I Got 9.9 Lower TTFT on a Real Android Phone by Reusing llama.cpp KV State

Local LLM inference has an expensive habit: It recomputes prefixes it has already seen. A system prompt. A reused RAG document.

Source: [Dev.to](https://dev.to/bossandboss/i-got-99x-lower-ttft-on-a-real-android-phone-by-reusing-llamacpp-kv-state-1ngi)