[nucliweb]

Stacked on #57 (it edits the lesson text #57 rewrites); retargets to main once #57 merges.

What

JPEG XL vs AVIF for web use-cases (Jake Archibald, 2026) uses the same method as this tool (equal SSIMULACRA2 targets), so it is a good outside reference for three things a reader needs to know before trusting a ranking:

• Rate-distortion mode: on cid22-friends.png, AVIF needs the fewest bits from 70 to 85 and JPEG XL is smaller from 90 up, so runs at 70 and at the default 90 name different winners. His photographs show the same pattern ("JPEG XL catches up at very high quality").
• Matched effort: effort moves sizes too, not only times, and unequally across codecs; he compares at maximum effort and finds "libjxl's higher efforts are much slower, use a lot more memory, and barely improve the result". Results at the defaults compare the presets most tools ship with.
• Scope and assumptions: ssimulacra2 is "a perceptual image quality metric by the folks behind JPEG XL"; it was tuned on independent subjective data (CID22, TID2013, KADID-10k, KonFiG-IQA), and dssim does not drive the search, so ColorVideoVDP is named as an independent cross-check where JPEG XL and another format are close.
• SVT-AV1 and JPEG lessons: cite "AVIF's 4:2:0 mode (half colour resolution) shouldn't be your default", which the 4:2:0 measurements in feat: replace the Kodak demo photographs with three from CID22 #57 back up.

Quotes checked against the post's text.