Cole Seifert(@colepseifert):I thought long context was what made agents slow. At @tetrateio's Inference Routing Summit, Sven Mawson of @SambaNovaAI said agent sessions hit cache 90-95% of the time, so 80-95% of the time goes to decode. The slow part is writing the answer, not reading the prompt.

Cole Seifert

@colepseifert

Building production LLM systems @ Accenture. Agents, RAG, evals, context engineering. Claude + Claude Code. SF AI meetups most nights. Notes from the room.

参加 September 2025

33 フォロー中 11 ファン

Cole Seifert@colepseifert

2026.11.05 17:30

I thought long context was what made agents slow. At @tetrateio's Inference Routing Summit, Sven Mawson of @SambaNovaAI said agent sessions hit cache 90-95% of the time, so 80-95% of the time goes to decode. The slow part is writing the answer, not reading the prompt.