That is one of the most compelling things I've ever heard about a search engine.
How well does kagi work for niche "reddit queries" like "best waterproof midi synthesizers reddit"?
Try it! After you'll be on a site that is just generated content, click backwards in the browser and remove the site from all your future kagi results. That's how search should work (and it should _not_ provide you a top 30 of random AI content).
My favourite feature is the customised down ranking I can do. No Pinterest etc for me.
Personally, if I could own Kagi stock, I would.
It's not just you.
If there are ads on a page, it means the advertisers own the author. He is not free to write what he wants, he is free to write only that which the almighty advertisers will tolerate being associated with. He will not write things which bite the hands that feed him.
I'm convinced writers like that will never write anything truly genuine. Chances are if you see ads anywhere you're reading self-serving generic clickbait content meant to attract attention and drive up ad impressions. It's not real, it's just "content", a generic square around which ads congregate like parasites.
Probably least a third of my queries are preceded by an !r now. A third of the rest are now question mark queries that activates their AI fast answer. It's like the google info box on steroids since it can answer any query and it works with lenses to restrict the fast answer to specific domains.
I originally signed up purely out of spite for the SEO scam that is Pinterest (Kagi lets you blacklist domains), but have since been repeatedly pleased with other rankings.
I love that MDN tops the list for DOM ish searches and w3schools is not even in the results.
Using Kagi often makes me forget how awful the Internet can be.
I’ve started using their quick answers to sort through the crud. In most cases, it catches and filters out the obviously-bought Reddit recommendations, surfacing bloggers and niche industry publications that did their own lab work.
One of them being vietnamcoracle.com, which is, without a doubt the best travel website for Vietnam (if you like to really go deeper.) None of the content is sponsored, just relevant ads (here you can rent a bike, book a tour etc) and they are discrete in general.
Between him researching new destinations and writing / updating/ replying to comments he just doesn't have a lot of time for the classic "money making job + sideproject".
I just imagine (hope) he's not the only website like that out there.
I don't currently use kagi but would prefer if there was an option to filter by ad type (everyone hates pop ups) and amount.
Anyway just my 2c.
With wolfram and the llm + more users I'm hopeful that in the longer term the price will go down and/or stay at 10$ over the longer term.
They aren’t eliminating it, just downranking it. Given Kagi’s search quality, ad and tracking density seems to negatively correlate with site quality.
So on Google you miss out on things that might otherwise be hidden gems because Google wouldn't rank it as highly.
There are tradeoffs on both sides.
That said, I am still quite opposed to js that tracks users across the internet for advertisement purposes. I do use ad blockers despite the fact they block a lot of less-harmful tracking by default, which I don't love, but it's too much work to differentiate between. (At least adblock users are the minority of internet users in general, so hopefully Analytics users still get enough data to be helpful for their purposes.)
I'd much rather get information from an individual or small group who is intrinsically motivated and is not just looking for the lowest bar of quality that won't make folks immediately bounce.
I also want to note that just putting Google Analytics on there doesn't kill the site's ranking. There are some sites that are just infested, and those are what get dramatically downranked.
I just tried your search, without even adding “Reddit” on the end. At the very top was a “Discussions” section, which had a tile for Reddit. The first result was also Reddit, with a few discussions nested under it. The the gearspace forum, followed by YouTube, then it bunches up a bunch of those 10 ten lists that pollute Google all into a section that is easy to use or skip, then sweetwater music, and then funny enough, your comment here on HN. It keeps going, but yeah… Reddit isn’t deranked to the point of not being used, Kagi sees the value in discussion forums when looking for the “best” something.
Do they filter out reddit or just small ma and pa sites with adsense?
best waterproof midi synthesizers reddit https://kagi.com/search?q=best+waterproof+midi+synthesizers+...
best waterproof midi synthesizers https://kagi.com/search?q=best+waterproof+midi+synthesizers&...
Feature request: block a domain?
"saab 900 power steering rack -reddit"
still returns a ton of reddit results.
https://help.kagi.com/kagi/features/website-info-personalize...
You can block specific domains, as well as make them rank lower. And if you ever want to ignore those rules, you can easily do that too!
Not affiliated with Kagi in any way, just a very happy user.
This is interesting. I guess we will soon read a longwinded post about it from the first person perspective.
"Kagi" goes first to get under his skin just a little bit.
Kagi on the other hand feels like an ideal subscription service. It feels like what Google search wanted to become.
[1] To force me not to leave open tabs I don't really care about, my default browser is FireFox Focus, which is strictly "incognito" mode, which means I'd have to manually log in to Kagi every time I did a search.
Even better though, now that Add Ons are opened up on Firefox for Android you can install Cookie AutoDelete. Then you just use regular tabs and can keep your Kagi cookies (and any other sites you regularly log into) but have it nuke every other site.
There is also a Kagi Add On. It didn't do pretty much anything when I first installed it last year but it might work these days.
Oh, and finally, there is the Kagi Session Link you can use that embeds a token in the URL so that you can more easily use it for something like a search provider in Incognito/Private tabs.
Oh that was you, thank you!
I don't understand your answer. Isn't that a good thing? Why are you missing trash on DDG?
I strongly recommend the Tab Wrangler extension. I set it to close any tabs that have not been visited in the last 6 hours. Of all the methods I've tried to deal with too many tabs, this has been the most effective.
FF Focus is my go-to browser, so most things I look up or do on the web vanish.
I have a second browser (Vanadium) with ~20 tabs open in different groups which I keep open perpetually. Things like my bank's site, my healthcare portal, other things I want to be able to stay logged into. Then there's one group for different articles I'm reading; these I close once I've exhausted the conversation and opened all the links I'm interested in.
(My desktop computer is a different matter entirely.)
Made me think and have to do some digging for the 'Kagi' out of Berkeley founded in the 90s where you could register your shareware purchases in the days before PayPal. Which seems to have been largely scrubbed from history outside of some WayBack snapshots.
https://www.macrumors.com/2016/08/01/kagi-shuts-down/
https://tidbits.com/2016/08/04/kagi-shuts-down-after-falling...
https://help.kagi.com/kagi/faq/faq.html#are-you-affiliated-w...
https://www.wolframalpha.com/input?i=2016-04-04+to+2019-01-3... = 2 years 9 months 27 days
https://www.wolframalpha.com/input?i=2019-01-31+to+2016-04-0... = 2 years 9 months 26 days
The duration should be identical but is off by one day.
One day a space mission will fail due to this bug.
Reported several times, they never cared.
both show an identical duration of 2 years 9 months 26 days (edit: despite reporting 1032 and 1031 days respectively).
Months are lunar. Days are solar.
Jan 2024 - Nov 2023 example (both cases show "2 months 1 day"):
https://www.wolframalpha.com/input?i=2024-01-31+to+2023-11-2...
https://www.wolframalpha.com/input?i=2024-01-30+to+2023-11-2...
Aug 2024 - Jun 2024 example (same bug):
https://www.wolframalpha.com/input?i=2024-08-31+to+2024-06-2...
https://www.wolframalpha.com/input?i=2024-08-30+to+2024-06-2...
The same thing doesn't happen from Jul 2024 - May 2024 (results vary by 1 day):
https://www.wolframalpha.com/input?i=2024-07-31+to+2024-05-2...
https://www.wolframalpha.com/input?i=2024-07-30+to+2024-05-2...
Date time arithmetic is weird ;)
https://www.wolframalpha.com/input?i=2019-01-31+to+2015-10-2... = 3 years 3 months 10 days
https://www.wolframalpha.com/input?i=2015-10-21+to+2019-01-3... = 3 years 3 months 10 days
I was a bit reluctant but I don’t regret doing it.
Really happy to see Wolfram added to it!
Recently I've been adding ? to a number of queries, to play with its knowledge graph and answering capabilities. It's been remarkably useful at surfacing information.
A search engine should show web results based on my query. That's it. Some summaries and highlights are useful, but show them in the sidebar, and make them optional.
If I need a calculator, I have plenty to choose from, including the Wolphram Alpha site.
If I need an answer to a question, LLMs do a good job at that.
Please don't make the common mistake of making your search engine "useful". My average session on your site should last seconds, which is the time it takes me to see the results, and click on the most relevant one. If you achieve that, you're doing a great job.
If I am paying for a product and a 90% majority of its features I don't care about and only care about 10% of it working really well I'm going to feel like I'm overpaying since I only use and care about 10% of the product, but pay for 100% of it.
And if I were a paying customer I'd keep getting more and more weary if I see the focus of the product keeps being this 90% I don't care about.
Automattic has almost 2000 employees.[0] I believe that number even automatically updates as folks come and go, but I could be wrong.
(I worked at Wolfram Research around a decade ago, when it was ~500 employees)
Sites have to make exceptions for Google, but likely wouldn’t care enough to allow other search engines in.
https://help.kagi.com/kagi/search-details/search-sources.htm...
I've crawled small web sites (even hosted on major providers) to archive them and never hit a wall.
[1] https://help.kagi.com/kagi/search-details/search-sources.htm...
> Kagi Search includes anonymized requests to traditional search indexes including Brave, as well our own non-commercial index (Teclis), news index (TinyGem), and an AI for instant answers. Teclis and TinyGem are a result of our crawl through millions of domains, focusing primarily on non-commercial, high-quality content.
[Edit] I guess it is possible that Google or Bing are hiding under "traditional search indexes including Brave".
If they announce that my monthly payments are going up, and I suspect it's because they're paying for partnerships with services I don't view as core to the product value that brought me in, I would feel cheated. And if they add those integrations and don't charge more, I'd have to wonder about that too: are they trading something for the integrations, or are they charging me too much?
However, as an early user, I've continued to support them and I think you've articulated really well why they're doing what they are doing -- and what made me not drop my subscription when I saw them integrate "GenAI" into Kagi.
Neat!
1: https://help.kagi.com/kagi/support-and-community/open-source...
- Kagi will rely on Wolfram's okay-ish stack to immediately add features to their engine
- Kagi will, over time, 1-by-1 remove the reliance on the (generally slow and expensive) Wolfram API with built-in features
- Once the meat of the features have been rewritten internally by Kagi in more performant and self-controlled ways, remove the dependence on the slow and expensive Wolfram API
Not saying that's a bad thing, but it's quite likely this is how it plays out. We've seen it before.
At that discount rate, a lifetime membership at $1,680 would be rational for Kagi ($600 individual). And at that price, I’d pay! Hell, do $3k flat (5% cost of capital; ~$1k individual) and throw in an annual in-person event and I’d call it a great deal. For Kagi, moreover, it would be permanent capital.
(I would also love for Kagi to launch a K-12 education product, possibly marketed at first to PTAs.)
[1] https://kagi.com/settings?p=billing_plan&plan=family&period=...
I strongly believe part of the reason Plex has gone down the shitter is because they sold too many lifetime memberships and expenditure has outstripped recurring revenue.
(https://kagifeedback.org/assets/files/2024-03-05/1709652268-...)
Does Kagi hang indefinitely for text searches for y'all? I experience this behavior about 1/4 searches. No error; just no page load. I am about to cancel my sub and switch back to Google.
Make sure to include example searches when this happens.
Hopefully the kagi team will see this though and chime in, or you should report the bug
Silly example: I just tried "average distance to moon/Eiffel Tower height" in Kagi and got "= about 1.167 million" (powered by WA).
I find that W|A works fine until it fails to parse my query and then I have no idea how to fix it, because the syntax is unclear.
https://wolfram.com/mathematica/new-in-8/free-form-linguisti...
3:38 - 0:10 = 8 hours and 32 minutes
I'm guessing it was trying to convert it to times of day instead of 3 minutes 38 seconds.
Interestingly if I click the quick answer button it gives 3:48 as a result.
Google - doesn't even try.
WA directly also had to nudge it to use it as a unit instead of a time of day.
Still some work to do here it seems!
But your query was ambiguous as written.
8 hours and 32 minutes is how long it takes to go from 3:38 pm to 10 after midnight? Perhaps the am/pm thing is also underspecified, so it chooses the shorter time delta?
Random aside: this blog is the first time I've seen a monospace font used for prose that doesn't look absolutely terrible; in fact, it's very readable for me. Inspecting with Firefox, I see the font is Menlo--and looking it up, it even seems like it's a web-safe font. Pretty interesting.
I have 10-20 keywords set up in Firefox for example. It works in chrome too.
But maybe I'm missing some use case for kagi bangs...
[1]: https://help.kagi.com/kagi/getting-started/setting-default.h...
I love the idea of Wolfram throwing his brainpower and associated ego behind this problem.
Bing couldn't answer "MangoldtLambda[11]" either, and it also couldn't answer "differentiate x^2 with respect to x", and didn't show a graph for "norway gdp".
🞬 blog.kagi.com (blocked ghs.googlehosted.com)
I have unlimited searches with Kagi, but I have still gotten into a habit of obsessively taking notes. I personally use Org Roam, but I think anything with hyperlinking and allows you to take notes non-hierarchically is important. This kind of information is worth searching the web once, taking note of, and then using your notes later to access.
While yes, realistically this is just a lonesome longwinded approach to bookmarks, I have personally never found myself going back to browser bookmarks. I actually refer to my notes all the time though.
> 16:28 - 14:50 = 22 hours and 22 minutes
although, Google never could not too.
> 14:50 - 16:28
Or
> 14:50 to 16:28
It makes sense when you think of it.
I’ve been using this for years, and it works wonders for datetimes too.
These prompts don't show on kagi? Seems like including them would be an avenue for improvement.
SearX has some advantages. It’s free, except for the compute and labor costs. It runs on our infrastructure, so SSO is easy and we have complete visibility on what’s leaving our perimeter.
Ignoring nuanced search quality evaluation, SearX main disadvantages are (1) it’s only available on our network, so if you set it as the default search that feature breaks if you’re not at the office or on VPN and (2) it’s slow. Users get a blank page for a meaningful moment before results show up.
Kagi is snappy AF and adds solid AI features. Their Universal Summarizer (which isn’t actually LLM powered) works well and the fastGPT is the new standard for open ended information retrieval IMHO. (I have a Perplexity subscription I don’t bother to use)
What I want from Kagi is (1) a single payer team plan that gets all the features, (2) more information & contractual guarantees about privacy. I argue that Google is a terrible default, but the brand is trusted while Kagi is still mostly unknown.
Now, the real benefit of SearX would be integrated custom search providers for in-house data sources. We haven’t done that, and I’m not confident we have the resources to do it well. This is also something I’d love to get from Kagi, but it’s not well-aligned with their current roadmap. I’m content to use them to search external data and pursue other solutions for internal stuff.
fantastic news, really glad to hear about this
I find that people often don't fully know how to search and blame Google for not having the feature. Google even implements things in the browser that people can take advantage.
Now it says unlimited, but I'm not sure "you and a swarm of a thousand digital clones" counts as fair use...
It obviously does not. You’re a commercial client.
>swarm of autonomous agents.
So....Google is right?