Notes on recommendation systems, information extraction, and the eighty percent of the work that is not training the model.https://alapin.fyihttps://alapin.fyi/writing/how-much-precision-does-intelligence-needhttps://alapin.fyi/writing/how-much-precision-does-intelligence-needA visual introduction to LLM quantization, from floating-point numbers to block scales, effective bits per weight and llama.cpp.Sun, 02 Aug 2026 00:00:00 GMTQuantizationMachine learningLLMsllama.cpp