I know it's frowned upon to say this at HN, but I'm beginning to wonder who's reading the actual page. At least to me the page seems pretty clear about it:

>The xor filters takes a bit longer to build, but once built, it uses less memory and is about 25% faster in some demonstration test.

When they say "faster," they mean for querying. Which is actually pretty useful for a lot of use cases where bloom filters are used today.

To be clear, I wrote specifically "construction performance".

The blog already stated that "build is bit slower", so i was mainly commenting that instead of "bit slower" it is actually order of magnitude slower (in some cases).

According to the paper, the query side performance is indeed faster, so they are not misleading in any way.