JuliusBrussee/caveman
View on GitHub ↗ ★ 116,834 Go
🪨 why use many token when few token do trick. Viral skill + proxy for coding agents that cuts 65% of tokens by talking like a caveman.
29 mentions on Hacker News by 29 people: 4 stories, 25 comments · first 2026-04-05 · latest 2026-10-29
20082026
-
…without any cringy flowery language that LLMs are known for. In fact, there's even a skill for it: https://github.com/juliusbrussee/caveman
comment by d3Xt3r · 2026-10-29 · in Ask HN: Should I be polite to my agent or should I treat it like shit? -
caveman mode good https://github.com/JuliusBrussee/caveman
comment by freeplay · 2026-09-16 · in PS5 Linux lead quits: "a bunch of noobs using LLMs" that "they don't understand" -
▲ 2
Respond terse like smart caveman. All technical substance stay. Only fluff diestory by Bluestein · 2026-09-10 · 0 comments
-
Same. I use this to fix claudespeak. Any hypothetical token savings are just gravy. https://github.com/JuliusBrussee/caveman/blob/main/skills/ca...
comment by eleventen · 2026-09-10 · in DeepSeek v4.1 Flash -
There is also Caveman[1] skill which rewrite prompts. 1. https://github.com/JuliusBrussee/caveman/
comment by nerdyadventurer · 2026-09-07 · in Portal by Spotify cut my Claude Code token usage by 90% -
…experience vastly better. I'm already a programmer, Claude shouldn't need to tell me like I'm 5. https://github.com/JuliusBrussee/caveman
comment by ARandomerDude · 2026-08-20 · in Vomit: Clean up Claude 5's token output with a separate LLM -
https://github.com/juliusbrussee/caveman
comment by mycall · 2026-08-15 · in Qwen 3.8 27B -
how about stuff likes https://github.com/JuliusBrussee/caveman is it also make the model less capable?
comment by firemelt · 2026-08-14 · in Why does Opus 5 feel worse to work with? -
This is what you want: https://github.com/juliusbrussee/caveman
comment by ApolloRising · 2026-08-12 · in Ask HN: Anyone have solution to Opus verbosity in Claude Code? -
Have you tried using the caveman skill ? :D Might be worth a try if you dont like long prose https://github.com/JuliusBrussee/caveman
comment by smnplk · 2026-08-09 · in How I use LLMs to learn complex topics -
Neat! But I still lean towards https://github.com/juliusbrussee/caveman .
comment by stefanfisk · 2026-07-25 · in A system prompt to get AI to stop pretending to be human -
…just cut down on overuse of specific terms; it cuts down on time spent digesting slop in general. https://github.com/JuliusBrussee/caveman
comment by bunderbunder · 2026-07-14 · in How to stop Claude from saying load-bearing -
…to answer like a cavemen, if llm talk like cavemen, the answers become shorter and more compressed. https://github.com/JuliusBrussee/caveman It's for getting it to output shorter answers, but also could help with your burnout.
comment by hoppp · 2026-07-09 · in I think I have LLM burnout -
…silently and suddenly you're paying latest Opus costs for 6 months old Sonnet performance. [1]: https://github.com/JuliusBrussee/caveman [2]: https://github.com/rtk-ai/rtk [3]: https://github.com/DietrichGebert/ponytail
comment by Zababa · 2026-06-19 · in The Token Compression Illusion: Why I'm Skeptical of RTK -
Yeah, it’s why the Caveman skill includes a Wenyan mode. https://github.com/JuliusBrussee/caveman
comment by grogg · 2026-06-18 · in DeepSeek Introduces Vision -
I want to see more operators try https://github.com/juliusbrussee/caveman How does it affect agent accuracy?
comment by armchairhacker · 2026-06-12 · in AI agent bankrupted their operator while trying to scan DN42 -
…configure model to be terse/concise with output style ? There are plenty of popular projects like https://github.com/JuliusBrussee/caveman which do it for you even. Input/Cache/Output ratios are use case and configuration dependent . Any…
comment by manquer · 2026-05-28 · in Anthropic raises $65B in Series H funding at $965B post-money valuation -
https://github.com/JuliusBrussee/caveman
comment by pas · 2026-05-28 · in I think Anthropic and OpenAI have found product-market fit -
Absolutely. I use caveman to help with that: https://github.com/JuliusBrussee/caveman
comment by anderber · 2026-05-03 · in Kimi K2.6 just beat Claude, GPT-5.5, and Gemini in a coding challenge -
▲ 2
Local SLM as a compression layer for cloud API callsstory by sideras56 · 2026-04-25 · 0 comments
-
If you’re enjoying this thread, you have to make a stop over here: https://github.com/JuliusBrussee/caveman
comment by DANmode · 2026-04-20 · in Claude Token Counter, now with model comparisons -
And what about with Caveman[1]? 1. https://github.com/juliusbrussee/caveman
comment by xd1936 · 2026-04-17 · in Measuring Claude 4.7's tokenizer costs -
…now, since there are not many workarounds to minimize token usage unless you use caveman [1]. [1]: https://github.com/JuliusBrussee/caveman
comment by XzAeRosho · 2026-04-17 · in Claude Opus 4.7 -
▲ 2
Caveman – why use many token when few do trickstory by kerblang · 2026-04-16 · 0 comments
-
…relevant by the day. I already enjoy reading its output more than vanilla so suits me well. [0] https://github.com/JuliusBrussee/caveman/tree/main
comment by cupofjoakim · 2026-04-16 · in Claude Opus 4.7 -
Just saying, I simply copied from https://github.com/JuliusBrussee/caveman Beyond the cost of tokens, I found it's very good at doing more with a limited context window before OpenCode hits context compaction.
comment by duguyue100 · 2026-04-12 · in Midnight Captain – A midnight commander inspired file manager -
Can't you use Claude caveman mode? https://github.com/JuliusBrussee/caveman
comment by storus · 2026-04-05 · in Running Gemma 4 locally with LM Studio's new headless CLI and Claude Code -
▲ 904
Caveman: Why use many token when few token do trickstory by tosh · 2026-04-05 · 366 comments