nx_webp_decode.nx

nx_webp_decode.nx -- SOVEREIGN WebP decoder, built from the first byte up (operator 2026-07-29: "build all the missing codecs... anything WebKit does and beyond"). WHY IT MATTERS: Wikimedia CONTENT-NEGOTIATES and serves WebP under .png URLs, so with PNG/JPEG/GIF/ BMP only, EVERY image on a wikipedia article decoded to nothing (measured: nsrc=11 spawn=6 dec=0). WebP is not optional in 2026 — it is the default web image format. LAYER 1 (this file, LANDED): RIFF/WEBP container + VP8L header + chunk routing. Exposes exact dimensions and the bitstream span, so the pipeline can size boxes and the KAT can prove the parse. LAYER 2 (next): VP8L LOSSLESS pixel decode -- LSB-first bit reader, canonical prefix (Huffman) code build, meta-prefix groups, colour cache, and the 4 transforms (predictor / colour / subtract-green / colour-indexing). That is what these thumbnails use. LAYER 3: VP8 LOSSY (boolean/arith decoder, DCT, prediction, loop filter) then VP8X/ALPH + animation. ★NOTHING HERE FABRICATES PIXELS: until layer 2 lands, decode_rgb reports NOT-YET (0), never a plausible-looking buffer -- a codec that guesses is worse than one that declines.