--- title: Talking to Copilot like a caveman | daily.dev description: 'Caveman prompting' is a technique to reduce LLM token usage by instructing AI models to drop filler words, articles, and pleasantries in favor of short, blunt... canonical: https://daily.dev/posts/talking-to-copilot-like-a-caveman-u5qsbinrw twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Talking to Copilot like a caveman | daily.dev og:description: 'Caveman prompting' is a technique to reduce LLM token usage by instructing AI models to drop filler words, articles, and pleasantries in favor of short, blunt... og:url: https://daily.dev/posts/talking-to-copilot-like-a-caveman-u5qsbinrw og:image: https://api.daily.dev/og/posts/u5qsbinrw.png og:image:alt: Talking to Copilot like a caveman og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Talking to Copilot like a caveman **[The Art of Simplicity](https://daily.dev/sources/bartwullems)** · 3 min read · 1 upvotes · 0 comments ## Summary 'Caveman prompting' is a technique to reduce LLM token usage by instructing AI models to drop filler words, articles, and pleasantries in favor of short, blunt fragments. A GitHub skill called 'caveman' packages this into an installable plugin for Claude Code, Copilot, Cursor, Codex, and other agents via a single npx command. It offers three modes: lite (removes filler, keeps full sentences), default (drops articles, uses short synonyms), and ultra (bare fragments with abbreviations). Real-world testing showed 15–30% token reduction rather than the advertised 65%, with longer sessions benefiting more than short debugging loops. The skill overhead (~1–1.5k tokens) can offset savings in short call chains, so measuring actual savings before budgeting around it is advised. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How much does caveman prompting actually reduce LLM token usage in real coding sessions? Real-world testing across actual coding sessions showed token reductions of 15% to 30%, well below the widely cited 65% figure. Short back-and-forth debugging sessions barely benefited because the skill instruction itself adds roughly 1-1.5k tokens of overhead, while longer sessions with explanatory answers, code reviews, and architecture discussions came closer to the 30% mark. _Measuring real token savings before committing matters when budgeting agent workflows on daily.dev._ ### What is the caveman skill for Claude Code and Copilot and how do I install it? Caveman is an installable skill/plugin that instructs an AI agent to drop articles, filler, and pleasantries, answer in fragments, and use short synonyms, while leaving code, commands, and error strings untouched. It installs with the command npx skills add JuliusBrussee/caveman and works with Claude Code, Copilot, Cursor, Codex, Gemini, and other agents, with lite, full, and ultra intensity modes. _Developers comparing token-saving agent skills can track new prompting tools like this on daily.dev._ ## Similar posts on daily.dev - [Why use so many tokens when a few do the trick?](https://daily.dev/posts/why-use-so-many-tokens-when-a-few-do-the-trick--uvekb2vuw) · Product Hunt · 4 upvotes · 1 comments - [Headroom vs Caveman: A Practical Guide to Reducing AI Coding Costs](https://daily.dev/posts/headroom-vs-caveman-a-practical-guide-to-reducing-ai-coding-costs-q8vwfnm3b) · Syncfusion · 0 upvotes · 0 comments - [So Expensive, A Caveman Can Do It](https://daily.dev/posts/so-expensive-a-caveman-can-do-it-xbgdgzytq) · Hackaday · 4 upvotes · 0 comments - [A GitHub tinkerer teaches Claude to talk less, and that may matter more than it seems](https://daily.dev/posts/a-github-tinkerer-teaches-claude-to-talk-less-and-that-may-matter-more-than-it-seems-2b1ct9c22) · InfoWorld · 2 upvotes · 0 comments --- Tags: [#ai-coding](https://daily.dev/tags/ai-coding), [#prompt-engineering](https://daily.dev/tags/prompt-engineering) [View this post on daily.dev](https://daily.dev/posts/talking-to-copilot-like-a-caveman-u5qsbinrw) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Talking to Copilot like a caveman","url":"https://daily.dev/posts/talking-to-copilot-like-a-caveman-u5qsbinrw","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/talking-to-copilot-like-a-caveman-u5qsbinrw"},"datePublished":"2026-07-14T05:45:12.907Z","dateModified":"2026-09-14T08:19:26.305Z","description":"'Caveman prompting' is a technique to reduce LLM token usage by instructing AI models to drop filler words, articles, and pleasantries in favor of short, blunt...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/9edb9fe8774baf699a44a182be52c244?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/9edb9fe8774baf699a44a182be52c244?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"The Art of Simplicity","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"The Art of Simplicity","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/52b68c70636d44f0901ae7cd107c9438","url":"https://daily.dev/sources/bartwullems"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/talking-to-copilot-like-a-caveman-u5qsbinrw","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-coding,prompt-engineering","timeRequired":"PT3M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"The Art of Simplicity","item":"https://daily.dev/sources/bartwullems"},{"@type":"ListItem","position":3,"name":"Talking to Copilot like a caveman"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/talking-to-copilot-like-a-caveman-u5qsbinrw#faq","mainEntity":[{"@type":"Question","name":"How much does caveman prompting actually reduce LLM token usage in real coding sessions?","acceptedAnswer":{"@type":"Answer","text":"Real-world testing across actual coding sessions showed token reductions of 15% to 30%, well below the widely cited 65% figure. Short back-and-forth debugging sessions barely benefited because the skill instruction itself adds roughly 1-1.5k tokens of overhead, while longer sessions with explanatory answers, code reviews, and architecture discussions came closer to the 30% mark. Measuring real token savings before committing matters when budgeting agent workflows on daily.dev."}},{"@type":"Question","name":"What is the caveman skill for Claude Code and Copilot and how do I install it?","acceptedAnswer":{"@type":"Answer","text":"Caveman is an installable skill/plugin that instructs an AI agent to drop articles, filler, and pleasantries, answer in fragments, and use short synonyms, while leaving code, commands, and error strings untouched. It installs with the command npx skills add JuliusBrussee/caveman and works with Claude Code, Copilot, Cursor, Codex, Gemini, and other agents, with lite, full, and ultra intensity modes. Developers comparing token-saving agent skills can track new prompting tools like this on daily.dev."}}]} ```