Compact answers and deliberate context selection for ChatGPT and Codex.
Reduce token use when the user asks for compact answers or lean context.