Kagi's "Small Web" offers a handpicked collection of more than 30,000 non-commercial, human-authored websites, including personal blogs, webcomics, and independent videos.
Their ‘Research’ AI is pretty decent too. I can get to the same ends eventually, but I can’t read and digest 25 different pages and then launch another 10 follow-up searches in 15 seconds. It’s summarizing and not simply inferencing so the hallucination rate is acceptably low and it also cites the sources so you can click a footnote to fact-check any important details.
I self-host most things, but I can’t self host a search engine. I much prefer paying monthly than having to fight the eternal battle against tracking and ads.
I have this on my projects lists, I’ve seen it recommended in a few places.
If I’m understanding the purpose (privacy filtering for searches), Kagi is a service fitting a similar role. Kind of like Headscale vs Tailscale. Is that right?
saving this comment for when i dive deeper down the rabbit hole later. at the moment i think im only self-hosting jellyfin and owntracks, and I have a couple llm models in LM studio but haven’t messed with them in a while.
Their ‘Research’ AI is pretty decent too. I can get to the same ends eventually, but I can’t read and digest 25 different pages and then launch another 10 follow-up searches in 15 seconds. It’s summarizing and not simply inferencing so the hallucination rate is acceptably low and it also cites the sources so you can click a footnote to fact-check any important details.
I self-host most things, but I can’t self host a search engine. I much prefer paying monthly than having to fight the eternal battle against tracking and ads.
You CAN self-host a search engine, look up searxng.
Yes it’s a metasearch engine, but so is Kagi for the most part, their own indexes don’t cover everything.
I have this on my projects lists, I’ve seen it recommended in a few places.
If I’m understanding the purpose (privacy filtering for searches), Kagi is a service fitting a similar role. Kind of like Headscale vs Tailscale. Is that right?
saving this comment for when i dive deeper down the rabbit hole later. at the moment i think im only self-hosting jellyfin and owntracks, and I have a couple llm models in LM studio but haven’t messed with them in a while.