I've also been using Kagi for ~1 month and god can I testify for how fantastic it's been. You have to TRY to find blogspam and the allowance of blacklisting domains plus some other handy search customization features make it an absolute joy to use.

It may lack "instant answer" widgets or other fancy search engine features but it gets the actual "search" part of the equation so right that I find it astonishing how I ever used DDG/Google in the past.

You can enable Instant Answers in the settings, it is not as complete as other search engines yet, but it is improving :)

Yeah I've flicked it on a couple times but rarely notice much change, though honestly it's not a search engine feature I've come to desire, as even if you get a widget the majority of the time it may parse incorrect data or not even show what you wanted (which happens a lot with google's widgets IIRC).

Same. Over the years I've trialed most search engines out there, but always find my way going back there after at most 2 days of trying them, because I end up adding "@google" before every query anyways because the results are bad.

With Kagi most of the results are what I'm looking for. If they are not, I'll still try "@google", but so far with very few queries Google's results were actually better. The biggest drawback is worse "smart cards" results, but I hope they keep those optional/unobtrusive anyway.

The strange thing is that the feeling Kagi gives me, isn't even unknown. It just feels like Google circa 2010.

I think the hardest thing with trying a new search engine is the constant question "i wonder if the old one would have done this better"

I wish there was an extension that would search simultaneously on different engines and present them in columns side by side.

Like this? https://chrome.google.com/webstore/detail/double-shot-search...

I've signed up for the beta, but it's hard to shake the feeling that signing in to a search engine is a mistake. "We respect your privacy", "we'll never sell your data", I've heard these claims before and they've almost always been lies. They can tell me that they don't maintain an eternal history of all my queries, but how can I ever verify that?

Kagi ultimately will be a paid service. As I've noted in the context of other services (email), by providing revenue in the form of paid services (and only paid, no freemium tier) the service doesn't have to implement ads at all, and thus can skip the pressure to deliver more and more data in exhange for better rates.

I guess it comes down to trusting that they'll sit on that big stack of user data, exactly the same stuff Google used to build a trillion-dollar company, and continue to decide day after day not to sell it. Will they continue to hold out if VCs get involved?

Yes, I'm aware that all the same arguments apply to DuckDuckGo, and that Google & Bing already explicitly do sell this stuff, but the stronger the promises, the more I demand to see proof.

I think part of it is also an implementation pressure. If you put your dev work into building a system around managing subscriptions and processing user payments, its less easy to flip a switch and start siphoning data to 3rd-parties when they come calling.

In theory, this would be visible to end-users as a halt in feature roll-outs, because the dev team has to pivot to building ad-tech.

I hope to god VCs don't get involved; if they do I'd be the first to bail. I'm hoping that the revenue model makes VC money allergic to them in general. Bootstrapping is preferred for this type of service (see also: Pinboard, sr.ht)

Neeva has a similar ad-free model. The experience is getting better all the time.

Umm... Google and Bing track the hell out of you whether you sign in or not. DuckDuckGo claims not to but the sites you hit through it certainly employ tons of fingerprinting.

Unfortunately with the advanced state of browser fingerprinting you are always "signed in" when you search on any search engine.

> how can I verify that?

Write and ask them. GDPR requires that users can retrieve all stored information associated to their profile/person. Most large internet companies have this functionality.

I love this search engine, it gives me the same feeling that Google did when it became a thing. Their business model after beta will be that users pay to use it, and it has no ads. This is a very encouraging sign, and personally I'll be willing to pay for quality search without ads. I hope enough other people feel the same to make Kagi profitable and functioning for years to come.

Looks interesting, but am I crazy for thinking that $10/month is an insane price to pay for a general purpose search engine? Surely Google wouldn't making anywhere near $10/month off of me even if I disabled adblock.

Those two sentences have no connection. You can gain a huge amount of consumer surplus, even as the seller reaps almost none of it. Google's gain from ads bears little relation to your gain from Google. (This is why people are so much better off in markets: it's actually very hard for a business to get more than a small fraction of consumer surplus.) As it happens, when people try to estimate your gains from a general purpose search engine, it usually comes in at like $100+/month (search 'willingness to pay search engine' and think about how long it would've taken you to find that in a physical library, and how you wouldn't've bothered in the first place because such a search would be impossibly expensive). So, $10/month would be a steal compared to not using a search engine at $0/month.

> So, $10/month would be a steal compared to not using a search engine at $0/month.

But that's a fantasy dichotomy as long as free search engines exist, and there's plenty of them. If you could somehow change the world and wipe them all out in an instant, I guarantee that people would scramble to provide alternatives. We will never live in a world where your only option is $10/month or no search. (Free & open source search engines already exist and you could host one at home or on a cheap VPS; there are also P2P search engines)

Of course people would scramble, but it has nothing to do with how much value you can screw users out of using advertising, because even if ads were worth $0 revenue, search is so valuable you could just plain charge users. This is why there is essentially no relationship between the value of search and the ad revenue. The value of the ad revenue could be $0, and the value to the user would still be $100+/month.

And because the consumer surplus is literally an order of magnitude or two more than the subscription fee quoted, that is prima facie a case that a subscription search engine could have a marginal benefit of >$10 compared to the free ad-supported engine. It, or the subset of searches you opt to use it for, only needs to deliver a little more value to be worth it. Quite aside from the problem of Google Search being increasingly jammed full of ads, wasting your time, or any distortion of ranking, people just plain dislike and avoid ads (https://www.gwern.net/Ads).

The value proposition is in surfacing a result that you wouldn't already have from the free search engine. Your personal calculus will of course vary but lets do a basic business case with the following assumptions

Free search engines work 95% of the time for your employees searches

Kagi can get a result in half of the remaining 5% (this is definetly the biggest assumption and I haven't had enough experience with kagi to say if this is realistic)

Your employee does 1 search a day and 30 days in a month (so kagi gets you 0.75 more completed searches a month).

It takes an employee 15 minutes to search manually through documentation or come up with a solved algorithm from first principals when the search fails.

In that situation your employees time needs to be worth less than $53.33 dollars an hour for the $10 dollar plan not to break even.

So play with the numbers how you want to make up your own mind but it does seem reasonable to argue there's a market for it at that price. Personal use where missing a result could have no cost is ofcourse another question.

edit: 0.75 not 1.5 searches a month extra

Sure, I'm not contesting that there's a market for it. Going along that line of reasoning, there's also a market for a group of experts you can phone and get an answer from at $100 / hour, and so on. But let's not push the goalposts too much :) My personal calculus says it's not worth $10 for me (I don't rely on search much for my work).

Their pricing is closer to $30 a month if you are searching more than a handful of times per day.

At $10 a month you have to pay per search if you search over 20 times per day or something.

The price is crazy only because you're used to not seeing the price you're paying (ads). I spend $10 on things way less valuable very often. A good search engine is at least as much value as intellij to me, and my company pays 5 times more than that per month for intellij.

The problem is that ads price discriminate. Google may not get much money off of you, but there are other users that are very valuable (think of a manager in a billion dollar enterprise searching for a subscription product to buy). Would you be okay with it costing 0.05% of your income?

Of course it will be hard to compete with ads as business model if the alternative doesn't allow for price discrimination.

You'd be surprised. Google makes more than that from showing ads to an average US user. Same for Facebook iirc.

Looks interesting.

In the FAQ they mention potentially charging around $10/month.

Not sure if I'm being entitled or anything, but I was expecting something more like the original WhatsApp model of a few dollars a year.

Perhaps I'm under-estimating how computationally heavy search is.

> Perhaps I'm under-estimating how computationally heavy search is.

It's because they aren't rolling their own search, they pay Google and Bing to do the search for them (via the Google and Bing Search API's which are charged), mix in a few results from their own crawler, and then reorder the results.

So they will always have a higher cost base than both Bing and Google, because they are paying for 3 different search indexes (including their own), plus Bing and Google's margins on the API, plus their own infra costs.

(Now if this is a sustainable model or not is another question...)

Kagi has a "consumption" section where they show how much your searches cost to perform.

I've done 20 searches this month (I haven't switched it to default) and it says I have incurred $0.25 (between $0.01 and $0.02 per search) so it would seem that it's very expensive for them to provide results at the moment. It would absolutely be unsustainable at a few dollars a year given these consumption numbers.

I'm not sure how they are counting the searches, mine is indicating more than 100 searches per day and I certainly don't access their website that much (maybe search suggestions count too?). Their results and overall product are much better than Google but I won't pay 30 usd per month for that.

Every time you make a 'search action' it charges you, including changing categories on the site or changing the filters on the search.

Lets say you search for "Photo Of A Mother" * Then you click "images" * Then you click "Sort by Recent" * Then you click "Licence -> Public" * Then you click "Size -> Large" * Then you click "Size -> Extra Large" *

Every time I have put a star in the above is a time where you would be charged a search (so the above would be charged as 6 searches). It's the same thing with switching to news, applying filters, blocking a site or boosting a site etc - I've validated this on my Kagi account by clicking actions and seeing what it does to the billing, and just by using the search engine and using the lenses feature for example you can quickly rack up loads of searches.

Now let's say you search for an error while you program, visit the first site, it's not got what you want so you click the back button and then visit the second site, that's not got the right answer so you click back and visit the third site... That's currently counted as 3 different searches rather than 1 search. If you open them in different tabs it's counted as 1 search though.

And with all this then you are suddenly over 25% through your daily allowance on the $10 plan. Even choosing to 'block' a site in the search charges you to block it. They are talking about charging $0.015 per search if you go over-quota (which is something like 20 searches on the $10 plan), but as far as I can tell if you are using it moderately heavily you will blast part the 20 searches and could end up with an eye-watering bill.

I think the team are great, and when I was on discord they were really receptive, but I ended up giving up on Kagi after trying to use it as a daily driver as I figured they wouldn't be able to find a good enough monetisation strategy for my level of usage (after hearing discussions in the pricing channel). The product is good, it's just that they can't offer it at a price I can accept (and I suspect they can't offer it at the moment at a price that the market can accept, unless they can reach a deal on API licencing or roll their own search).

That pricing model is insane. I think it will remain a niche product only used by the wealthy unless they can get it down to somewhere like $5/month for unlimited searches.

Yeah, it's not that hard to get results as good as Google when you are paying Google and Bing to give you search results.

It IS hard to get results as good as Google at a price where people are willing to pay for it.

IMO I think the issue with Kagi is that their current architecture has made it easier for them to get great results (as they are just taking their results from Google + Bing and adding their own special sauce) but it makes it much more difficult in the long run to be cost-competitive.

Their nearest competitor (Neeva) does offer $5 a month search, has a free tier, and rolls their own engine, so I don't really see how Kagi's current model is sustainable unfortunately (but maybe they will be tremendously successful and I will look back at this post and feel silly!).

The premise is that for people willing to pay for a search engine, quality of results trumps everything.

I've been using Kagi for a while now too. I like their stance on privacy, but I don't really like that I have to create an account to use it.

https://kagi.com/privacy

Any reasonable justification for requiring sign-up/a user account?

Presumably because, as they move out of beta, they want people to start paying to use their search engine. Thats the only way they can afford to be ad-free.

I'm 100% happy to pay for an ad-free search engine that doesn't sell my data. I don't really like having all my search terms being linked to me via an account, but I suppose that's already happening anyway. I guess I picked the right day to stop looking at clown porn.

I guess how would you implement monetization without an account in some form ?

You could just allow a given IP address ? But that's just as trackable and has tons of downsides with using from other locations.

They could do the SaaS model and give you an API key or unique URL allowing up to 10000 searches per month which you could share with friends/your company

That's probably a worse business model, but it would be really interesting to hear what other monetization ideas Kagi considered or is considering.

Works at an enterprise level (e.g. exchange rate providers) but would be a bad fit for a B2C product.

Managing customer expectations on api key usage (esp. if that key is publicly visible e.g. URL parameter vs. HTTP header) is not worth it unless you have higher-priced products.

Also api keys would mean you might have to prevent re-selling, etc. Furthermore, they could still analyze api key usage to get the same historical data on you as if you logged in.

Would still link searches to your API key.

That's missing the point. If the same API key is used by 1000 people at a company, you don't have the expected 1:1 mapping of search terms to individuals with an account.

An individual account which isn't permitted to be shared is different.

You could sell tokens that get used up after each search or expire after certain amount of time since first use. Browser extension could store tokens and provide them to website as needed in random order. Tokens could be resold so no tracking by payment processor

To buy a token you still need payment details. How would this be different to just buying a subscription?

If you buy from reseller they have no way to correlate payment details to search queries. With subscription, however, service provider have your payment details and knows everything about your queries so it can build a profile on you.

yeah but they've been in "beta" for four years now.