That's plausibly true. I've definitely seen absurd token costs abused and wasted. I've also seen simple problems require absurd token counts regardless of prompt quality. Which factors made this problem require 1000x fewer tokens than they used?