So I've been using Kagi for a while now as a paid search engine. I always thought it's $25 a month plan was a little steep for search, but a) I got work to pay for it, and b) startpage nee google was getting less and less useful, and bing and whatever used it has... well been worse for me always.
Anyway, I just got told that they've now adjusted their pricing / added features to Ultimate, and I think (at least now) that's actually added a lot of value if you're into the more advanced LLVM / AI models / chat. I have also been paying $20 a month through work for ChatGPT Plus. I might drop that because Kagi now lets you chat with / use GPT4 as well as Claude2 and a Google LLVM model with the one $25 a month, in addition to all the search and AI Search (with sourcing) together.
I don't know how well paid search is going to ever do - it might be a short term tool. But for now, not having ads in the search, a straightforward pay for service model that seems to work just as well with their stated privacy goals, and getting multiple AI LLVM is pretty cool "one stop shopping" if you will. I also like giving a shot to less ad based models for Internet services that I can't see how they don't become privacy invasions.
Well, the reality is, search costs money. Quite a lot of money it seems.
So that is either paid for by you, or by someone else. Nobody is going to run search as a charity. So it's going to be paid for by parties interested in paying for your attention.
Even if you run ad blockers or use meta search engines like searx, you are going to be finding results by companies that have paid to be there.
I am a heavy search user. My search quantity is reasonably large just from personal use (I'm a curious dude, what can I say?) but my professional use of search as a software developer is staggering some days. My anecdotal experience is that that Google search has been declining in quality for years, and especially over the last two or three. DuckDuckGo is a nice alternative for privacy (potentially), but I while I find myself feeling less in a walled garden with them, I don't actually find their results to be any better than Google's.
I have tried Kagi recently. So far, I really like it. I genuinely feel like I get good results (read: find something quickly that is relevant to what I searched). I love their lensed searches that let you search the indie-web, and I love that they let you add weightings to websites that you trust.
It is expensive, no doubt. But for a certain audience that relies on quality web search, prefers to not be walled in by paying search engine optimizers and values paying for a product rather than opting to be the product, Kagi offers a solution.
Having said that, I would love to see the cost come down and make it more accessible to the many and I appreciate that for most people, the "free" search engines are good enough.
I see all your points and fully get it. I just naively wish something new will pop up by popular demand, some breakthrough idea out of the box, like some kind of open source search engine supported by p2p network or federated instances where everyone would contribute resources. If such demanding projects like operating system or social media can be open source then I don't see why search engine couldn't be.
I have a feeling that you mainly wish for something that you can use for free.
Reality is, money needs to come from somewhere. It will either come from you, or someone who might have different motivations than you do.
That's not reality, that's a mindset promoted by corporations for the last decade. How much do you pay for using Lemmy, Linux, Mastodon and other FOSS?
How the fuck do you think Lemmy, Linux and Mastodon are sustaining themselves, if not with money?
I donate to open source projects that I feel need the support, because unlike you I don't take them for granted.
And what exactly would stop people from doing the same for search engine project?
You tell me. What's stopping you from donating to SearXNG?
SearxNG is kind of a bad example, since they are only a meta search engine and don't do any of the heavy lifting.
I wonder how big the database needs to be for useful search, and how much new data is added every day. Is this something that could realistically be run locally or widely mirrored?
Nope, not realistic for "mirroring". Federated could be possible, but I wouldn't have high hopes about (good) latency and coverage.