I find it difficult to wrap my head around how Privacy Pass actually works. It seems pretty involved. Kagi‘s implementor actually explained it over on hackernews, but that‘s only a summary.
Since it‘s an RFC standard, I don’t doubt that it works, but I hope someone smarter than me actually checks Kagi‘s open-sourced client code, which should be enough to ensure anonymity, according to the standard.
I‘ve been using Kagi since September 2023.
The biggest difference to other search engines is really the quality of the results. Before Kagi, I‘ve used DuckDuckGo for a few years, but had to switch to Google more often than not, as I just didn‘t find what I was looking for. Most alternative search engines only use Bing‘s index, which just isn‘t any good in my experience. Kagi‘s results are built from quite a few different sources as well as their own index. Seeing how crappy Google‘s results are at this point, I don‘t think you can get better search results right now.
Also, the ability to rank and block specific domains in my results is something I really like and wouldn‘t want to miss anymore.
Lastly, I really enjoy using a search engine that isn‘t being optimized for advertisers, but for the user. Kagi‘s team is very receptive for feedback. For example, they’ve implemented an icon for search results with paywalls, because users asked for it. There are many small, nice details like that, which can save you a few clicks or just improve your general experience.
I unfortunately can’t really see how a browser could still be nice to use and properly resist fingerprinting.
The site https://amiunique.org/fingerprint tries to fingerprint your browser and lists the used attributes along with their uniqueness within their dataset. And while a browser could pretty reliably lie about its User Agent or Platform, it’s often just necessary for a modern website to know, for example, what your view-port’s resolution is or what kind of audio/video codecs your device supports. Going through my own results, I’d say combining these necessary data points is probably enough to identify me, even though I’m pretty privacy-conscious.
Maybe I’m overly pessimistic, but I think preventing fingerprinting would need a regulatory instead of a technical solution. Unfortunately that doesn’t seem very likely anytime soon.