Does telling an LLM to "be concise" actually save you money? We measured it across 9 models. Compressing the output can save you money and keep accuracy, compressing the input prompt does not. [R]

LLMs are too verbose and with a black box model the only things you control are what goes in and how you tell it to write back. Yesterday Claude Code shipped a "concise output style" where Claude keeps things short. We already have a paper out about this! We tested both channels, shortening the inpu