token-optimization

The hard half of context engineering is subtraction: deciding what to keep out of your AI's context, not just what to put in.