⬢github Go · 121.5K ★ +12.7K since we first saw it · pushed 15 h ago · Apache-2.0
JuliusBrussee/caveman
why use many token when few token do trick. Skill + local proxy for coding agents: shorter answers, smaller tool output, every original recoverable.
Caveman is a prompt skill plus proxy for AI coding agents that makes them write tersely, in caveman-style speech, cutting token usage by roughly 65% and lowering LLM API costs. It installs with one command, works with 30+ agents, and ships as an npm/PyPI middleware you can also embed in your own apps.
Why now: It went viral — #1 on GitHub Trending and Hacker News — and was backed by an Adobe Research paper and JetBrains testing showing big cost savings with no measurable quality loss.
Who it is for: Developers using AI coding agents like Claude Code who want to cut token bills without changing tools.
aianthropiccavemanclaudeclaude-codellmmemeprompt-engineeringskilltokens
Stars over our 694 snapshots: 113.9K to 121.5K, since 29 d ago.
Where people talked about it
- ⬢github JuliusBrussee/caveman 4 d ago
API: https://socialmediatrends-api.osmike.com/v1/repos/JuliusBrussee/caveman