I desperately wanted Kagi to be my new search engine, but alas, no dice. At the end of my 100 search trial I wasn't even sure it was a practically better option than DDG. And believe me, I tried: I even coerced myself to stop adding !g and made it my browser's default search engine. Sometimes the results were good, and it reliably searched for what I asked it to, which is nice. OTOH though, it felt like it didn't have a deep index and, when I searched for things that were obscure but I knew existed, I sometimes found myself needing to try DDG and Google after.
Unfortunately, before being concerned about carbon footprint, the product needs to come first. I have no idea how an actually good search engine can be profitable and sustainable without ads and tracking, but it doesn't do any good to have a lower carbon footprint than your competitor if you're dead in ten years.
Surely eventually, something will happen. Someone will have a breakthrough moment. As of now, all that's happening in my view is that Google is getting so much worse that it just lowers the expectations for competitors. Maybe eventually it will get so bad that you may as well never use it, but that's not exactly a fantastic outcome.
Is building a good search engine and index from scratch in 2023 simply infeasible? I know Google went through great pains to optimize their index and had a couple novel tricks for ranking results... but the internet is infinitely more massive in every dimension, despite the death of literally libraries of useful blogs, forums and topic-specific websites, and now it seems even Google, with their massive budget, is struggling to deal with SPAM on top of ruining the search experience in general. It's easy to just blame enshittification, because duh, but I wonder if more specifically, the Internet is literally just collapsing under the weight of its own growth and an increasingly weak signal to noise ratio.
I think it depends a lot on the user. I never found DuckDuckGo to have good results, I think because it's based on Bing, and that remains laughable to this day (I have recently had to use it a bit and holy hell it's shit). I am comparing to startpage, which is I guess a more anonymous Google.
Kagi is at least as good for my day to day use, though I can't say it's obviously better. What it does have that I find useful is more ability to rank forum posts and the like higher, where they often have the real answers. Along with the "lenses" which let you kind of say what sort of stuff you're looking for - like only forum posts.
Right now, they also added to the Ultimate level a bunch of AI models you can chat with, which actually have pretty up to date knowledge it seems, and wider knowledge. For example, Claude 2 via Kagi was able to both help me as a DM with making a D&D 3.5e NPC (GPT4 can't do that alone, or at least didn't for me, and the plugin I tried to use that was suggested just didn't work at all), all the way to actually quickly finding JunOS commands and giving useful info about various outputs and clarifying what to do - which so far seem to be correct.
But the actual value to me is I can also ask GPT4 from Kagi, and Google Bison, so I don't need multiple subscriptions. I don't actually know how long this will hold true, but for now Kagi does a lot of stuff in one place that's pretty useful for me.
I typically use search engines not to answer questions but to search for things, like for example, maybe I want to find an old Qt Quarterly because something on HN made me think about it, or I'm looking for really deep MSDN documents about Text Services Framework to research something about how IME works on Windows. My biggest issue with Kagi is that whatever indices it seems to pull from seem less deep than both Bing and Google. Now, I could be wrong and maybe it's actually bigger or it uses one of them for some of its results, but my experience was that I was able to get results on DDG when Kagi couldn't return any for a given query.
Generally, I find that Google is a bigger index with a better ranking algorithm on top of it. However, DDG/Bing do generally seem to have an impressive index, and one of the biggest differentiators for most search engines that are not Google is that most search engines that are not Google will generally allow you to search for what you want to search for. I feel like with Google, it's very difficult to force it to always include all of the exact terms I'm specifying; it wants to search for something else. Sometimes I'm not even sure it's able to differentiate between two similar looking terms!
With all of this said, Kagi definitely was nice in terms of the fact that it felt like it was doing a good job of searching for what I asked for and an alright job ranking them. I gave it my best shot. But, alas, I never got to the point where I felt comfortable not searching Google afterwards. Even as much as Google pisses me off, it feels like their index is just far and away the largest, and it's not even really close.
A lot of people seem happy with kagi, myself included. I haven't used g! for months now. I really like some of their tools like the panel of more obscure results.
Yeah, I'm very confused by this post. The last few days/weeks Kagi has taken _very_ long to respond. I end up refreshing the page because it doesn't even show a loading indicator -- just a blank results page.
Interesting that there's several corroborators here. My own Kagi searches have been consistently speedy. I'm a heavy Kagi user on a Kagi for teams plan, in the UK, if any of that is relevant.
The carbon footprint of an LLM writing a ton of useless corporate email is much lower than the all in footprint of employees writing these useless emails themselves.
however what i found to be an inconvenience and something i do more often then i thought was google businesses for opening times, phone numbers and addresses. Which is nicely presented in a google search and not so in Kagi.
The title is editorialized, and the CO2 data is just garbage. Trying to estimate the CO2 footprint just from the transfer size makes no sense.
Drilling through the links to the source[0], their methodology is to take the estimated total energy use of the Internet (networking, servers, clients) and divide it by the estimated volume of end user traffic in the Internet, and multiply it by the worldwide average CO2 emissions / unit of energy. And that is supposed to give them a meaningful CO2-per-byte measure.
You can check to which google cloud location you connect and see the latency, by opening the settings side bar on the main page and have a look at the bottom.
They don't even include the marginal query costs. It's just the cost of sending the data to the user's web browser. And it appears to assume the data is served over a mobile phone network.
Possibly to corporations that have to perform carbon reporting or individuals who it’s important to.
Several tech companies (Google, for example) use ElectricityMaps data and models [1] to orchestrate their workloads around carbon intensity, for example.
One of these search engines scores a direct hit on the search "Kagi is 80% faster and 70% less carbon intensive than Google" and the other produces dozens of nonsense snippets that are barely related.
wonder if they have large enough data to search like Bing, DuckDuckGo or Google otherwise the statistical data shown is moot and purely marketing but not of real life?
If DDG or Kagi had an index large enough to cover the long tail of the web like Bing and Google do, we'd know about it. They'd have large data centers of their own, not to mention the papers and patents covering all the optimizations necessary to crawl, index and serve that without losing money. Instead, that work is outsourced to Teragoogle and whatever its Bing counterpart is called.
"The web is big. You just won't believe how vastly, hugely, mind-bogglingly big it is."
So all about 1g co2 per search. Compare to 150–200g per km of driving. Id say replace one car trip this year with bicycling, and you’ll be good to use Google.
If you mean searching for bars, shops, opening hours, etc. It does not work that well, I think this works only on kagi maps currently. But searches in my language (german), even about things near me works very well.
I give zero fucks about carbon footprint of my search engine. I care about giant barges hauling 9 billion bullshit items from china unleashing untold amounts of garbage and pollution into the ocean.
I also enjoy kagi and have been trying to use it more recently
Yeah it's an essential part of the "guilt economy" for products to advertise marginal improvements to environmental metrics via their purchase. People want to feel they are better people by using A if it is X% less evil than B, and will pay a premium for A. Something funny about it is that as long as environmentally better products are more expensive, poverty will always be implicitly demonized.
> This reminds me of people who use dark mode then claim they are saving electricity because black uses less energy. Lulz.
I mean, that's literally true for OLED devices. It's in the bucket of "who cares if you burn 10W on a screen when your appliances are burning multiple kW at once", but it is strictly true.
Also, weirdly enough, the recent huge spikes in temperature we’ve seen worldwide has in part been due to regulations put in place to prevent some giant barges from burning the dirtiest fuel possible, preventing huge amounts of sulfur dioxide from blanketing our oceans. That gas actually causes cloud seeding and the lack of it recently has significantly increased global temperatures. Scientists are now scrambling to figure out a less toxic way of emulating exactly the effect those huge barges had to block the suns radiation. Just wanted to throw that in because these conversations are can be more complex than they seem on the surface.
The existence of major transgressors of laws/regulations/norms/ethics shouldn’t factor into our care for reducing the prevalence of minor transgressors.
FWIW, the scope 3 GHG emissions of Alphabet seem to be similar to a major airport like London Heathrow.
Consumerism is going to be a difficult battle to fight.
One small way software engineers can try to fight it though is by trying to reduce electronic waste through having a well-maintained high-quality polished Linux distro for mobile phones, so that we’re no longer forced to suffer a lack of software updates due to a “end of life” determined by the manufacturer.
Unfortunately, before being concerned about carbon footprint, the product needs to come first. I have no idea how an actually good search engine can be profitable and sustainable without ads and tracking, but it doesn't do any good to have a lower carbon footprint than your competitor if you're dead in ten years.
Surely eventually, something will happen. Someone will have a breakthrough moment. As of now, all that's happening in my view is that Google is getting so much worse that it just lowers the expectations for competitors. Maybe eventually it will get so bad that you may as well never use it, but that's not exactly a fantastic outcome.
Is building a good search engine and index from scratch in 2023 simply infeasible? I know Google went through great pains to optimize their index and had a couple novel tricks for ranking results... but the internet is infinitely more massive in every dimension, despite the death of literally libraries of useful blogs, forums and topic-specific websites, and now it seems even Google, with their massive budget, is struggling to deal with SPAM on top of ruining the search experience in general. It's easy to just blame enshittification, because duh, but I wonder if more specifically, the Internet is literally just collapsing under the weight of its own growth and an increasingly weak signal to noise ratio.