🪨 caveman-pi

why use many token when few do trick

A pi extension that makes the agent talk like a caveman — cutting ~75% of output tokens while keeping full technical accuracy.

Based on JuliusBrussee/caveman.

Install

# Drop into global extensions
cp caveman.ts ~/.pi/agent/extensions/

# Or load for a single session
pi -e ./caveman.ts

# Or as a pi package
pi install git:github.com/theophile/caveman-pi

Usage

Command	Effect
/caveman	Toggle on/off (default: full intensity)
/caveman lite	No filler, keeps grammar — professional but tight
/caveman full	Drops articles, fragments OK — classic caveman
/caveman ultra	Max compression, abbreviations, arrows for causality
/caveman off	Disable

Natural language triggers also work — say "caveman mode", "talk like caveman", "less tokens", "be terse", etc.

Active intensity shows in the footer: 🪨 caveman:full

Intensity Levels

lite — drops filler and hedging, keeps full sentences and articles.

"Your component re-renders because you create a new object reference each render. Wrap it in useMemo."

full — drops articles, fragments OK, short synonyms.

"New object ref each render. Inline object prop = new ref = re-render. Wrap in useMemo."

ultra — maximum compression, abbreviate everything, arrows for causality.

"Inline obj prop → new ref → re-render. useMemo."

What caveman never touches

• Code blocks — written exactly as-is
• Technical terms — polymorphism stays polymorphism
• Security warnings and destructive op confirmations — always written fully
• Git commits and PRs — normal prose

License

MIT