity. Could never figure out why it needs to exist. If I want citations, I’ll just ask one of the LLMs I mentioned to provide them.
Also, there news app is pretty underwhelming.
On the other hand, it doesn't monetize any of the data you give. Six of one, half dozen of the other, yeah?
Re: image gen... it's a search engine. Why would I need my refrigerator to toast my bread?
The privacy buffer between the LLM providers and the user is a large part of the appeal. Having it hooked up to a search engine outside of the data broker space makes it uniquely attractive.
Given that window dressing, having toasted bread in a kitchen that isn't selling my data is something I want.
Because at least for my own usage of Google the LLM started out as an interactive search with substantially better context filtering that could tune the results to my desired technical level. However I promptly started just having it explain the subject matter to me rather than spending 30+ minutes consulting various docs and forum posts because it makes for an excellent secretary/tutor combo provided you vigilantly watch for misinformation.
So to answer your question, while charts might not be particularly useful for a search engine a tutor certainly benefits from them.
I see the slowness and now the news that they might remove the assistant for the $10 plans as evidence of how costly it is to run LLMs and by extension how unsustainable it must be for OpenAi, anthropic, Microsoft etc to be offering such performance for free or very low prices. Surely something has to give soon.
I didn't consider that bundling Search & Assistant maybe puts them in a tricky spot among some users who revile LLM features, and others who will utilize them to the cap. To the degree that the former is subsidizing the latter, or costing them customers (probably not a ton): I can see why separating the two offerings makes sense.
Though I'm sympathetic to the users for whom this would basically be a strict downgrade in featureset.
The big advantage of this is that instead of getting commentariat feedback, you will see people’s wallet votes on whether they value kagi for being the best of like 6 options for search, or for being a distinctly middle of the pack option in a field of millions of thin llm wrappers.
I’ve used it quite a bit, but just as a natural language frontend to their search engine. I don’t see it as a true alternative to generic chat bots like Claude or ChatGPT, and certainly not a coding tool. I would likely not subscribe to Assistant.
You can look at it other way around - why to have ChatGPT when you can use chatgpt through kagi. One of the advantages being that they claim to be "privacy" layer that doesn't give other companies metadata about you.
For a beautiful 6 months before the frontier labs were able to search as part of their responses, Kagi had an edge here. This was when the “AIs hallucinate!” meme was thriving and my response was “Wait, your AI AGENT doesn’t provide references?” and I felt very cutting edge then.
Now that frontier labs have search, generalised tool use, memory, I no longer use The Agent. Haven’t done so for months. I still find great value in Kagi’s FastGPT even though this overlaps with many search engines today. It works well and I don’t pay with my privacy. But there’s no Agent edge any more, Claude wins.
Ultimate user here. I assume this doesn't kill "quick answers" for pro users - which I use frequently when I need a quick summarization. For assistant use directly, I've been thinking about stepping back from ultimate, as I use claude for AI rubber ducking; which works better than all of the LLMs available on assistant.
That's right:
> Kagi Assistant's web tool uses Kagi Search, and that has nothing to do with this subscription plans discussion, we're not changing anything there. The same applies to LLM-powered features in Kagi Search, like Quick Answer.
I have maxed out my ultimate usage before and when that happens the quick summarization tools did not function indicating I hit my limit. I assume it would affect those, but that might be part of how they break it up.
To be honest, while it's great that it uses Kagi as its search engine and that improves the results, I find myself more and more using Claude and Gemini over Assistant while still using Kagi as my main manual search engine. They should probably focus more on their main product instead of LLMs.
I completely forgot this existed.
I like the occasional feature of appending a question mark to your query to get a nice summarizer to comb through the internet so I don't, but I only use it a few times a month.
Why is this on a whole separate domain and why can't I use my Kagi account to comment? Why do I have to sign up here? Why haven't I seen any messaging from Kagi directly about this? This seems really weird.
Those domains have been around since forever (also https://orionfeedback.org). I guess it was easier to just set up a new platform instead of integrating it, or building something from scratch.
I haven't checked in a while, but I'm sure there's been conversations about this on discord as well (https://help.kagi.com/kagi/support-and-community/discord-ser...). I'm too busy to read up on every little thing, so I'm glad this happens elsewhere, off-site, and I just get the big changes through my RSS feed.
I know I'll sound like a Google shill, but I really don't know how this could compete against what Google AI Studio offers for now. Studio even has free quota (while being completely opaque).
I once feel Kagi is some fresh air that we need for search engine. But now it seems to be more and more directly competing against the other LLM web summery products and imo Google/Gemini is light-seconds ahead in that space. The AI Overview is an ultra water downed version of it and it's still somewhat usable.
Perhaps this move is signaling that they'll be back focusing on search itself more though (I hope so).
Huh, I didn't realize they had a standalone AI mode. I've used the summarizer a few times and found it to be occasionally useful.
Please keep Assistant, it's an easy to use alternative to OpenAI, and is not intrusive. I will say that I find the GUI infuriating though... It's never clear where the cursor is when you start a new chat. Don't understand the low contrast layout with too much whitespace.
I’m a pro subscriber. I like Kagi, but I think it is getting lost. Instead of focusing on being the best search engine in a world where other search engines have become deeply enshitified, they are joining the enshitification train.
Start by removing features and adding a different subscription. Next add one ad from curated sources. Then add ads everywhere.
I will probably cancel my subscription soon and spare myself the disappointment.
I've switched to lumo if I need an ai so that's fine by me.
I've been a Kagi Pro user for several years now, and to be honest, the models from the pro tier aren't that useful compared to the free versions of Gemini & ChatGPT.
For me, I pay for Kagi pro for search without Google/Bing enshittification, their Translate (which I use quite often while I'm learning German/working in German - better for me than Google Translate), and their Summarizer. I pay for Claude, and also occasionally use OpenRouter for my AI needs.
Well, the solution is easy if users just paid for AI features separately.
I honestly think that AI goes against Kagis's main purpose. I see it as a tool to search the Internet. Carefully craft what you think is relevant, try to remove as much AI generated crap from your results, etc.
By using AI you are doing the opposite. You are letting some random AI get the results for you.
That's assuming Kagi assistant is the AI that searches the Internet. I don't know, I have never used it. I use Kagi search every day though
I use the assistant daily, often instead of search. The search is good, but the sad fact is searching www has been getting less interesting fast and there simply isn't much of value to find (maybe if they added an option to search in the Wayback Machine? I would pay extra for that!). Adding a LLM on top of search results and to fire off multiple searches makes it possible to sometimes squeeze a little bit of actually useful information out of the cesspool that is modern web.
Without the Assistant I would probably go back down to the lowest kagi tier.
While the assistant is a decent add on, it fails as a standalone product. It just isn't in the same league as Claude, ChatGPT, Gemini, or even LeChat, and I’d hate to see them waste too much development time on it.
I recommend Kagi focus on being a sidecar to AIs rather than try to offer it in any way; I'm hoping the API comes out and I can get proper search results for my local LLM.
Unless they're building their own LLMs, best not to annoy people who already see random LLMs everywhere.
Feels like a rug pull.. smells like enshittification