"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)."
Honestly, I just want my money to be used to improve good old match-keywords-against-index search.
(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 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.
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 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 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.
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.
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.
Discussed at: https://news.ycombinator.com/item?id=46716806
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.
> 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.
Thank you.
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.
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).
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.
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.
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.
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 am interested to learn if anything else is coming besides a billing change. Like will ultimate/assistant subscribers get access to MCP?
Unless they're building their own LLMs, best not to annoy people who already see random LLMs everywhere.
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.
Re: image gen... it's a search engine. Why would I need my refrigerator to toast my bread?
So to answer your question, while charts might not be particularly useful for a search engine a tutor certainly benefits from them.
Given that window dressing, having toasted bread in a kitchen that isn't selling my data is something I want.
Also, there news app is pretty underwhelming.
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.
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 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.
Without the Assistant I would probably go back down to the lowest kagi tier.
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)
The quick model is there to be mainly quick.
> 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.
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?
Though I'm sympathetic to the users for whom this would basically be a strict downgrade in featureset.
> 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.
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.
https://kagifeedback.org/d/1338-provide-a-plan-without-ai-fu...
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.
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.
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.
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.
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.