a-french-anon

You don't read Lisp via the parentheses, but via the indentation. It's almost like Python but the (s and )s are for the computer to see.

Tip from a Frenchman: Camembert is an overglorified version of Brie. The one you need to see is where the "king of cheese" is made: Roquefort-sur-Soulzon.

I think it's saner to be literal here. Reference counting is a way to automatically collect garbage for you. Why shouldn't it be called a GC then?

Electric Miata when?

This doesn't look like it has wired support?

Do you just want to use "sets of haplogroups with similar phenotype" instead? I mean, you can argue all day that the existence of the orange color somehow means that there's no difference between yellow and red, but you're just going to look foolish.

Which is kinda ironic because Toyota is probably the only one that did hybrid from the ground up with their HSD transmission.

For decoding, yes, dav1d. But for encoding, a lot more SIMD kernels in libaom/SVT-AV1 written by other people.

(And lots of kernel shared between decoding and encoding, of course)

The third for me is that LLM usage has exploded for the development of core Julia itself (just look at the PRs).

Yes, but for speed which means a lot of SIMD plumbing, the larger workforce certainly benefited AV1 codecs!

Sorry for the dumb question, but I don't understand something in these graphs. Since it's x=speed and y=bd-r, is this test done at a fixed target quality/size?

What do you think about the fact that most perceptual quality improvements in SVT-AV1 have been almost only driven by external, non-paid enthusiasts such as yourself while the maintainers have been basically only targeting VoD and RTC?

Compared to JXL (which admittedly has almost nobody working on except veluca these days...) which was focused on it since the beginning, I think it really matters.