Caveman prompt cuts agent output tokens 44% and call cost 33% in one test
gethackteam · x · 2026-10-24
The viral caveman project (115K stars on GitHub) makes coding agents talk in minimal "caveman" style, with a proxy that cuts token usage up to 65%. In a same-model, same-task test, adding one instruction — "Minimum words. Strip filler." — dropped output tokens from 1,174 to 516, cutting estimated whole-call cost by 33%. Available as a skill or proxy for agents like Claude Code.
Related event: 'Caveman' Prompt Trick Cuts AI Token Costs by Up to 65%(2 posts)→
More from coding & agent
- PL researcher admits Opus 5.5 is already a better programmer than him — tweetsatpreet · 2026-10-25
- Free LLM Dojo launches: 50+ interactive labs and 490 AI-engineering interview questions — sharpeye_wnl · 2026-10-25
- mlx-decision runs decision models locally on Apple Silicon, up to 6x faster than PyTorch — armbues · 2026-10-25
- How do teams share memory across multiple AI agents? — belhelha · 2026-10-25
- Vibe Monitor ships after 41 beta iterations: local stdio MCP adds architecture checkpoints to coding agents — Either-Alternative32 · 2026-10-25
- KNOWS benchmark: best computer-use agent fully solves just 3.6% of tasks — anmarasovic · 2026-10-25