“The prevailing strategy among LangChain customers is to bet on the continued, significant decline of LLM inference costs over time.”