Kagi Teams
blog.kagi.com
blog.kagi.com
Difficult to say if Kagi gives better results, but what's for certain is that it doesn't give worse results.
I am happy to pay 120$/year to support the business model.
The freedom of clicking on the first few search results without thinking whether or not those are ads and/or scams is priceless; a calm and premium experience.
In other words, there are fewer low quality results which raises the average quality of the results you’re presented with
Kagi does great with 99% of my searches.
I never click on ads. Back when I used Google, if I searched for something specific and saw both the ad and the search result I wanted- I just scrolled down to the search result. My wife would just click on the ad, and I think most people do. Beyond that specific example, she clicks on other types of ads as well. For me, I just see them as invasive and hostile. I’m not the type of user Google cares anything about, and maybe that’s why I perceive a drop in quality results.
most normal people have been have been shifting their search habits. tiktok (banned soon), youtube, reddit (though many use google to search reddit since their native search isn't good)
Which, frankly, I'm willing to pay for. I think it's a good product.
docker run -d -p 9017:8080 --name searxng_data -v "${PWD}/searxng_data:/etc/searxng" -e "BASE_URL=http://localhost:9017/" searxng/searxng:2024.8.31-769b2ed03
Its basically no maintenance after this setup. I can mess with all the settings (except results format) from the main interface.
Also as a former Google employee... I dig the idea of contributing funds for some competition in this space.
I think the code is still running, just not publicly. Give me access to that!
I doubt this is true given how the content on the web has changed.
It is fun to be nostalgic though!
Yes, because there’s more stuff and different stuff than there used to be.
What makes you so sure the old algorithm would’ve worked as well with more and different stuff to sift through? Other than the fact that you really want to believe it.
> it’s clear that the algorithm is skewed for political reasons and likely also economic reasons
Bud, it’s a business. This has always been true.
The new $10 all you can eat plan is so much better in that regard. I don't think I would still be using it if they hadn't introduced that.
But the fact you can use kagi to filter out links with ads/trackers inside them,means sometimes great results rise up, and that can be very valuable.
Primarily I've found the source of issues like this are the lack of indexing anymore, and it's not just Kagi. A lot of tech discussions now have their own Forums or Discourse that are poorly promoted and (seem to?) have a robots.txt that says not to index them. I often have to manually go to those sites to run searches and find results from years ago. Also a lot of prohects keep moving away from Forums or equivalent, and using Discord, Matrix, or IRC, none of which end up searchable. No one is going to use those the same way no one is going to call you on the phone, and everyone just has to keep asking the sme questions if they do use it, so I always take that as "we don't support this software/product anymore" and rarely expect to find anything technical about it.
Despite having so little paying users. Not even 35,000 lifetime paying users.
They can really thank the HN echo-chamber, because without it they would have been forgotten a long time ago, nobody knows about Kagi outside of here.
There are plenty of billion dollar companies that make users pay for something that has always been free (note taking apps just off the top of my head).
Searching for real time information that happens within a day is way worse than google. I realise how much I missed the shopping search of Google. Google map is still irreplaceable. Google search has better UX, I particularly missed the favicon that identified the site next to URL.
Image search is also way inferior
This is fair. On the other elements, I’ve found Kagi to be superior. In particular, having shopping results separated from recommendations (and summarising them with quick answers). As for favicons, uprank or even anchor your favourite domains [1]. That’s what the favicon was proxying.
[1] https://help.kagi.com/kagi/features/website-info-personalize...
But you didn't mention the things that Kagi does better than Google which is in my experience: everything else.
Google is still "free" though so there's nothing stopping you from adding !g to your Kagi shopping searches.
I have a bookmarklet for Kagi's Universal Summarizer, and if a long article or video doesn't immediately tell me "what it's gonna tell me" then I just click it.
javascript:Object.assign(document.createElement("a"), {href: `https://kagi.com/summarizer/index.html?url=${encodeURIComponent(window.location)}`, target: "_blank"}).click();
Ime image search is far better in kagi. I found some specific images that i was looking in google _for months_ and i could not find. Just because they were some older blogposts probably and google penalises older posts, and because of the AI generated cesspool that it is becoming.
- Professional ($10/user/month): Unlimited search
- Ultimate ($25/user/month): Unlimited search plus LLMs
You can mix and match users on different tiers.
Still, I'm only using Kagi Search and not any of their LLM Agent stuff.
I am still waiting for better pay-per-use LLM access, rather than $15-20/mo subs.
I am from the other side. My LLM bill would be hundreds of dollars in month if I would use APIs but these flat prices really work for me.
Suggestion: it would be cool if the Teams plan included a “personal” plan you can use with a personal account for each member. 1Password does it this way - if you have 1Password at work, you can “link” your personal account to it to get a family plan free of charge - but it’s linked for just this purpose, your employer doesn’t get access to your personal 1Password.
This is great to keep professional and personal accounts separate.
For example, I use this to put Encyclopedia Brittanica results above Wikipedia when the former has a matching article. Similarly, I prefer all3dp results for 3d printing searches.
It's the only way to fix search: designs which are detectable attempts at monetization should be nerf-able by the user. A new search game must be set up where seeking money kills your site rankings in search. I only care about sites where the webmaster is paying and working to tell me something he cares about, for free, because he is obsessed. No I don't want to buy your stupid product/service, I don't have money to spend and spending it on anything except survival and advancement is counter to my strategic interests. I want autist site and forum results, I do not want marketing-driven money making results... period.
If the site makes money, it prima facie doesn't have my interests at heart. This heuristic works really well for me.
Based on a cursory look Team Ultimate is the same as individual Ultimate, feature-wise.
- AI Assistant with access to leading LLM models, grounded in Kagi Search
- Universal access to:
- OpenAI
- Anthropic
- Google
- Meta
- Mistral
However, based on that page, the team plan cannot be prepaid annually for a discount, so if you are willing to do that the individual plan is 10% cheaper.
That's a bold claim
It's a fair assesment, unless you're nitpicking about "all" in there.
Kagi plus a few good LLMs (I use local ones via ollama because I do that, and also Claude sometimes) gives you research capabilities that are as good as the pre-enshittification Internet and often better. I use LLMs like brainstorming partners and research assistants and then use Kagi to search for real data to verify results and make sure the LLMs aren't hallucinating.
Going back to Google I'm sometimes shocked at how absolute trash it is. I'm sure they're maximizing their internal KPIs.
Of course I know the real problem with Google is that I'm not the customer.
When I install Linux on a new computer, I install Firefox from its package manager, log in with Firefox Sync, and this switches my default search engine to Kagi. So I never once visit Google, not even to download Firefox.
I just tried going to Google, and it is simultaneously satisfying and saddening to be met with a cookie disclaimer pop-up: Saddening because I'm reminded that ubiquitous search access comes at the cost of tracking, and satisfying because I never once visited Google on this one-year old computer (after which the pop-up disappears).
don't you have your identity (through payments) tied with every search query you make when using kagi?
Specifically, I trust the CEO and founder.
If I didn’t trust anyone, I’d visit DuckDuckGo using uBlock Origin via a Mullvad VPN paid for with coinjoined crypto.
They fail in a nasty way too - nobody expects you to block them - so you'll simply see blank pages, etc. Likewise for Firebase which it seems 40% of startups use.
Been using Kagi since its beta and I’ve never used Google intentionally since.
Really a minor annoyance tho.
Nothing especially sensitive, I just consider it basic online hygiene.
I regularly compare results for the same query between Kagi and Google just to check if still makes sense, 18 months in and every single time I do this comparison I feel more validated to pay for it.
Of course, totally personal experience but I've got so used to the quality of results from Kagi that Google has become almost unusable in comparison.
An AI equivalent would be the JetBrains AI for the same price which seems to be integrated to CLion but I'm still not completely sold about it.
I’m also curious about the LLM integration with Kagi. I currently pay for a Claude subscription but would be happy to drop that and pick up the $25/mo Kagi if LLM access was similar. Is there a dedicated chat app? Interface?
You can provide custom instructions and set up multiple assistant flavors. For example, you can have one that highlights controversial information about a topic, another one that jokes about everything, etc.
All the assistants have optional web access. They use Kagi search to look up relevant information for you. You can apply a search lens to restrict web results to particular domains.
You can set the assistant to use different LLM APIs. They currently support Anthropic (Claude 3 Haiku, Claude 3.5 Sonnet, Claude 3 Opus), OpenAI (GPT 4o, GPT 4 mini), Google (Gemini Pro), Mistral (Pixtral, Large) and Meta (Llama 3.1 405B).
Kagi maintains their own LLM benchmark. Their goal is to (1) ensure that models haven’t ever seen the questions before and (2) emphasize reasoning and instruction following, which they’ve found to be most important for the assistant. https://help.kagi.com/kagi/ai/llm-benchmark.html
Not a bad idea, just quite hard to pull off when Microsoft starts pushing Bing and Copilot to their enterprise clients more aggressively.
The amount of effort to get anything paid approved is often similar whether it's $1 or $100. Above that, it probably gets exponentially harder but that's fair.
I think the original comment pointed out that justifying any new purchase (especially for something that's generally taken for granted as free such as web search) is hard, not necessarily whether it's $10 vs $20.
Anonymity > decent search 8 days a week!
Must be your X echo chamber. I haven't heard about Perplexity in ages.
> Do they both search -> put results in an LLM as input text?
No. Kagi does have optional LLM-based answers, but at the core they present as a search engine and not a chatbot like Perplexity.
Perplexity is yet another AI startup trying to find a way to monetize LLMs. Will they survive when the bubble pops? Idk. Do I trust them with my data? Fuck no. Just look at all the advertising they're investing in and ask yourself how they plan to make that money back.
For the unaware, Yandex Search is tightly integrated with the Kremlin.
This was a good nudge to check in on how close Yandex is to the Russian government. Answer: way closer than they were 5 years ago. Russia basically nationalized them, and routinely modifies search results to align with the government’s preferences.
https://www.zois-berlin.de/en/publications/zois-spotlight/th...
https://kagifeedback.org/d/5445-reconsider-yandex-integratio...
> yandex is about 2% of the cost of your subscription
Source link: https://kagifeedback.org/d/5445-reconsider-yandex-integratio...
https://kagifeedback.org/d/5445-reconsider-yandex-integratio...
Apparently this has been happening for a long time but only recently drew attention when they announced integrating images. The founder cites the 2% number as if it's supposed to make it all okay, but when it comes to principled objection to funding Russian propaganda, it's hard to argue that we should be okay as long as the number per subscriber is small.
I've been a Kagi subscriber from the beginning and am up for renewal, and I've got to say here (since I can't say so there now that it's locked) that this is giving me pause. I've fought back against boycotting Russian-founded companies before (anyone remember when JetBrains came under fire?), but this is Yandex, not some little company that happens to have a Russian founder.
Edit: I just want to draw attention to the fact that this subthread was the top comment and now has been downranked to the bottom. Whether that was moderator action or algorithmic, I disagree with the decision.
When a founder's/owner's moral attitude is this unprincipled, supporting their company will only make the problem worse. "Immorality scales with the business".
We are seeing this with Reddit and fake users. Kagi owners, employees and (prospective) users should ask themselves: "what good would even the best search engine be, if the world is already corrupted?" You will find what you support.
i like that they are using a diversity of sources with a diversity of biases for their information. i don't want them to just get their information from a single biased source in the west like literally every other search engine out there.
Why do people always turn to whataboutism rather than engaging with the criticism directly?
The desirability of diverse perspectives in search has no bearing whatsoever on that subject.
A non-deflective response would be something like: "I know paying money to Kagi indirectly supports Russia in its war against Ukraine, but I don't give a shit about that. What really matters is that my search results must include Kremlin propaganda alongside what I consider to be US propaganda".
Now, I suspect your response is going to be something along the lines of "But Google is American and America has done lots of bad things too." So to save you some time, I'll respond in advance: That too is a deflection. It is true, but it also irrelevant to the point at hand—many Kagi users do not want their money going to the Russian state.
My perspective is also that search engines like Google are bad, but they at least aren't run by the US government in the same way Yandex is run by the Kremlin.
"You can't criticize Kagi using Yandex because the US is also bad" is why it's whataboutism, and why that's not an effective argument.
"XYZ is also bad, so your criticism of ABC isn't allowed."
It would help somewhat if the founder could clarify whether funds are going to Yandex Russia or the newly split-off Nebius. If it's the latter I could at least satisfy myself that I believe the companies to be sufficiently distinct.
I really hope they reconsider their arrangement.
https://www.zois-berlin.de/en/publications/zois-spotlight/th...
> Volozh resigned as Yandex’s chief executive in June 2022 and left Russia; in August 2023, he publicly condemned Russia’s invasion of Ukraine. After lengthy negotiations and in the face of mounting political and legal pressure, in February 2024 Yandex announced the sale of its business in Russia to an investment fund consisting of investors with close ties to the Kremlin
I'm sure the Kagi team is reading this thread: you need to seriously reconsider your business relationship with the Kremlin. I somehow doubt the quality of your search is so dependent on that one search provider to make the (apparent) 2% of subscription prices worth it.