caveman

To reduce token consumption in AI responses, you can use compression techniques that strip away conversational filler and redundant prose while preserving technical accuracy and code structure. This ensures long-context tasks use fewer output tokens without losing critical logic.

Compressed communication does not remove code blocks or error strings from output. The process specifically targets non-essential linguistic elements and pleasantries, maintaining full technical accuracy and code integrity throughout the response.

To start compressing AI model responses, activate the caveman mode by typing /caveman. This immediately initiates ultra-compressed responses that persist across the entire session until explicitly stopped.

Available compression intensity levels for minimizing output tokens include lite, full, ultra, and classical Chinese variants. These six distinct levels range from professional-tight to extreme classical Chinese compression to suit different needs.

You should not use ultra-compressed communication during security-sensitive or high-risk technical operations. The system implements automatic clarity triggers that revert to normal prose in these scenarios to ensure safe and clear execution.

Your chosen compression style will stay active for the entire session. It features context-aware persistence that maintains the selected compression level across multiple interactions until you explicitly stop the mode.