Top AI Product

Your coding agent burns tokens saying “I’d be happy to help!” before every answer. Caveman kills that. It’s a skill/plugin — not an app, not an API — that you drop into Claude Code and 30+ other agents, and it rewrites how the model talks: short fragments, no filler, no hedging. Same answer, 65% fewer output tokens.

What it actually does

The trick is dumb and effective. Caveman only touches natural-language explanation — the “let me walk you through this” fluff. Code, commands, file paths, error messages, git text all stay byte-for-byte intact. Six intensity levels ship, from lite to ultra, plus a wenyan classical-Chinese mode. There’s /caveman-compress for shrinking memory files (~46% input savings) and /caveman-stats to tally what you’ve saved. Runs fully local, zero telemetry.

Why it’s blowing up

Nearly 3,000 stars in a day, 83k total — one of GitHub’s fastest-climbing repos right now. It hits the one nerve every agent user shares: token bills. Fair warning — savings are output-only, and the skill adds ~1–1.5k input tokens per turn, so already-terse sessions can go net-negative.

You Might Also Like

• Caveman Scores 333 hn Points for Making Claude Talk Like a Caveman Does it Actually Save 75 of Tokens
• Microsoft mai Code 1 Flash is Live in Github Copilot 60 Fewer Tokens Than Comparable Coding Models
• Openviking Treats ai Agent Memory Like a File System and 9k Github Stars say its Working
• 27k Github Stars in Weeks Learn Claude Code by Shareai lab Breaks Down ai Coding Agents Into 12 Lessons
• Addy Osmani Open Sources Agent Skills 19 Workflows That Make ai Agents Code Like Google Engineers