195 comments

[ 3.2 ms ] story [ 242 ms ] thread
I can't stress how much better Kagi is than Google, I'm a paying user. No more scrolling through tons of ads. Filtering by time and language is great, no more GitHub and Stackoverflow clones in the result. Thank you, I actually haven't realized how small Kagi still is, given how polished that product already is.
I'm in the same boat as you. The results and polish justifies my monthly payment to them.

Given how impressed I am with Kagi search maybe it is time to check out Orion.

Eh, ads are not something one has to put up with unless one chooses so, and filtering by language worked well for me with DDG (can’t remember about google, it’s been so long). I do enjoy their actually unique features, though ;)
By ads I believe the parent means junk domains "top10cameras2022.com" and not intentional ads.
Actually, I meant both. The junk domain part is also easily solved in Kagi with Website ranking adjustment and blocking.
Some searches with Google have become absolutely unusable, a flood of paywalled quora and random blogs that blabber without answers just to try and sell you a solution to a problem you don't actually have.

Kagi allows you to remove all that trash and actually find what you need.

Does it block pinterest by default? ;-)

-- Aside - why dont we have a keyboard key to switch an entire line from ALL CAPS to all caps or to All Caps... (This would be a great key)

-

Is there a way something like Kagi can also search google and ddg and ad results to a search then rank the relevant ones based on which ones users click in the results in addition to whatever/wherever Kagi indexes from? (or is that how it works - because the goog crawler is massive and hard to replicate is it not?)

Serious questions, as I do not know...

I think that sort of auto-ranking would require more tracking than they want to do.
> Is there a way something like Kagi can also search google and ddg and ad results to a search then rank the relevant ones based on which ones users click in the results in addition to whatever/wherever Kagi indexes from? (or is that how it works - because the goog crawler is massive and hard to replicate is it not?)

They don’t connect any searches/clicks to user accounts. But you can rank sites yourself like this: https://i.imgur.com/Po7mG7Q.png

> Does it block pinterest by default? ;-)

One of the first things I tried with them. I have the option to down-rank websites I don't like. I can block or lower their ranking.

I saw that in the video on the front page. It looks like a killer feature. I use DDG currently and am starting to get annoyed how Wikipedia is almost always the top result for any broad term.

I'm not sure it's a $10/mo killer feature, though.

I believe it does search using multiple other engines (google, DDG, Bing?) and lots of other ways as well (archive.org searches) but for ranking you have to choose to do that yourself
Agreed.

Hated what g! was becoming, so I started using ddg, liked it at first but had to often modify my queries with a g! but I still used it, the experiance was't great, but it somehow worked, then the controversies around ddg started to come out. At the end of the day you gotta keep the lights on, if your users are not your customers then someone else is going to have to help you pay the bills.

Used kagi since it was recommended here on hn. Used it for a week(-ish) and instantly fell in love with it, the lenses feature, the ability to rank search results meant I could block out websites I hated seeing on my search results and boosting the ones I cared about. Now I am a paying customer.

Oh yeh, one more thing, for the past few months I don't remember a single instance of me modifying my queries with !g. Now I promote them every chance I get, I even got called a paid shill on reddit a few weeks back :D .

What are the DDG controversies?
Not a controversy, but my gripe with ddg is whatever they do when I click on a result that puts the title and icon of the target page in my history, but when I click on the link in history it takes me to the ddg search results page, not the actual page.
This drives me crazy but I thought it was just my computer
The CEO and founder of ddg announced on twitter that they will start moderating content on the Russian invasion of Ukraine [1]. While I agree disinformation is a problem, I want my search engine to treat my like an adult, just do your job of retrieving the information I asked for and let me decide what is mis-information and what is not.

It starts with a cause we all can get behind, sometimes its "think of the children", another time it's regulating information from people supporting orcs, and who knows, how do I know things I care about might get suppressed in the future? Stop the nonsese, let me be the judge of what I consume.

The other instance I remember was the whole ms tracking situation[2], if your whole selling point is "Privacy simplified" then do just that without all the nonsese. There were a few others (#duckduckgone) before I stopped using them.

[1]https://twitter.com/yegg/status/1501716484761997318 [2]https://twitter.com/thezedwards/status/1528808759027331072

I think competition in web search is desperately needed and Kagi is promising. I used it during the free public beta and the search results were comparable to Google - sometimes better, sometimes worse. I sincerely hope they continue improving their search engine and other products, and make a well deserved profit from that.

However, this seems a little disingenuous to me:

> We do not log searches or associate them with an account, by design. This is a software architecture decision (we simply don’t need this) and there is no simple switch one can flip to enable this in case of a court order.

I have no reason to doubt this that their software is like that currently. But I also know that you don't need to change your software architecture to log user queries.

Yes, but let's not yet again confuse "search engine" with search service (proxy or metasearch if you prefer). I'm biased so will resist the temptation to say more and only point you to the comments which are currently at the bottom of this discussion: https://news.ycombinator.com/item?id=32677872

Having said that I congratulate Vlad and his team. Competition and innovation, in browsers and web search, is indeed needed. We welcome, and are willing to support, all new challengers making genuine efforts.

I’ve been using Kagi for three months now, and 27 queries per day sounds about right, it could be more but when I open incognito tab on Brave it asks for a login which is super irritating, so about 15 queries a day still go to google. So far Kagi results are so good I forget about it until the end of the month $$$.
If the extension doesn't work, set the kagi search url to include the token as given in kagi settings.
I pay for Kagi after a recommendation here on hn.

I have been very happy with it and how clean/fast it is to find what I want.

$19 per month for single user without regional pricing is just insane given localised results (here in Japan) are really bad.
$19? Kagi is $10/month. Family and business plans coming soon.
Individual Plans

Kagi Unlimited - $19/mo or $180/year ($15/mo) or $288/biennial ($12/mo) - Original Kagi experience, with unlimited searches

Edit: I see I misunderstood the original comment. Disregard my message. :)

I don't know where you live but for me it's 10 USD per month.

Says Premium | USD $10 /per month | Unlimited ad-free, private search.

On their /pricing page.

I was hoping there’d be a conversation here about the price increase but so far every comment is about Kagi in general without ever reading the actual blog. They also list a $5 a month plan. What’s not clear is when they’re launching this and if they’ll grandfather in the current customers.
"If such change to Individual plans is to occur, we plan to grandfather-in all early adopters (meaning all current and future paid customers, up until this change) allowing them to keep their existing subscription price as long as they don’t cancel it."

It mentions on the blog post that they'll grandfather in existing customers.

It says on the announcement that they will grandfather in existing customers.
That's what they want to increase to, but they say they'll grandfather old users into the old prices.
(comment deleted)
It is currently, but the article linked here mentions that they are considering an increase because the current price is likely to not be sustainable.
Hadn't heard of Kagi before this, but seriously considering it. $10/mo I might be able to swing. I would not pay $19/mo but that's just me.
(comment deleted)
Three things:

1. You don't know what privacy is. "We believe that Kagi has the most user-friendly privacy policy of any search engine out there. However, privacy and anonymity are not the same things. For example your parents can know everything about you and still fully respect your privacy." Ask any kid if privacy is their parents knowing everything. You're using it as a marketing term.

2. This is illegal public solicitation and an SEC violation: "anyone passionate about Kagi and our vision for the web will have the opportunity to invest as little as $5,000 USD and join the ride with us. It would be structured as a SAFE note and we would use the funds to expand the team and accelerate our product vision."

3. Stop abusing HN to shill your paid search engine. "We are pleased to see that Kagi is propagating fast through Hacker News, Reddit and Twitter with word of mouth from existing users." Kagi shills hijack every single Hn discussion thread about search. You is just as bad. Just. Stop.

It looks like you have some substantive points, but can you please not be aggressive like this on HN?

I appreciate that you're watching out for the quality of the site, but you broke the site guidelines here, which is no doubt why users flagged your post.

Beyond that, your points would be much more persuasive if you made them more neutrally, without this kind of attack.

https://news.ycombinator.com/newsguidelines.html

I am sorry @dang I apologize for the aggressive tone of my comment and not making a better argument without the anger. The promotion felt blatant. I let it get to me (every comment I believe at the time - and many other recent posts). It was wrong and I'm sorry.

Fwiw this was the source of the self-promotion.

https://kagifeedback.org/d/887-comment-on-hn-discussion-of-k...

Just to let you know, I flagged your original comment because, as dang said, it was too aggressive, personally attacking me. I’m enthusiastic about Kagi, but certainly not a shill.

The kagifeedback thread you posted was also by another (imo too enthusiastic) user of kagi with no relation to the devs.

Hi Danny

1. Attention was brought to us for the unfortunate phrasing of that sentence and it was edited in the post

2. I am curious what rules are violated by mentioning plans for an investment round?

3. While I am glad this discussion happened, it was not submitted or solicited by any Kagi team member. We published a blog post this morning and our users (I assume) took it from there, as these things tend to happen.

Finally the post you linked to in your second post was from a clearly excited user (not Kagi team member) on our public forum.

I really like how listicles are compartmentalized to one section of the results, instead of taking over the entire page.
Paying user of Kagi here. I recommend it to all my friends. They have a family plan on the roadmap which is great for me. I have left Google search altogether and never looked back.

Over time I have managed to block most of the websites which annoy me the most from the results. For example I have blocked w3schools and boosted MDN or the annoying ads disguised as blogs or websites which pollute the search results by mere SEO enhancements.

They mentioned that for family plans, parents could share "lenses" with other parents. I would love for this to be extended to everyone so there's a "lens marketplace" that people could share their curated lists of sites so that filter lists of shitty websites can be crowdsourced.
That would be great, for example if I could find a lens that helps me search in topic I care about. Imagine Google’s :site but for a specific and custom group of sites with custom rankings.
W3Schools is not as bad as it used to be.
I tried many search engines, but I always ended up going back to Google for many queries. Kagi it's the first one that I really feel like Google doesn't add anything any more. Sometimes if I can't find something I will try google, but I never actually got better results, which means my query was just bad.

I am just so happy with Kagi I can't explain. I am not a huge fan of subscriptions and I carefully curate what I pay for monthly. I love paying for Kagi! I wish them only the very best!

Same here. Perhaps your happiness is due to Kagi being (right now anyway) an entity whose incentives line up with what's best for its end-users?

This could definitely change, but today, it looks like a genuine attempt to break out of the terrible local maximum (ad-supported Internet) we're stuck in.

Exactly the same here. Every once in a while I think "let's see if Google has anything better for this query", and it never does.
I would really love to use Kagi full-time. I'm currently on DDG, but something about them that I can't quite put my finger on rubs me the wrong way. Maybe it's the amount they apparently spend on advertising, or maybe it's that they're just sourcing their results from Bing.

Anyway, at $10/mo I’m not quite prepared to put my money where my mouth is. $5/mo would be much more attractive, and I would guess they'd have better profit at that price. How much does each user cost them?

Edit: Thanks everyone for the pricing info! I guess in a world where we expect search engines to be free, we don't realize how expensive they actually are to run (at least at smaller scales like Kagi).

Intuitively I would have expected the expensive part to be building the index and for each query to cost fractions of a cent. I am extremely surprised to learn that each query apparently costs them $0.125.

It says it costs them $12.50 in computing costs per 1,000 queries. It says the average user makes 27 searches per day, including the 10% of free users limited to 50/month, so probably about 30 searches/day for paid users.

$12.50 × (900 ÷ 1000) = $11.25/month/user

One thing that I actually love about Kagi, is that they tell you how much you cost them. In the settings, under billing, there is a usage tab where it says how much your queries cost. Apparently last month I "used" $2 more than I paid, hopefully I am making it up with the word of mouth I spread.

There's more info in their FAQ https://kagi.com/faq#cost

Wow I'm more than $10 in the red so far after being subbed for 3 months, with an average of around 1200 search per month!

I really hope that their high costs are due to some inefficiencies (e.g. with their current scale) they haven't been able to work out yet, otherwise that does not bode well for the concept of a paid search engine... or maybe I'm just an extreme outlier? I pretty much only ever search for documentation and things like that while I'm working, but I apparently do that a lot lol.

The price they're using seems to be $0.0125/query. I'm assuming that includes business expenses, like dev/admin/salary costs, and not just the raw costs of executing a query.

I guess a price hike to ~$15/mo would cover my particular usage, and I might be willing to pay that much, although I don't think many others would. $10 is already more than a lot of people are willing to pay just for search. Maybe if they bundled some other services into the subscription it would be more appealing? People expect search to be free, but are willing to pay for other services that are probably much cheaper to operate (VPN, password manager, cloud storage, etc)

> I'm assuming that includes business expenses, like dev/admin/salary costs

It actually does not, it covers infrastructure, but not salaries. Salaries for Orion/Kagi devs are currently paid by Vlad out of pocket, he mentioned that on Discord.

I’d say with 1200 you are slightly above average from what people shared on discord (I only hit $10 in costs in August for the first time).

When the post was discussed on discord, I mentioned that $19/mo would be too much for me, on the other hand I could do $12/mo with a 2-year pre-pay (not that I have to, thanks to the grandfathering)

You are, that is a fantastic way to look at it!
I feel the same about pricing; $5/mo is about what I’d expect to pay. I’ve been trying it at the current rate though to see how it goes. A fantastic thing they do is tell you how much you’re actually costing them. My billing page says:

     Month  Searches    Cost
  Jun 2022       875  $10.94
  Jul 2022       773   $9.66
  Aug 2022       842  $10.52
  Sep 2022        10   $0.12

  Your total payments: $30.00
  Your total usage: $31.25
  Your payments to usage balance: -$1.25 [Tip the difference]
This is fantastic, and I hope it can be a sustainable practice whether it's them or other larger companies. Transparency is positive when they're the little guy with a nerd audience, but most companies depend on information asymmetry to mess with our sense of value.
made me curious how many search queries I'm doing: averaging 2000/month:

    sqlite3 ~/.mozilla/firefox/*default*/places.sqlite "select strftime('%Y/%m', last_visit_date/1000000, 'unixepoch') as yearmonth, substr(p.url,9,14) as search_backend,count(1) as hits from moz_places p where (p.url like 'https://duckduckgo%' or p.url like 'https://www.google.%/search%' or p.url like 'https://www.qwant%') group by search_backend,yearmonth order by yearmonth asc,hits desc;"

add any search engine you use with another 'p.url like'
> Anyway, at $10/mo I can't quite put my money where my mouth is.

Starbucks coffee fallacy :) I.e. it's cheaper than a Starbucks coffe so it must be cheap enough.

> I'm currently on DDG, but something about them that I can't quite put my finger on rubs me the wrong way.

Have you given Brave Search a try? I honestly didn't expect to switch, but after trying it out for a while I ended up making it my default and haven't looked back. Brave is mostly using its own index (in my case, for 87% of my search results).

> Anyway, at $10/mo I can't quite put my money where my mouth is. $5/mo would be much more attractive, and I would guess they'd have better profit at that price. How much does each user cost them?

Not really. Between free and paid users, TFA says they average roughly 27 searches per user per day; 1000 searches costs them $12.50. If each user (TFA says roughly 10% of searches come from the free tier) is averaging 810 searches in a month, that puts their bare cost at close to $10.13 per user per month, not including salaries and other operating costs.

I'm one of roughly 2.6k paid users at the moment; they estimate the need for 25k users to be sustainable. There's so much about google and most of its current alternatives that legitimately disgusts me- voting with my wallet is the only option I really have.

>that puts their bare cost at close to $10.13 per user per month, not including salaries and other operating costs.//

You're saying it costs then 1⅛ cents per search excluding wages and other operating costs.

That seems like they must be vastly over-paying for something? What does that money pay for and who is paying the wages+operational costs?

Edit: this comment [0] indicates every search is actually 3 searches, two at commercial resale rates (Google, Bing) and one against their own index; that probably explains it.

[0] https://news.ycombinator.com/item?id=32678059

> You're saying it costs then 1⅛ cents per search excluding wages and other operating costs.

All I can do is relate what the article says :).

> That seems like they must be vastly over-paying for something? What does that money pay for and who is paying the wages+operational costs?

Well, they use a couple of paid APIs for their queries (as you noted in the edit of your comment). They say in TFA that they're currently breaking even on bare search costs and that the founders are picking up the tab for most of the rest of the expenses.

Based on other data here, it seems like it's $0.0125 / query
The $0.125 is fascinating. According to my stats I query Google between 2500 and 3500 times a month. That would be a $300/mo. bill for Kagi?

I guess most people don't query as much as I do, but seems like it would need quite a number of people underquerying to make up for users who would be similar to me?

It is $0.0125. So more like $30/mo
> Orion is zero-telemetry by default so auto-updates are disabled by default

This is a very surprising choice. I'm also leaning towards the side of irresponsible. Web browsers are probably the biggest attack surfaces users have. They are running untrusted code thousands of times a day. That isn't even counting all of the other attack surfaces such as HTML rendering, media decoding and playing, WebGPU type stuff and more. The stream of vulnerabilities has been constant so far and there is no evidence that the stream is over.

If this was my product I would disable everything but auto-updates, and make a clear policy about what data I track on update downloads.

Trading security for privacy likely results in less privacy once you get pwned.

I have used Orion to test it out and it prompts you if you want to update when a new update is available. Coming from a Unix background I prefer this method over a forceful update.
Hmm, something is wrong with their wording then.

> The way we measure this is through auto-update “pings” to our updates server.

> This measures only users who have enabled auto-updates

> auto-updates are disabled by default

This seems to say that they won't do auto-update checks by default. But I guess they are doing auto-update checks but not tracking the checks, just a "ping" (whatever that means if it isn't an update check).

We are not doing auto-update checks by default (as that would be a form of telemetry). The pings we count will originate only from users that manually enabled them.
> This is a very surprising choice.

I believe that it's not too surprising if you know that most of their users are from HN.

The HN comments on every telemetry-related post that I've seen have been violently, overwhelmingly, religiously against telemetry (not tracking - telemetry proper) in any form.

Even if those comments aren't representative of the views of the readers as a whole, if you're building a company and those comments are your biggest insight into your users' feelings - you're going to do stuff like this.

(and yes, it shouldn't have to be said that this inhibits security, but then again most of the anti-telemetry HN commentators are fine with that)

> and yes, it shouldn't have to be said that this inhibits security, but then again most of the anti-telemetry HN commentators are fine with that

The end result of security at all costs is zero user agency so this is not at all surprising. Like anything, auto updates are a tradeoff that is neither without cost nor provides 100% protection.

The choice to enable auto-updates is prominently present to users on onboarding. So while default off, it is one click away from being turned on for every new user using Orion.

Also there is a question of the real risk. WebKit and macOS are generally very secure. We had to intervene only once in the last 4 years with a (medium) security related WebKit update (and we shipped it a week sooner than Apple did with Safari).

We believe that the two facts highlighted above minimize any security risk.

On the other hand every telemetry, including auto-update pings, sends personal information (at the very least user's IP address) to the browser vendor. Almost every other browser on the market is ad-supported (directly or indirectly). Telemetry and ad-supported businesses can be a problem.

We do believe that browsers do not have the right to do this (at least if they want to call themselves privacy-respecting). In our view it should be on users to decide if they want to trust the browser (browsers are "user agent" after all) and every telemetry, including this, should be opt-in.

Simply put, if a browser is not zero-telemetry it can not be called privacy respecting, and any browser using telemetry by default should at least disclose it to users on the first run.

Google has a search API. Does Kagi use this on the backend?
This is the most important question. Are you piggybacking off of existing search providers, or is this a search engine from the ground up?
Looks like they do. From their FAQ[3]:

> Our searching includes anonymized requests to traditional search indexes like Google and Bing as well as vertical sources like Wikipedia and DeepL or other APIs. We also have our own non-commercial index (Teclis), news index (TinyGem), and an AI for instant answers.

[3] https://kagi.com/faq#Where-are-your-results-coming-from

They use bing and google APIs (that’s where the cost mainly comes from as every querye requests from both), their own search index, and smaller indexes.
I've been a subscriber for a while now - completely replaced Google, even for esoteric tech queries.

Only feedback I have is, if someone on the team gets some free cycles, please make the calculator a little more user friendly. Mostly looking for the ability to Kagi search an equation, and then tack on additional operations to the equation in the results page. Right now I have to reach for the mouse, click the small text field and type ans*. Would be nice if it just inferred that I want to use the previous answer.

(comment deleted)
In regard to anonymity, I'm curious if the kagi users can form a closed tor-like network that will insure that all queries come from an authenticated user and at the same time the query source is obfuscated.
I joined the Kagi beta two weeks before it ended and have been a paying user ever since. I find the price to be quite steep but so far it have no regrets and will most likely continue my subscription. I average over 2000 queries per month.

- Search result quality is at least equal to Google for most queries and far superior for queries involving discussions or technical topics. Sadly, it is inferior for queries in my area of expertise (Swiss law).

- Kagi is the only alternative to Google that I know of (and I have tried many) which is able to produce good regional results and mix them - when appropriate - with international results. Most privacy focused search engines are very US centric.

- Blog spam and copy cats are not completely gone but the situation is at least bearable, especially after continuosly filtering them and boosting credible sites instead.

- Customization options, filtering options, speed and site layout are also far superior when compared to Google.

Apart from the lower quality results from queries involving my area of expertise, the only thing that still keeps me using Google Search are the reviews from Google Maps.

All in all, I recommend supporting this product.

I also have been using since beta, and agree with you. For me, the ability to boost and block certain domains is a killer feature. (There are probably solutions for other services, but to have it built-in is great.)

For developer-related queries I`m pleased with the results!

Odd you say the price is steep - I have been using it mostly exclusively for a few months now and I haven't even thought about it cost. It must come out to under 0.1¢ per search.
> It must come out to under 0.1¢ per search.

I was struck by their pricing page [0] which says that "it costs us about $1 to process 80 searches" and that "at USD $10/month, the price does not even cover our cost for average use". I don't think they're going to be right that beta testers will do more searches than average users, and I think that they'll attract heavy searchers (the ones most noticing and disliking Google's quality). I would be surprised if I search less than 50 times a day, and that's probably just during work. If I was a paid Kagi user they'd be losing money on me. They say their price might have to go up, I don't see how they're going to avoid that.

[0] https://kagi.com/pricing

I agree that it was an impressive product but couldn’t justify it based on the cost of living where I live and relative to other things I subscribe to (including my Fiber connection). I do hope things like this succeed.
That sounds unsettling, given the normal trajectory for software companies running at a loss (get VC, adjust goals to match VC desire, etc). For a privacy focused company it feels like they already lost the normal leverage line item of doing "something" with user data.

I dunno, "if you're not paying, you're the product", but how does that fit if you are paying, but what your paying isn't actually enough? Whats the other product?

Hope they do well.

I can afford it but when compared to other stuff it just seems steep:

- Half the price of my fiber subscription

- Double the price of my VPS subscription

- Same price as my mobile subscription

I’m jealous of your other subscription costs…
Honestly there has to be something wrong with software pricing. Slack cost per user is similar to giving everyone in company sim card, while cell operators have to maintain both, massive physical infrastructure AND software infrastructure. Seems like most of SaaS environment is horribly inefficient.
I'm using Kagi in France. First time I'm not actively looking back to be on Google (which sometimes I still do). Congrats and continue your effort !
"For example your parents can know everything about you and still fully respect your privacy."

This was raised in a flagged comment, so I still felt it was worth highlighting. I don't find analogies work well for technical decisions. I usually prefer that a company tell me what is and is not private and anonymous so that I can made a decision on if I should use an application and on how I will use it. There is no need to justify the decision you made and in some respects lecture us on why the decision was justified. Just be transparent about what is and is not and that's enough.

Heh, I actually raised that on their discord because it’s so cringy :D
Removed the analogy from the post since, agreed it can sound condenscending and wrong (missed that in the initial draft).
I pay for Kagi on principle. I’ll just come out and say it though: Kagi is missing 30% of the internet, at least. It may just be that Kagi is in the landing phase of “land and expand”, but I do think that long term, and generally, search engines need to be indiscriminate. Sure, interpreting queries and providing flavored results is fine especially when I’m paying so I know the flavor is in my interest, but excluding entire categories of content is not.
Can you elaborate? What is missing?
I love this browser but I have problem with missing autocomplete on text inputs and with latest update I can’t play any video on youtube. Also would be possible to not animate address bar when entering url?
You can post these to orionfeedback.org
Been a paid Kagi user for 3 months and it's been a wonderful replacement.

If I could ask for one thing, though I don't think it is up to Kagi. I use my Android phone with the Microsoft Launcher and Edge and I for the life of me cannot figure out how to make it Kagi the default search engine.

Besides that, it has been a great service and well worth it! I wish the Kagi team well and I look forward to buying a shirt or hoodie in the future! :)

Kagi has been worth the money for me. It's got a lot of room for improvement, but compared to DDG (which I like) it's given me everything I could ask for. I don't like the subscription business model most of the time, especially for software, but in this case there is a definable cost associated with me using the product, so it makes sense.