[00:03.120 --> 00:08.400] It's a very big honor to be a part of this 30-year tradition, so thank you for that. [00:09.760 --> 00:13.180] Before I begin, how many people have heard about Kagi? [00:13.880 --> 00:16.320] Just... oh, oh, okay. [00:17.860 --> 00:24.080] I was hoping to convert some of you to Kagi, but it seems like... to pay for the trip. [00:25.600 --> 00:26.280] That's great. [00:26.900 --> 00:36.900] Well, yeah, this will be sort of a story, how did we do what we... how did we get started with this, and how do we do some of the things we do? [00:37.140 --> 00:40.120] Just trying to approach it from a hacker's mindset. [00:40.820 --> 00:47.420] And to do that, I wanted to establish some credentials first with this community, since this is my first time here. [00:48.040 --> 00:49.340] This is how I got started. [00:49.460 --> 00:52.520] This is the literally first thing I saw on a computer screen. [00:53.440 --> 00:56.220] Raise your hand if you share a similar experience. [00:56.760 --> 00:58.320] Okay, a couple of hands there. [00:59.080 --> 01:01.280] Yeah, this... this is 83. [01:01.880 --> 01:03.020] I was literally mind blown. [01:03.400 --> 01:06.180] And from that moment on, I knew I'm gonna be in computers. [01:06.800 --> 01:13.920] And then some 30 years later, I... I had an honor to have a dinner with David Braben, who's one of the guys signed here. [01:14.560 --> 01:21.940] And he continued to... to blow my mind by discussing how to drive a comet into Mars to terraform it. [01:23.160 --> 01:24.360] So fun stuff. [01:25.460 --> 01:31.100] Ten years after that, I was in college, and I got drawn into the world of muds. [01:31.860 --> 01:34.020] Mud is a multi-user dungeon. [01:34.260 --> 01:36.260] It's an Internet text adventure game. [01:36.900 --> 01:39.620] And this is the reason I dropped out of college. [01:40.000 --> 01:44.260] I played insane amounts of muds, and then I created... [01:44.880 --> 01:50.340] One... the screenshot on the right is... is actual my own mud, which is still up and running. [01:51.680 --> 01:52.280] And... [01:52.280 --> 01:54.140] Very interesting quote here. [01:54.460 --> 01:57.500] Text is the highest bandwidth medium we have to the imagination. [01:57.980 --> 01:59.260] I think resonates... [02:00.180 --> 02:04.260] Pretty well across all things that I do, especially nowadays. [02:05.100 --> 02:05.700] But... [02:05.700 --> 02:06.700] Um... [02:06.700 --> 02:07.680] I thought it... [02:07.680 --> 02:08.460] It... [02:08.460 --> 02:09.400] It is pretty remarkable. [02:09.900 --> 02:10.300] And... [02:10.300 --> 02:10.920] This is... [02:10.920 --> 02:11.600] This is... [02:11.600 --> 02:15.480] Quote by Richard Bartle, who's one of the co-founders or the co-creators of mud. [02:15.820 --> 02:22.680] And he talks about reading text on the screen and that there is nothing else as powerful in terms of activating your imagination. [02:23.880 --> 02:31.600] And I also find it interesting that we sort of went full circle from text to 2D to 3D to AR and VR. [02:31.880 --> 02:32.360] And now... [02:32.360 --> 02:33.760] Thanks to AI. [02:34.280 --> 02:34.520] Uh... [02:34.520 --> 02:34.940] We are... [02:34.940 --> 02:37.040] Kind of back to text again. [02:37.300 --> 02:39.960] So I find it just interesting that this is happening. [02:41.120 --> 02:41.640] Um... [02:41.640 --> 02:43.620] And I also happen to be... [02:43.620 --> 02:44.100] Uh... [02:44.100 --> 02:46.920] Part of the generation that was promised jetpacks and... [02:46.920 --> 02:47.640] And... [02:47.640 --> 02:48.020] Find cars. [02:48.880 --> 02:49.400] And... [02:50.020 --> 02:50.540] Uh... [02:50.540 --> 02:54.220] And we have this little itch, like, why is that not happening? [02:54.360 --> 02:56.380] And so we are doing these crazy things, I guess. [02:58.160 --> 02:58.680] Um... [02:58.680 --> 03:01.560] And, you know, I was always... [03:02.340 --> 03:03.020] Uh... [03:03.020 --> 03:10.400] Annoyed by hundreds of thousands of smart engineers working on things like optimizing ad clicks and whatnot. [03:10.780 --> 03:11.380] And not... [03:11.380 --> 03:12.720] Not delivering on this promise. [03:14.260 --> 03:14.940] Um... [03:14.940 --> 03:16.820] So today we are here to talk... [03:16.820 --> 03:18.580] About search mostly. [03:19.160 --> 03:19.840] And... [03:19.840 --> 03:21.100] The fact that it's broken. [03:22.080 --> 03:22.760] And... [03:22.760 --> 03:25.860] Especially recently, if you are in this field and... [03:25.860 --> 03:30.580] It's almost like every week there is a publication that's talking about how Google is getting worse. [03:31.880 --> 03:32.360] Um... [03:32.360 --> 03:33.400] And that it's dying. [03:33.740 --> 03:34.500] And it's useless. [03:34.780 --> 03:35.420] And things like that. [03:35.760 --> 03:36.180] And... [03:37.540 --> 03:38.020] Um... [03:38.020 --> 03:39.560] There is a pretty good reason for that. [03:39.680 --> 03:43.860] And the founders of Google knew about that all the way back in 1998. [03:43.860 --> 03:46.720] And these are the quotes from... [03:46.720 --> 03:49.620] From the seminal white paper that basically defined Google. [03:50.400 --> 03:50.800] That... [03:50.800 --> 03:52.040] Uh... [03:52.040 --> 03:52.120] Uh... [03:52.120 --> 03:54.960] Green and Page wrote at Stanford in 1998. [03:55.340 --> 03:56.700] So they clearly... [03:56.700 --> 04:05.380] Here recognized that adding advertising into a search engine will lead to bad outcomes. [04:05.380 --> 04:08.740] You will have worse search results. [04:09.260 --> 04:09.780] Uh... [04:09.780 --> 04:10.160] The... [04:10.160 --> 04:11.900] There will be a conflict of interest. [04:12.820 --> 04:13.400] Uh... [04:13.400 --> 04:16.920] There's actual incentive to provide poor quality search results. [04:17.100 --> 04:20.020] Because you want the user to keep searching and keep clicking. [04:21.340 --> 04:21.630] Um... [04:21.920 --> 04:24.500] And this is normally where my presentation would end. [04:25.000 --> 04:26.040] Or should end. [04:26.560 --> 04:26.960] Uh... [04:26.960 --> 04:29.320] Because these guys figured it out at the beginning, right? [04:30.240 --> 04:30.820] Um... [04:30.820 --> 04:32.120] But that's not how it played out. [04:34.780 --> 04:35.360] Uh... [04:35.360 --> 04:36.020] It was already pretty big. [04:36.200 --> 04:37.840] And they adopted the... [04:37.840 --> 04:39.400] The ad-based business model. [04:39.740 --> 04:41.640] And roughly ten years... [04:41.640 --> 04:42.640] From that moment... [04:43.400 --> 04:44.080] Uh... [04:44.080 --> 04:45.640] The ads were just everywhere. [04:46.200 --> 04:46.680] Um... [04:46.680 --> 04:48.360] And it was not more a search engine. [04:48.620 --> 04:52.540] And the original mission to organize the world's information become... [04:53.800 --> 04:54.480] Um... [04:54.480 --> 04:56.080] Organize the world's ads, basically. [04:59.080 --> 04:59.760] Um... [04:59.760 --> 05:03.360] What is also revealing is that... [05:04.760 --> 05:06.800] You know, after many years... [05:06.800 --> 05:07.360] Um... [05:07.980 --> 05:08.440] Of... [05:08.440 --> 05:08.840] Of... [05:08.840 --> 05:09.220] Of... [05:09.220 --> 05:10.440] Of such kind of practice. [05:11.220 --> 05:13.560] And shareholder pressure and things like that. [05:13.740 --> 05:15.580] It was revealed recently in the DOJ. [05:16.440 --> 05:17.040] Uh... [05:17.040 --> 05:19.080] Anti-trust case against Google. [05:19.240 --> 05:20.140] Some internal documents. [05:20.520 --> 05:24.740] When they replaced the head of search with the guy who basically ran ads. [05:25.060 --> 05:25.200] So... [05:25.200 --> 05:27.520] So head of ads became head of search. [05:27.520 --> 05:29.480] And the reason they did it... [05:30.340 --> 05:30.900] Was... [05:30.900 --> 05:33.380] They wanted to increase the advertising revenue. [05:33.680 --> 05:34.560] And what the guy did... [05:35.060 --> 05:35.520] Is basically... [05:36.380 --> 05:37.520] Ordered search to... [05:38.100 --> 05:38.460] Become worse. [05:38.780 --> 05:40.980] Because their internal testing... [05:40.980 --> 05:41.880] Uh... [05:41.880 --> 05:43.360] Has shown that people... [05:43.360 --> 05:44.740] Largely didn't care. [05:44.920 --> 05:45.720] And there was no other... [05:46.300 --> 05:46.760] Uh... [05:46.760 --> 05:47.880] Replacement on the market. [05:48.120 --> 05:48.680] So... [05:48.680 --> 05:50.440] So they just made search worse. [05:50.600 --> 05:51.520] Which made people... [05:52.100 --> 05:52.620] Stay... [05:52.620 --> 05:53.380] And... [05:53.380 --> 05:53.780] Uh... [05:53.780 --> 05:55.880] Spend more time on Google searching for stuff. [05:56.140 --> 05:58.400] And every interaction with Google is obviously... [05:59.140 --> 05:59.700] Uh... [05:59.700 --> 06:01.180] An opportunity to have... [06:01.180 --> 06:02.620] More ads shown to them. [06:05.180 --> 06:06.020] My own... [06:06.300 --> 06:06.840] Uh... [06:06.840 --> 06:07.460] Journey... [06:07.460 --> 06:08.160] With... [06:08.160 --> 06:09.240] Sort of waking... [06:10.060 --> 06:10.620] Um... [06:10.620 --> 06:11.640] From the matrix... [06:11.640 --> 06:12.460] Becan... [06:12.460 --> 06:13.300] In 2018. [06:13.620 --> 06:15.900] And this was the exact search query... [06:15.900 --> 06:18.480] That I typed in one day in 2018. [06:18.500 --> 06:20.160] And I woke up from the matrix. [06:21.620 --> 06:22.180] And... [06:22.180 --> 06:22.980] Uh... [06:22.980 --> 06:23.720] What's... [06:23.720 --> 06:25.320] This is... [06:25.320 --> 06:25.560] You know... [06:25.560 --> 06:26.980] A typical Google query. [06:27.220 --> 06:28.280] I was searching for GitLab. [06:28.460 --> 06:29.520] Which is a GitHub alternative. [06:30.000 --> 06:31.240] And Google presented me... [06:31.240 --> 06:33.140] With a sponsor link... [06:33.140 --> 06:34.000] For the site. [06:34.340 --> 06:34.980] And then... [06:34.980 --> 06:36.120] Exactly the same... [06:36.120 --> 06:37.060] Organic link... [06:37.060 --> 06:37.640] For the site. [06:38.360 --> 06:40.200] And for the first time... [06:40.200 --> 06:41.360] After using Google... [06:41.360 --> 06:43.220] For probably 20 years... [06:43.220 --> 06:44.080] I realized... [06:44.720 --> 06:46.120] That something's wrong with this. [06:46.280 --> 06:47.880] And the thing that was wrong... [06:47.880 --> 06:48.760] Is that it... [06:48.760 --> 06:49.220] Assaulted... [06:49.220 --> 06:50.300] I realized it... [06:50.300 --> 06:51.560] That... [06:52.800 --> 06:53.500] Aslee... [06:53.500 --> 06:54.320] I realized... [06:55.460 --> 06:56.160] Like... [06:56.160 --> 06:57.740] Nobody would do that... [06:57.740 --> 06:58.260] This... [06:58.260 --> 07:00.360] If my best interest was... [07:00.360 --> 07:01.300] Their goal... [07:01.300 --> 07:01.400] Right? [07:02.100 --> 07:03.540] It was pretty profound. [07:03.760 --> 07:05.040] It took me 20 years... [07:05.040 --> 07:06.100] Um... [07:06.100 --> 07:06.580] But... [07:07.300 --> 07:07.660] Um... [07:07.660 --> 07:08.380] But after that... [07:08.380 --> 07:11.360] I couldn't look at Google results the same way anymore. [07:13.340 --> 07:22.600] The other thing that I realized also was that my kid was just starting the elementary school in California. [07:22.600 --> 07:24.240] They all get Chromebooks. [07:24.600 --> 07:36.180] And I was pretty terrified of her being profiled for the next 12 years and targeted with ads, attempting to change her behavior and things like that. [07:37.220 --> 07:40.320] So these sort of things happen at the same time. [07:40.520 --> 07:48.620] And the third one, I read this quote from Ian Bremmer as a political scientist just about that time. [07:48.780 --> 07:57.020] And he said, the idea that we get our information as citizens through algorithms determined by the world's largest advertising companies by definition of dystopia. [07:57.520 --> 08:06.800] So all these came together and it became apparent that something's wrong. [08:07.260 --> 08:17.700] And not only that something's wrong, but something's fundamentally wrong with one of the most important tools that I and many of us in society use every single day. [08:18.120 --> 08:30.240] We are basically depending on a search engine that is run by one of the largest private companies in the world whose incentives are not aligned with our own. [08:31.260 --> 08:37.240] And it has profound impact on us as users, as people, but also on a society. [08:40.220 --> 08:43.680] And so this [08:46.780 --> 08:51.980] is a quote that came out last night as I was having a couple of beers with my buddy here. [08:52.700 --> 08:54.540] And I thought I could reuse it. [08:54.800 --> 09:03.120] And because it's an interesting quote, and Robert, Hailein may or may not have said this exact thing. [09:03.760 --> 09:07.720] But there's this question, why do this? [09:08.520 --> 09:13.020] Like, why would anyone be crazy enough to attempt to create a paid search engine? [09:14.760 --> 09:16.120] Well, that's the reason, right? [09:16.360 --> 09:18.120] Somebody has to. [09:18.800 --> 09:19.720] And so we did. [09:21.820 --> 09:26.600] The easy thing was trying to figure out, so how do you change search? [09:26.760 --> 09:28.840] How do you do this? [09:31.100 --> 09:34.260] And the formula is pretty simple, right? [09:34.360 --> 09:44.540] There's a thing that the user wants, and you should be giving that thing to the user, avoiding all the crap that we see in search every day. [09:45.220 --> 09:46.940] So the formula is pretty simple. [09:48.820 --> 09:51.200] Treat search as a user-centric thing. [09:52.020 --> 09:56.200] So put the user in the driving seat, optimize the product for them. [09:56.360 --> 09:59.880] Every other almost product in the world is optimized for the user. [10:00.020 --> 10:00.720] Search was not. [10:01.940 --> 10:06.520] Remove intermediaries between the user and the provider. [10:06.660 --> 10:10.080] And this is especially important in the context of information consumption. [10:10.920 --> 10:13.760] And those intermediaries are obviously advertisers. [10:14.200 --> 10:27.620] So you do not want to have an intermediary when you're searching for information for yourself, for your family, you know, for scientific, educational, health, whatever reasons you do not want to have intermediaries. [10:27.780 --> 10:30.160] You want to have direct communication. [10:30.640 --> 10:33.320] You need to build a business model that aligns incentives. [10:33.580 --> 10:35.600] That business model is pretty simple. [10:35.600 --> 10:39.380] It's called paid or paid subscription model. [10:39.600 --> 10:41.420] It's used by many, many other products. [10:41.820 --> 10:49.380] And what it does is directly aligns incentives because when you do that, the user becomes the customer. [10:49.520 --> 10:51.460] And if your search sucks, they go away. [10:51.800 --> 10:54.260] So you are directly aligned with them. [10:54.860 --> 11:02.300] And a nice thing you get for free with such business model, you get privacy basically being a byproduct of the business model. [11:02.420 --> 11:10.820] You don't have to specially optimized for it because at that stage, user data becomes a liability and you want to have as little as possible of it. [11:10.940 --> 11:13.160] You have no value in having user data. [11:13.340 --> 11:16.580] The only thing you want from the user is $5 a month or $10 a month. [11:17.180 --> 11:23.380] So removes or makes your product privacy respecting almost by default. [11:25.100 --> 11:28.880] And so we started with a fairly small team. [11:29.040 --> 11:30.260] We bootstrapped this whole thing. [11:33.680 --> 11:36.440] And this is a picture just from the COVID years. [11:37.300 --> 11:42.760] Four of us from the early team just taken near here in Met actually. [11:43.360 --> 11:52.280] And yeah, very small sort of let's leverage things that we can leverage and let's try to build this thing. [11:53.800 --> 12:06.080] And I want to show you a couple of things that once you start thinking about putting user in the center of everything, what does that user experience start to look like? [12:06.520 --> 12:11.520] So I want to show you a couple of Kagi features and we will later talk about how we do them behind the scenes. [12:13.200 --> 12:16.640] One of the most used features is Kagi. [12:17.040 --> 12:20.340] And this is, by the way, how Kagi search results look like. [12:20.580 --> 12:23.740] We like this black and white newspaper theme. [12:25.560 --> 12:34.640] So one of the basic features and one of the most popular ones is that you can basically block or raise any result in your search results. [12:34.800 --> 12:39.060] And it's once you do that, for example, here I'm blocking medium, it's gone forever. [12:41.640 --> 12:42.920] And why is this powerful? [12:43.120 --> 12:46.880] Well, it gives you the user, the agency, to moderate your search read. [12:46.920 --> 12:48.440] We don't have to do that for you. [12:48.640 --> 12:58.380] And as a matter of fact, in Kagi, we're trying to do as little as possible in terms of intervening with results and giving users the tools and agency to do that themselves. [12:58.620 --> 13:05.180] So there's a particular site you don't like or a group of sites, a couple of clicks, and they're gone forever. [13:07.000 --> 13:18.660] We actually even publish a neat public page where you can see the most blocked and most promoted sites in Kagi. [13:18.840 --> 13:22.080] So there's a bunch of Pinterest and some political sites. [13:22.700 --> 13:23.920] It's pretty interesting. [13:24.160 --> 13:28.320] And as a new user, you can quickly sort of add them to your own block list. [13:30.520 --> 13:37.780] Another thing we do is we have this feature called Lenses, which basically allow you to get a subset of results from a web. [13:39.000 --> 13:42.440] So in this case, it was small web, which are personal blogs. [13:42.820 --> 13:45.040] These are academic results for cloud code. [13:47.540 --> 13:49.380] And you can create your own lenses. [13:50.000 --> 13:55.220] So we have a bunch of pre-built ones, the forums, the forums, programming. [13:56.140 --> 13:58.640] So it just searches the subset of the web. [13:58.940 --> 14:02.240] And it's very, very useful when you want to find something specific. [14:03.200 --> 14:05.120] I'm showing a couple of examples here. [14:13.720 --> 14:16.840] PDFs, Petiverse, and I think one more cyber security. [14:17.400 --> 14:18.740] Interesting for this crowd. [14:20.980 --> 14:22.420] So that's Lenses. [14:23.820 --> 14:25.520] Let me show you another one. [14:26.780 --> 14:31.920] We, for example, show you when we believe the content is behind a paywall. [14:32.600 --> 14:40.180] And we surface direct links to comments on Hacker News and Reddit for some results that are being discussed on these things. [14:40.420 --> 14:44.040] These are just small things that you don't see in any other search engine. [14:44.320 --> 14:45.600] But why not? [14:45.840 --> 14:46.860] People love them. [14:48.940 --> 14:51.600] We also group listicles together. [14:52.300 --> 14:59.500] So the listicles are those, you know, top five best something something 2025 blog posts. [14:59.680 --> 15:01.800] And the algorithm for this is like super simple. [15:01.920 --> 15:03.320] We are not doing anything fancy. [15:03.600 --> 15:11.960] Like the algorithm says, if, you know, top five or top numbers, something best, something year, group them together. [15:12.300 --> 15:16.020] Like it's nothing fancy, pretty hacking way, but it works like charm. [15:16.580 --> 15:21.460] It allows you to skip all this, you know, if you're searching for something like this. [15:23.960 --> 15:27.020] There are other areas for innovation. [15:27.500 --> 15:29.780] Once you start thinking about what the user wants. [15:29.980 --> 15:31.500] So here's a video search. [15:31.680 --> 15:33.820] You can block channels. [15:34.200 --> 15:37.820] So you can block domains in web search, but video search. [15:37.960 --> 15:39.960] You can, for example, block channels. [15:40.160 --> 15:42.300] You can also search different providers. [15:47.780 --> 16:01.180] We even built a feature where we would hide the... you can switch the clickbait thumbnails to the non-clickbait ones when you're looking for video results. [16:03.080 --> 16:04.220] Because why not? [16:07.720 --> 16:18.180] We have a feature where you can create basically regular expressions to change the URL of results to the different URL. [16:18.940 --> 16:24.800] Here example, redirecting anything ready.com to old ready.com, which is what I use. [16:24.920 --> 16:26.820] But you can do that for programming documentation. [16:26.820 --> 16:29.780] If you want to always go to the latest version or something like that. [16:32.160 --> 16:33.800] We have this nifty thing. [16:34.000 --> 16:39.420] If you add a question mark at the end of the search query, we show you an AI result. [16:39.620 --> 16:46.040] So all AI in Kage is on demand only and invoked by the user. [16:46.160 --> 16:49.560] So this is a quick way to just get an AI answer to a question. [16:50.860 --> 17:00.760] And we do have our own full assistant, which is basically shown here as the interface to that assistant. [17:00.760 --> 17:06.080] We have access to all the world's top LLMs and they're all grounded in Kage search. [17:06.260 --> 17:07.320] We're pretty proud of this product. [17:10.060 --> 17:19.460] So I want to also show a few things how we do this behind the scenes because although a lot of work, none of this is rocket science. [17:19.900 --> 17:28.040] And you know, with a relatively small effort, a small team, we have been able to do all this and continue to do all this. [17:28.880 --> 17:38.200] So one of the things that we started doing from the beginning is removing SEO garbage or low quality sites. [17:38.380 --> 17:41.120] And it's a pretty sort of low tech effort. [17:41.540 --> 17:44.100] We have a crawler that goes around the web. [17:44.420 --> 17:51.160] It uses a headless browser with uBlockOrigin loaded. [17:51.500 --> 17:56.680] And then we read how many ads and trackers uBlockOrigin blocks on that site. [17:56.700 --> 17:59.580] And that's our signal how crappy the site is. [18:00.040 --> 18:05.580] So the more ads and trackers that are on the site, the crappier it is. [18:05.780 --> 18:07.420] And this actually correlates... [18:11.220 --> 18:16.500] This actually correlates really nicely with the quality of the content in any testing that we did. [18:17.700 --> 18:22.280] And if this is above certain threshold, we just kick it out. [18:22.300 --> 18:24.140] We don't index the site. [18:25.160 --> 18:40.680] And opposite to that, for example, if we know that something is a really good link, for example, it was top of hacker news 10 years ago, but it's not on the web, we will try extra hard to crawl it and go to Internet Archive. [18:41.200 --> 18:44.800] So those will also be surfaced in Kagi search results. [18:45.040 --> 18:46.840] And these are two examples of this. [18:48.140 --> 18:55.880] We show how many trackers we detected on this site and blast from the past is this section in Kagi. [18:55.880 --> 19:00.400] So pretty simple thing that you can do, and we do, and it works like charm. [19:01.580 --> 19:10.260] And basically how Kagi works, we have a number of external APIs that we use because we don't have the money to build the index of the entire web. [19:10.360 --> 19:11.900] We are talking billions of dollars there. [19:12.100 --> 19:14.000] So we use external indexes. [19:14.080 --> 19:17.220] Then we process them through things that I just described. [19:17.340 --> 19:18.900] We try to remove the crap. [19:20.240 --> 19:23.320] Then we plug in the results from our own index. [19:24.080 --> 19:25.920] Then there's user personalization. [19:26.060 --> 19:27.180] That's the most important part. [19:27.440 --> 19:31.100] So user deciding what they want in their search feed. [19:31.220 --> 19:35.940] They're personalizing their search feed using lenses and other features. [19:36.300 --> 19:37.900] And finally you get the search results. [19:38.080 --> 19:41.020] So this is full end-to-end flow of how Kagi works. [19:42.680 --> 19:48.140] We have something called the Kagi Small Web, which basically is an initiative to surface human web. [19:49.260 --> 19:50.720] Anyone can add a website. [19:51.240 --> 19:53.520] And we surface them very high in our results. [19:53.740 --> 19:57.280] So here's an example of the UI of the small web, which you can access. [19:57.880 --> 20:05.040] But also this link, if you are looking for basic for encoding, surface is number three in our search results. [20:05.240 --> 20:10.380] So anything that's in small web, we basically surface very high, usually position two or three. [20:11.800 --> 20:20.200] And this is a conscious choice because we do want to platform personal blogs and basically humans. [20:20.560 --> 20:22.340] Our mission is to humanize the web. [20:23.640 --> 20:35.840] And people react to this very well, too, because this is often a way to find hidden knowledge, hidden gems, actual people writing on the web that does not exist in any other search engine. [20:35.840 --> 20:39.640] We just don't have the incentive to show these kind of results. [20:40.420 --> 20:49.080] Kagi Small Web is open-source, and we also open-source the fast feed parser, which we use to parse a thousand of RSS feeds. [20:51.420 --> 20:56.960] One other thing we do are the crowdsource block lists, like GitHub is full of these. [20:58.360 --> 21:01.800] So you have a block list for crappy, spammy websites. [21:02.100 --> 21:04.980] We just use them out of the box most of the time. [21:06.480 --> 21:08.980] I'm not sure if we are using these actual ones. [21:09.120 --> 21:10.460] I'm just giving them as an example. [21:12.180 --> 21:19.100] There's one to block AI generated images or domains with AI generated images, which we use in our image search. [21:20.260 --> 21:24.260] But we are now working on a model that will be detecting actual images. [21:24.460 --> 21:31.660] But this I'm just showing as a shortcut how you can achieve superior search quality. [21:31.860 --> 21:34.020] Like the information is out there. [21:34.100 --> 21:37.140] You just need to get it and implement it in your search engine. [21:40.140 --> 21:42.840] We built something called Kagi Privacy Paths. [21:43.080 --> 21:49.640] And one of the questions we get is, how can you guarantee my privacy when I need to log in? [21:49.740 --> 21:52.460] And you need to log in because you need to pay for Kagi. [21:52.600 --> 21:54.120] You cannot use Kagi for free. [21:54.740 --> 21:55.920] And that's the whole point. [21:56.140 --> 21:59.480] And so we build this technology, which is pretty cool. [21:59.760 --> 22:04.260] It uses state of the art cryptography. [22:05.220 --> 22:14.740] And what it basically does is guarantees from a technological perspective that we cannot know who the user using Kagi is. [22:15.420 --> 22:17.420] So first you generate some tokens. [22:18.100 --> 22:21.860] And then you can query Kagi using the token. [22:22.100 --> 22:25.840] And we don't know who the user generated the token is. [22:26.700 --> 22:31.320] And we have a pretty good blog post if you're interested in this. [22:31.380 --> 22:35.880] It's pretty novel use of this technology for this use case. [22:36.400 --> 22:39.120] And we open-source most of this work as well. [22:39.160 --> 22:41.580] You can find the extension in the backend libraries. [22:42.520 --> 22:46.240] And when you pair this with Tor, you basically have full anonymity on Kagi. [22:48.920 --> 22:51.020] We built a couple of other things. [22:51.300 --> 22:52.840] So we're not just a search company. [22:53.080 --> 23:03.020] We are thinking of Kagi as an ecosystem company trying to replace the whole big tech ecosystem, not just the search part. [23:03.460 --> 23:12.720] So we're building a web browser called Orion, which is WebKit based, zero telemetry, with a business model of pay to own your browser. [23:12.980 --> 23:23.300] Again, browser is even, I guess, more important than search because we do spend 10 hours a day in a browser and maybe an hour a day in the search. [23:23.640 --> 23:32.820] And yet we somehow allowed that most of our browsers are being paid for by somebody else, basically. [23:34.420 --> 23:38.400] There was this time maybe 20 years ago when you could pay for your browser and that's gone. [23:38.920 --> 23:40.440] And with Orion, we are bringing that back. [23:40.720 --> 23:45.340] There's a free version and then you can basically just opt to support the project. [23:45.480 --> 23:47.980] There are no new features you get if you pay for it. [23:48.140 --> 23:54.000] But you pay and it's $150 for a lifetime license and then you own your browser. [23:54.260 --> 23:54.560] And that's it. [23:54.660 --> 23:55.380] That's the business model. [23:56.060 --> 23:59.860] We have macOS and iOS and the Linux is coming in March 26. [24:00.440 --> 24:01.680] It works right now. [24:02.560 --> 24:05.140] We have another thing called Kagi Translate. [24:05.600 --> 24:10.220] Very cool, super high quality translation, tons of features, Kagi style. [24:10.520 --> 24:15.040] So we took a look at translation and did what we think translation needs. [24:15.280 --> 24:17.180] So I'm inviting you to check that out. [24:20.680 --> 24:23.440] Just want to show you relatively where we are. [24:23.640 --> 24:25.760] These numbers are not 100% accurate. [24:26.200 --> 24:36.920] It was a slide from six months ago or so and I was just too lazy to update the numbers but it gives you sort of the relative standing of things. [24:37.360 --> 24:41.460] And we are about a million searches per day today. [24:42.540 --> 24:45.340] These other guys maybe increased a little bit as well. [24:46.020 --> 24:51.760] So it may look small but I'm pretty proud of what we have built, to be honest. [24:52.060 --> 24:59.040] And every single one of the million searches done at Kagi today are paid for and delivered by the user. [24:59.680 --> 25:03.400] All that other stuff is free stuff, free searches. [25:03.720 --> 25:07.800] So we are about nine seconds of Google searches a day right now. [25:11.400 --> 25:12.100] Not bad. [25:15.260 --> 25:16.300] And that's all for me. [25:17.020 --> 25:18.120] Thank you so much. [25:31.610 --> 25:32.810] Thank you Vladimir. [25:34.010 --> 25:36.010] Again everybody we're opening up for questions. [25:36.270 --> 25:37.990] If you'd like you can come get in line for questions. [25:38.170 --> 25:38.770] We've got two mics. [25:39.750 --> 25:42.430] I will be watching the Matrix channel. [25:42.930 --> 25:45.470] Please tag Manchicken if you'd like to ask a question. [25:45.690 --> 25:48.250] I'll try and pull it out if you forget to do that but I can't promise. [25:51.430 --> 25:52.130] Thank you. [25:52.590 --> 25:59.410] Can you tell about international results, Russian, Chinese, how well does it work? [26:00.310 --> 26:01.690] The coverage is pretty good. [26:02.150 --> 26:15.910] The best way to... I am inviting you to check it out and see for yourself but we don't have any problems with international results because we do surface results from other indexes that cover them pretty well. [26:18.590 --> 26:20.970] I have a question for Matrix real quick. [26:21.530 --> 26:26.310] DuckDuckGo similarly promised an alternative to Google but seems to have been shitified as well. [26:26.870 --> 26:31.830] Where do you see them as fitting into the broader Kagi versus Google narrative? [26:33.910 --> 26:58.570] I respect what DuckDuckGo did and they started in 2008 when you know we didn't even think about alternatives to Google and I think they did a lot of very important work but I think at the end of the day they are also sort of the victim of the ad-based business model which prevents innovation in [26:58.570 --> 27:01.790] search just because the incentives do not align. [27:02.990 --> 27:12.890] And once you have a user that pays $10 a month for search like we do you better have all these things or they're gone because there are so many free options. [27:13.170 --> 27:22.090] So I think they were they have been the victims of the ad-based business model although maybe not a you know the intention was definitely to do something different than Google. [27:22.370 --> 27:23.230] So I wish them luck. [27:23.370 --> 27:27.470] I think they have a place in the ecosystem but we are doing something different. [27:29.850 --> 27:31.110] Thank you for the talk. [27:31.570 --> 27:40.950] I'm curious to hear your thoughts on how your browser Orion positions itself versus Mozilla Firefox and other privacy-oriented browsers. [27:42.370 --> 27:45.730] Well Mozilla Firefox is not a privacy-oriented browser. [27:46.610 --> 27:48.550] Let's just get that out of the way. [27:48.730 --> 28:01.470] When you start Mozilla Firefox it would send a ton of telemetry back and you cannot call yourself privacy-respecting if you're sending users personal information and when you do telemetry you send at least the IP address. [28:02.570 --> 28:04.790] So let's get that out of the way. [28:06.150 --> 28:12.010] The second thing Mozilla is financed directly by their biggest competitor. [28:12.630 --> 28:14.710] So that doesn't work long-term either. [28:16.150 --> 28:19.270] So I'm really not trying to position. [28:20.910 --> 28:22.970] We are just trying to do our own thing. [28:23.130 --> 28:26.810] So I believe browser is the most intimate tool you have on your computer. [28:27.070 --> 28:29.210] You should be owning your browser. [28:30.090 --> 28:31.650] With Kagi you can do that. [28:32.370 --> 28:34.490] And then it's also fast. [28:34.710 --> 28:35.490] It's zero telemetry. [28:35.610 --> 28:36.590] It supports extensions. [28:37.370 --> 28:43.590] So we're just trying to build a very, very good common sense, no BS browser. [28:44.290 --> 28:44.990] That's it. [28:47.890 --> 28:49.250] Another question from Matrix. [28:49.510 --> 28:53.830] Economics dictates the cheapest product always attracts the largest user base. [28:54.130 --> 28:55.430] Quality is not a factor. [28:55.630 --> 28:58.150] Facebook, Google, and many others have proven this. [28:58.490 --> 29:01.810] Is a large user base not a factor in your business? [29:02.730 --> 29:03.850] Luckily it isn't. [29:03.990 --> 29:04.870] Thank you for the question. [29:05.310 --> 29:11.770] We are not here to kill Google or be in any sort of way a Google killer. [29:11.950 --> 29:16.130] We are just here to do the right thing for our users, for my own kids. [29:16.310 --> 29:17.870] This is the reason I started the company. [29:20.110 --> 29:21.590] It is a private company. [29:21.770 --> 29:23.650] So obviously there is growth concerns. [29:23.810 --> 29:25.270] There are people working there. [29:25.550 --> 29:30.110] But ultimately if you ask me whether we are successful, I would say we already are. [29:30.290 --> 29:31.110] They're profitable. [29:32.210 --> 29:33.350] People love what we do. [29:33.790 --> 29:38.010] And so the world domination was never a factor. [29:38.770 --> 29:39.730] Let's put it that way. [29:40.870 --> 29:49.630] Yeah, so obviously the big selling point for this would be at least like for me the quality of the search results and like how all those are curated. [29:50.530 --> 29:51.550] I just wanted to ask. [29:51.710 --> 29:57.050] There is, I know there's a blog post by Dan Liu where he compares like a lot of the major and also some of the smaller search engines. [29:57.430 --> 30:04.810] And like specifically like puts them through like different like real world tests of different search cases to see like how quickly they could find results. [30:04.810 --> 30:14.770] And what he found was, I know that, and just in case like I don't know if you've read this or not, but what he found was that like Kage yields like similar results to like DuckDuckGo to Google. [30:15.310 --> 30:20.230] Just on the off chance, have you read this and would you know like if you have any response to that? [30:21.370 --> 30:24.450] Yeah, that blog post I think happened a year or two ago. [30:26.550 --> 30:30.810] I don't think that was accurate back then, certainly now. [30:31.330 --> 30:39.890] And I know he received a throng of emails from Kage customers basically fixing things in his blog post. [30:41.350 --> 30:50.770] You know, we invite independent testing and at the end of the day every single customer of Kage would not be paying if the results were not better. [30:52.910 --> 30:56.290] So yeah, I just invite anyone to try and see for themselves. [30:56.630 --> 31:00.530] That's the best, you know, best way to do that. [31:03.730 --> 31:07.390] Hi, big fan coming on two years of membership. [31:07.550 --> 31:07.790] Thank you. [31:07.910 --> 31:08.350] Enjoy it. [31:08.550 --> 31:15.110] And I have also recently started getting a lot of use out of the Assistant feature. [31:15.110 --> 31:18.330] It's my primary like LM interface at this point. [31:19.170 --> 31:22.190] It's also a loss leader for a lot of companies in the space. [31:22.350 --> 31:28.310] I'm sure perplexity and you know, all these other companies are losing buckets of money providing these services. [31:29.110 --> 31:39.930] Are you in a similar state with Assistant or is it, you know, in isolation a profitable product for you as something that's included in your subscription? [31:40.290 --> 31:45.010] Yeah, so the reason it's loss leader for all these companies because they offer it for free. [31:45.110 --> 31:49.750] They offer it for free because of the growth or the imperative to grow. [31:50.070 --> 31:51.230] We don't have that. [31:51.490 --> 31:56.530] And the only way to access Assistant is if you are a Kagi customer. [31:56.950 --> 31:58.550] So you pay for your use. [31:58.730 --> 31:59.890] So there are no... [32:00.770 --> 32:02.070] It's not a loss leader. [32:02.270 --> 32:04.950] I think it's something that makes a product stickier. [32:05.250 --> 32:13.770] Now, when you have search and then you have AI and you can merge the two, you can merge access to all LLMs with the best search around. [32:14.010 --> 32:16.990] I think it produces superior results. [32:17.210 --> 32:21.170] So for us, it's a value add. [32:21.890 --> 32:27.250] And is it a direct pass through like usage cap that you have as part of the subscription to keep that in check? [32:27.450 --> 32:32.850] Because you've got these cases with like people racking up the equivalent of like thousands of dollars on cloud code. [32:32.850 --> 32:38.210] Yeah, you cannot do that in Kagi because we limit you by what you pay for your subscription in tokens. [32:38.510 --> 32:42.350] You can buy more, but most people never hit those limits anyway. [32:42.590 --> 32:43.970] But we had... [32:44.610 --> 32:48.470] We did not have those limits before we found about the abuse. [32:48.710 --> 32:51.230] So we had to put them in place, obviously. [32:51.810 --> 32:52.550] And so... [32:52.550 --> 32:53.150] Cool. [32:53.290 --> 32:53.990] Thanks very much. [32:54.130 --> 32:54.390] Thank you. [32:55.550 --> 32:56.270] Hi, Mudd. [32:56.430 --> 32:57.050] Thanks for the talk. [32:57.990 --> 33:02.070] You guys clearly have some very like talented people on the team, developers. [33:02.390 --> 33:05.450] Like I'm really impressed how like simple and effective these features are. [33:05.570 --> 33:09.690] Like the regex site swap and then uBlockorigin. [33:09.790 --> 33:12.290] Those are like really elegant ways to solve the problem. [33:12.590 --> 33:13.570] Like what's your like... [33:13.570 --> 33:16.570] How do you attract those kind of people that think that way? [33:16.670 --> 33:21.030] Because that's like, I think that's a different way of approaching software than a lot of developers are used to. [33:21.130 --> 33:22.350] So like, what's your like pitch? [33:23.190 --> 33:25.570] That's... Can you guys stand up? [33:25.650 --> 33:26.750] We have three of them here. [33:33.470 --> 33:34.710] That's a pretty good question. [33:36.570 --> 33:43.650] Well, you certainly had to be crazy to come work at Kagi in the beginning, which is how the guy with the cap came on board. [33:45.350 --> 33:47.930] As hey, we're going to make this paid search engine. [33:48.570 --> 33:49.710] Okay, let's do it. [33:51.410 --> 33:53.370] So there's certainly an element of that. [33:54.030 --> 33:58.810] And we do look for people who are just naturally curious. [33:59.390 --> 34:01.590] But we also put them through a lot of... [34:04.210 --> 34:05.970] Our interviews are not the easiest. [34:06.150 --> 34:06.970] Let's put it that way. [34:07.310 --> 34:10.210] And we make people work for it. [34:10.390 --> 34:19.690] Because we really want to end up only with those who are really persistent, really want to come and work at Kagi. [34:20.590 --> 34:24.410] And by that, I mean really, really, because we don't pay much too. [34:25.730 --> 34:27.690] Because we are funded from user subscriptions. [34:27.950 --> 34:28.250] That's it. [34:28.650 --> 34:31.050] The salaries grow as subscriptions go. [34:31.370 --> 34:40.130] But, so I think this is what produces the talent and just just being frugal and, you know, have to think outside the box, I would say. [34:40.930 --> 34:53.430] But yeah, mostly thanks to that hiring process, which is, you know, a lot of engineers are, I would say, it's not what they expect. [34:53.690 --> 34:56.710] We are definitely not like Silicon Valley type companies. [34:57.430 --> 35:00.450] Everything is, you know, honey and milk. [35:00.810 --> 35:02.970] No, it's super hard to work at Kagi. [35:04.370 --> 35:07.390] But yeah, 50 of them do that. [35:07.630 --> 35:08.710] I'm thankful for that. [35:10.550 --> 35:12.070] Hey, Vlad, thanks for the talk. [35:12.830 --> 35:14.570] I'm a Kagi customer, love the product. [35:15.490 --> 35:30.930] Going back to AI a little bit, one thing that, and maybe this is something I missed, but I couldn't really find on the site, is when we look at all those different AI models, how is Kagi kind of, I guess, influencing privacy behind them, right? [35:31.010 --> 35:32.450] Are you guys using dedicated tenants? [35:32.790 --> 35:41.870] Are you, when you make agreements with these companies, are you opting out of those questions that Kagi customers are asking to be training data in the future, things like that? [35:41.870 --> 35:43.510] And can you speak to that a little bit more? [35:43.770 --> 35:43.790] Yes. [35:44.550 --> 35:55.250] So most of the API providers have, in their terms of service, when you use them through the API, that they will not use the queries to train data or to store them. [35:55.410 --> 35:56.650] So we do that. [35:56.870 --> 35:58.730] And we work with providers who have that. [35:58.910 --> 36:04.030] And when they have an option to, you know, opt out and things like that, we like to do that. [36:04.130 --> 36:09.450] We actually have a page on the in the documentation that details every LLM that we work. [36:09.630 --> 36:11.010] What is the actual provider? [36:11.010 --> 36:12.290] What are the terms of service? [36:12.550 --> 36:13.810] And how is privacy preserved? [36:14.130 --> 36:17.610] So we tend to work only with those that preserve the privacy. [36:18.650 --> 36:19.090] Awesome. [36:19.290 --> 36:19.550] Thank you. [36:19.770 --> 36:20.030] Thank you. [36:22.150 --> 36:23.550] I have two questions. [36:23.850 --> 36:30.970] The first is, I know you're not trying to conquer the world, but how big do you think the market is for privacy conscious products? [36:31.250 --> 36:36.750] And can it sustain like a whole ecosystem of similar privacy conscious products? [36:36.950 --> 36:46.350] And the second question is, just curious, like how much are people on Kagi Assistant, how much are people using open-source model sources, closed source? [36:48.810 --> 36:52.790] Well, we don't pitch Kagi as a privacy product. [36:53.110 --> 36:56.450] As I said, privacy is basically a side effect of the business model. [36:56.450 --> 36:58.170] Or let's say like non and certified. [36:59.610 --> 37:01.450] For that, I think it's pretty big. [37:02.710 --> 37:05.210] 700 million, I think, is the latest data. [37:05.410 --> 37:07.630] People use an ad blocker in the browser. [37:08.970 --> 37:12.930] They call it the largest protest against the business model in history. [37:13.150 --> 37:14.010] But that's not paid, right? [37:14.610 --> 37:20.130] No, but I'm saying that, you know, there's a protest against the business model now, whether they pay or not. [37:20.510 --> 37:27.490] There are, I think, 35 million people paying for YouTube premium, which is to avoid ads most of the time. [37:29.670 --> 37:37.730] So, you know, when I think about Kagi, I think there are 5 million people out there who would use it today if they knew about it. [37:39.090 --> 37:41.570] But we don't run, you know, a lot of marketing. [37:41.590 --> 37:43.430] It's mostly word of mouth and things like that. [37:43.590 --> 37:45.370] So, I think the number is growing. [37:45.850 --> 37:47.790] And it takes time for these kind of things. [37:47.970 --> 37:50.110] I often compare it to organic food. [37:51.130 --> 37:54.750] And, you know, 40 years ago, we used to eat crap. [37:55.270 --> 37:56.370] And at least I did. [37:57.010 --> 38:03.270] And then there were some crazy people saying, you know, you need to, you know, grow organic food and do this. [38:03.270 --> 38:06.770] And now I'm paying like three bucks for my avocado at Whole Foods. [38:07.050 --> 38:08.350] And a lot of people are. [38:08.490 --> 38:10.170] So, it took like 40 years for that. [38:10.390 --> 38:20.130] So, I don't think this will take us as long because these big tech companies are really shooting themselves at the foot every single day at an accelerating rate. [38:20.530 --> 38:22.570] It's just the crazy stuff they're doing. [38:23.110 --> 38:32.050] So, I think more and more people will wake up from the matrix, as I call it, and just, you know, try and start to value the information they're getting. [38:32.250 --> 38:34.510] Like, we put a very little value on information. [38:34.630 --> 38:35.590] We just took it for granted. [38:36.470 --> 38:48.590] But, you know, we are discovering as a society, information is super valuable, you know, just at a society level because it, you know, dictates elections and our medical and science and education, but also at a personal level. [38:48.790 --> 38:54.670] If I have access to better quality results, I'm going to be more competitive at what I do with my work and things like that. [38:54.870 --> 38:58.230] So, I think that society is waking up to this fact. [38:58.390 --> 39:00.790] And I do expect to see Kage competitors in the future. [39:02.690 --> 39:03.170] Yeah. [39:03.850 --> 39:11.610] And about the open-source versus closed source models, I don't have the data with me, unfortunately. [39:11.610 --> 39:19.350] I don't know if we even track, but you can ask some of the people from the team, maybe they would have access to that. [39:20.910 --> 39:21.430] Thank you. [39:25.150 --> 39:25.750] Excuse me. [39:25.910 --> 39:31.730] So you mentioned how it's obviously quite problematic for Mozilla that they're funded by their largest competitor. [39:32.070 --> 39:42.090] And so that got me thinking about how, so you have your own small index of the web, or you have an index of the web, which seems like maybe it's not the main thing. [39:42.190 --> 39:46.130] Then you also have all these other indexes that you're relying on, which are provided by your competitors. [39:46.450 --> 39:47.450] So how does that work? [39:48.210 --> 39:52.290] And you know, how do those agreements work? [39:52.410 --> 39:56.010] And is it feasible for you to build something as big as that? [39:56.090 --> 40:01.750] I think there were some claims during the Google antitrust case that like, it's just really hard to do that for some reason. [40:01.930 --> 40:06.570] It's like, you know, it used to be that anyone could just go make a web crawler, but I guess, is that harder now? [40:06.570 --> 40:07.630] Great question. [40:08.470 --> 40:13.250] So yes, we are basically using indexes of our competitors. [40:13.970 --> 40:19.190] The way we mitigate some of the risk is that we use a lot of them, five or six. [40:19.470 --> 40:24.030] So for, you know, business continuity reasons and also for enhancing search results. [40:24.190 --> 40:25.610] We pay for a lot of that. [40:25.850 --> 40:26.950] Users don't see that. [40:27.050 --> 40:28.310] It all happens in split second. [40:28.650 --> 40:30.690] But that is what increases our cost. [40:30.750 --> 40:35.070] And we do that both for search quality and also business continuity. [40:35.490 --> 40:39.410] So if, you know, we get cut off from one, we lost 15%. [40:39.410 --> 40:45.370] Not, you know, in case of DuckDuckGo, I think they are mostly using Bing, for example. [40:45.510 --> 40:50.690] So if they get cut off there, it's like somebody says submarine tied to a boat. [40:51.330 --> 40:53.170] The submarine dies. [40:53.650 --> 40:54.350] That's it. [40:55.550 --> 40:57.370] So there is that. [40:57.590 --> 41:02.250] The second part of your question about is it feasible to create a [41:06.890 --> 41:09.330] index that's the entire web today. [41:10.750 --> 41:12.630] And, you know, it's a pretty deep topic. [41:12.930 --> 41:18.830] We are doing a lot on that stuff, although we are focusing on non-commercial web. [41:18.930 --> 41:23.550] I think largely that train has left the station in early 2000s. [41:24.750 --> 41:32.150] And the way to do that is exactly the antitrust because Google did build that by being a monopoly. [41:32.750 --> 41:39.530] And I don't think it's feasible to achieve the same level of quality, or at least level of quality the users would pay for. [41:40.190 --> 41:43.390] Like, I cannot imagine anyone paying for Bing, for example. [41:43.870 --> 41:47.190] So there's a level of quality that's needed. [41:47.550 --> 42:00.510] And we have expressed our opinion on this in a pretty long blog post discussing remedies for the antitrust, which I believe, you know, divesting Chrome and things like that are useless remedies. [42:00.810 --> 42:05.430] What is really needed is for free access to Google index and data. [42:05.650 --> 42:11.570] So any startup in the world can build on top of it. [42:11.770 --> 42:14.130] So then you would have thousands of startups. [42:14.310 --> 42:20.390] You could have your search engine for your own town, or search engine for cats, and things like that. [42:20.490 --> 42:27.590] That would proliferate the creation of search startups, which is the goal of the Sherman Act and the whole antitrust. [42:27.850 --> 42:30.030] And so we deeply believe that is the way to go. [42:30.230 --> 42:30.970] And we'll see. [42:31.370 --> 42:33.310] The verdict should come soon. [42:33.450 --> 42:34.670] So we'll see what happens. [42:34.930 --> 42:36.890] Or the remedies will come soon. [42:37.930 --> 42:51.040] So... From Matrix, we have, uh, how much accessibility of Kagi is, uh, sorry, how much is accessibility of Kagi considered? [42:51.200 --> 42:53.880] Have you done user testing with disabled users specifically? [42:54.180 --> 42:56.860] Do you plan to publish an accessibility statement? [42:57.160 --> 42:59.720] Do you have a roadmap for accessibility improvements? [43:00.620 --> 43:02.820] Thank you very much for that question. [43:02.820 --> 43:06.680] And I'm proud to say, we are super big on accessibility. [43:07.460 --> 43:16.740] Uh, when I used to run our product roadmap, uh, anything regarding accessibility would be pushed to the top of the queue no matter what. [43:17.540 --> 43:21.720] And, uh, uh, we do have a lot of blind users. [43:21.720 --> 43:27.940] They, they participate in discussions in our forums, and they are first to tell us if there is something off. [43:28.100 --> 43:29.900] So we do pay a lot of attention. [43:30.100 --> 43:38.060] We haven't published any official documentation on that, or things like that, which, uh, I think is a great idea, and we'll do that. [43:38.280 --> 43:41.920] And we are also thinking about expanding accessibility. [43:42.720 --> 43:50.620] Not, not in a typical sense, but, um, for example, you should be able to access search results through email. [43:51.620 --> 43:52.520] Nobody did this. [43:52.660 --> 44:01.520] Like, why wouldn't you be able to email search at Kagi.com and get search results, or ask your assistant something, or translate something. [44:01.520 --> 44:11.100] So we are also thinking about accessibility in terms of, um, you know, how do you access the product, not necessarily how do you use it. [44:11.360 --> 44:21.780] So accessibility pretty high on our list, and I invite the person asking this to, to get in touch, uh, directly, and, you know, happy to chat about that, and work on opportunities to improve. [44:24.160 --> 44:24.700] Hi, again. [44:25.160 --> 44:30.700] Um, what are, like, some existential risks, perhaps, to Kagi, or the business? [44:31.080 --> 44:36.520] Like, because, like, earlier you kind of talked about, there's, like, a kind of a tap at the top of external APIs. [44:37.020 --> 44:41.480] Is there a concern that they could, if you guys get big enough, they might shut off the tap? [44:41.880 --> 44:44.240] Or, I'm just, this is kind of an open-ended question. [44:44.480 --> 44:59.240] Like, technical... I mean, for most of them, we are a distribution channel, so, you know, they get... So, for a typical search engine, a... our typical user is somebody who uses an ad blocker. [45:00.160 --> 45:04.920] I would say 95% of Kagi customers probably have an ad blocker, one way or the other. [45:05.440 --> 45:09.860] And so the search engines are not making any money, the typical ones. [45:09.980 --> 45:12.220] At least they get something through us. [45:12.500 --> 45:17.420] So, in a way, we are monetizing those users they wouldn't... otherwise never would. [45:17.680 --> 45:19.100] So, I don't think there's an incentive. [45:19.400 --> 45:24.860] And again, we really... I, I respect Google's, uh, people and technology. [45:25.000 --> 45:28.400] I think they have wonderful, uh, people and technology. [45:28.580 --> 45:32.940] It's just that their business model is profoundly bad for, for society and users. [45:33.280 --> 45:34.720] That's my only complaint. [45:35.060 --> 45:41.840] Um, but, you know, I don't, I don't have, uh, intentions to kill Google or dominate the world in that sense. [45:41.980 --> 45:45.160] So, I don't think of us really as threat. [45:45.280 --> 45:50.000] And we have actually approached them and said, you know, we are happy to share some of our insights. [45:50.140 --> 45:52.460] You guys should launch a paid version of Google. [45:52.620 --> 45:53.600] We would love that. [45:54.440 --> 45:57.600] Because that would educate the market that there's such thing as a paid search. [45:58.260 --> 46:03.880] Like, we are right now number one on Google for paid search engine, but nobody's searching for that. [46:04.120 --> 46:06.180] So, we don't have a lot of views for it. [46:06.520 --> 46:08.040] Um, anyway. [46:08.240 --> 46:11.420] So, those are just my thoughts about how, how I think about those things. [46:11.660 --> 46:16.180] So, in terms of existential threats, it's mostly us screwing up in a big time. [46:16.760 --> 46:19.020] This is what I tell my team all the time. [46:19.280 --> 46:23.540] We have everything laid for us. [46:23.860 --> 46:25.720] A lot of tailwind. [46:25.720 --> 46:28.680] Uh, it's, it's only us who can screw this up now. [46:29.060 --> 46:29.080] So. [46:29.440 --> 46:29.920] Okay. [46:29.980 --> 46:30.300] Thank you. [46:30.500 --> 46:30.700] Thanks. [46:33.880 --> 46:38.720] The paid model may be financially prohibitive for people around the world. [46:38.940 --> 46:46.780] Is there any plan to have paying customers pay a bit more to help subsidize the cost of users that may find it financially prohibitive to use it? [46:47.460 --> 46:48.380] Wonderful question. [46:48.680 --> 46:52.220] So, there are two ways that we are thinking about this. [46:52.220 --> 46:59.720] One is, there is a gift feature right now in Kagi where you can just give subscription to friends and family. [47:00.760 --> 47:14.620] The other most scalable one is the initiative we just launched called Kagi for libraries, where we are basically giving free access to library patrons and participating libraries, hopefully around the world we have. [47:14.860 --> 47:20.240] We just launched this last month, I believe, and we have already a queue of about 50 libraries. [47:21.140 --> 47:30.060] So, anyone who cannot afford search but still wants to use premium search, wants to use Kagi, will be able to go to their local library and access it for free. [47:30.600 --> 47:37.060] So, yeah, this, this, this is our way to, to, uh, you know, solve, solve this problem. [47:39.520 --> 47:39.960] Hello. [47:40.480 --> 47:52.380] Um, all across the websites, especially some of the small websites that you're looking at, are getting increasingly aggressive about blocking crawlers, which obviously has an impact on the search business as well. [47:52.560 --> 48:00.300] Even though, of course, you try not to block a search crawler, but an AI crawler, eventually this ends up getting merged. [48:00.520 --> 48:02.160] Do you see this as a concern? [48:02.340 --> 48:11.800] And do you have any ideas around tactics for search engines and like, uh, sort of post crawler world, not really, or at least crawlers as we have now? [48:13.400 --> 48:17.860] Yeah, uh, so I'm not worried about small web. [48:18.220 --> 48:25.640] Most of small web or the human web are people who are doing that, um, in addition to their career. [48:25.900 --> 48:28.940] So, they're just creating content, do not expect anything in return. [48:28.940 --> 48:36.080] And they see a lot of value in, in, in our initiative because we surface them and bring them, uh, readers. [48:36.380 --> 48:48.960] But, um, I think on, uh, you know, if you zoom out, you have this concept now of AI, uh, basically crawling websites, synthesizing answers, giving them to users. [48:49.220 --> 48:55.820] And this is something that may become unsustainable for the most of the web. [48:56.120 --> 48:58.200] And there will be changes for sure. [48:58.520 --> 49:11.300] Uh, the small web will not change, but those websites that counted on search engine traffic to monetize their users, usually through ads, unfortunately, um, or affiliate marketing and things like that will have a big problem. [49:11.440 --> 49:12.600] And they are already a big problem. [49:13.080 --> 49:29.960] Um, so they will, you know, in my view, many of them will either have to switch to a subscription model and really stand behind the content that they are creating, because if it's not valuable enough that nobody will pay for, all they will cease to exist. [49:30.320 --> 49:41.220] Um, there is another model that we are thinking about, uh, and we started thinking about even before our lens, um, the model is profit sharing. [49:41.440 --> 49:46.640] Basically, we want to profit share with anyone appearing in our search results. [49:46.800 --> 49:48.300] That was the idea from day one. [49:48.480 --> 49:55.860] And nowadays it's a profit share with anyone that our AI assistant takes content from, and it would work automatically. [49:55.860 --> 49:59.460] There will be no fancy publishing deals to sign. [49:59.620 --> 50:07.900] If you are a website and you show up in CAGI answers, we just attribute you proportionally to, say, the text that was used. [50:08.140 --> 50:12.880] And then you can sign up at CAGI Webmaster Tools, and there's a check waiting for you at the end of the year. [50:13.280 --> 50:15.060] So we want to automate all of that. [50:15.400 --> 50:19.820] Um, obviously this, uh, starts to make sense once we are bigger. [50:20.120 --> 50:24.320] Because if, if the check is for two cents, then it doesn't solve the problem. [50:24.480 --> 50:26.340] So there's a lot of considerations. [50:26.580 --> 50:30.960] We are thinking deeply about this and, you know, happy to connect after to go into more detail. [50:32.200 --> 50:32.720] Sorry. [50:32.900 --> 50:33.940] I apologize, friend. [50:34.100 --> 50:40.120] I'm going to take the last question from, uh, Matrix, uh, without doxing yourself. [50:40.260 --> 50:41.120] I'm adding that part. [50:41.400 --> 50:45.860] Uh, how do I get in touch directly to follow up with Vladimir? [50:46.560 --> 50:54.180] Yeah, here's the, uh, uh, Vlad at CAGI dot com is the easiest. [50:55.800 --> 50:56.360] Excellent. [50:57.120 --> 50:58.360] Thank you very much, Vladimir. [50:58.960 --> 51:01.640] Let's give a round of applause to Vladimir. [51:02.040 --> 51:02.680] Thank you. [51:02.900 --> 51:03.160] Thank you.