Perfecting anti-aliasing on signed distance functions

This looks like a really nice roundup and also kudos for collecting the small code snippets at the end.

Using Oklab in anti-aliasing is something I haven’t encountered before. I didn’t digest this article yet in full but I would’ve liked to see how much difference it makes in color images. For the black & white examples the sRGB gamma curve would’ve been a better-motivated choice (Oklab does a cube root), since that’s what the pictures will be displayed on. But I’m open to the idea of AA being better in a perceptual color space, and really appreciate the empirical comparisons between the alternatives.