Show HN: TurboQuant-WASM – Google's vector quantization in the browser

This tackles embedding bloat in a way that actually matters — the kind of thing that makes a 1.5GB index basically unusable on mobile. TeamChong squeezed float32 vectors 6x using WASM and Google's quantization trick, and the real flex is skipping the training entirely. Most quantization approaches need hours of preprocessing, but this just wants a dim and a seed, then you're ready to encode. Direct dot product on compressed data without decoding is the move here, and relaxed SIMD support in modern browsers finally makes it viable. 211 stars in a few weeks isn't bad signal for something this niche.