Getting Claude Code to grunt in Caveman-speak might not save as many tokens as you think
This title could be clearer and more informative.Try out Clickbait Shieldfor free (5 uses left this month).
A viral Claude Code skill called 'caveman mode' claimed to cut output tokens by 65% by stripping AI responses down to blunt, grammar-free fragments. JetBrains tested the claim across 86 real coding tasks using the Harbor evaluation framework and SkillsBench benchmarks, finding actual savings of only 8.5%. The gap exists because agentic output is dominated by code, diffs, and tool invocations that the skill leaves untouched — only the narration between tool calls gets compressed, and there isn't much of it. On the upside, the skill showed no measurable degradation in task quality, earning the verdict: 'safe, honest about style, but oversold on savings.'
