I benchmarked caveman against two words

I benchmarked caveman, a popular Claude Code compression plugin, against the prompt "be brief." across 24 dev questions. Full write-up with per-category breakdowns and the bits I cut from the video: https://www.maxtaylor.me/articles/i-benchmarked-caveman-against-two-words Benchmark harness (open source, strategy-agnostic): https://github.com/max-taylor/cc-compression-bench Caveman plugin: https://github.com/olafgeibig/caveman Chapters 00:00 Intro 00:08 Testing setup 00:27 Quality results 00:36 Token usage results 01:16 Caveman features 01:59 Outro