Show HN: TurboQuant-WASM – Google's vector quantization in the browser
SIMD vector compression — 3 bits/dim with fast dot product
Our Take
{"problem_it_solves": "Float32 embedding indexes are too large for mobile RAM, take minutes to download, and gzip only saves ~7% due to high entropy", "target_customer": "Developers building browser/edge applications needing vector compression without training overhead", "use_cases": ["Vector search", "Image similarity", "3D Gaussian Splatting compression", "LLM KV cache compression", "Real-time indexing", "Browser and edge deployment"], "pricing_details": "Open source (MIT license)", "differentiator": "No training step required - unlike PQ/OPQ, just init(dim, seed) and encode any vector immediately. Each vector is self-contained for streaming data.", "why_now": "Browser support for relaxed SIMD is now available (Chrome 114+, Firefox 128+, Safari 18+, Node 20+)", "traction": {"notable_metrics": "211 stars, 7 forks, 96 commits"}}
Key Facts
The people behind Show HN: TurboQuant-WASM – Google's vector quantization in the browser
Links
Want products like this in your inbox every morning?
Five products. Every morning. Written by someone who actually cares whether they're good or not. Free forever, unsubscribe whenever.