I personally was wrong in that I thought AVIF would be the last traditional codec, and future progress would come from ML based compression. But JPEGXL is very impressive. I do think we are close to the end of the road for this approach though, as neural net based compression appears to outperform by many multiplies. I expext everything (major) after JPEGXL to be ML based.

it is tragic they do not use SVT-AV1 https://github.com/AOMediaCodec/SVT-AV1 that makes AV1 dead on arrival