Show HN: Turboquant.cpp – Quantize embeddings to 1-4 bits, no training (400 LoC)github.com/RunEdgeAI2 pointsandrewmikhail7 days ago