Open-source 'caveman' plugin slashes LLM token costs by up to 65%

A free open-source plugin called "caveman" is gaining traction for its ability to drastically reduce token consumption in large language models, particularly for coding tasks. Developed by Julius Brussee, the plugin forces models like Claude Code, Gemini CLI, and GitHub Copilot to adopt a terse, telegraphic style, eliminating conversational filler and articles. Initial tests suggest savings of up to 65% in output tokens, translating to significant cost reductions, especially since output tokens are often more expensive than input tokens. AI

IMPACT This tool could significantly lower operational costs for AI users by reducing token consumption, especially for coding and reasoning tasks.

RANK_REASON The cluster discusses a plugin that modifies the output of existing LLMs to reduce token usage, which is a tool-level improvement rather than a new model release or core research.

Read on dev.to — LLM tag →

• 3DNews
• 404 Media
• Anthropic
• caveman
• Claude Code
• Gemini CLI
• GitHub
• GitHub Copilot
• Julius Brussee
• Nvidia
• OpenAI
• Shayne Sweeney

• product
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →