Coinbase CEO shares how to keep AI spend flat as token use soars

Coinbase CEO Brian Armstrong revealed how Coinbase halved its AI spend while token usage grew "exponentially". According to him, key tactics include defaulting to cheaper open-weight models like GLM 5.2 and Kimi 2.7 via an LLM gateway, AI-driven prompt preprocessing for optimal model selection, keeping context lean and raising cache hit rates from 5% to 60% in tools like LibreChat.