Inference systems notes for practical AI teams

Technical essays on TTFT, cache design, model serving, hardware trends, low precision, and the operational shape of frontier inference APIs.