Companies slash AI costs 65% by making Claude and Codex talk like cavemen

Companies Deploy Caveman Plugin to Combat Soaring AI Costs

A growing number of companies are turning to an unconventional solution to tackle escalating AI costs: making their large language models communicate like cavemen. The tool, simply called 'caveman,' transforms the typically verbose responses from Claude, Codex, and Gemini into stripped-down, essential-only answers1

1

. Developers at OpenAI, Nvidia, and GitHub have adopted the caveman plugin, with a senior OpenAI employee contributing code to add support for OpenAI's Codex tool1

1

.

How the Caveman Plugin Reduces Token Usage

Creator Julius Brussee developed the tool in early April after noticing that excessive token consumption was driving up his bills while using Claude Code. "I made Caveman back in early April because I was using Claude Code heavily and noticed a lot of my token spend was going to unnecessary prose," Brussee explained, citing "pleasantries, hedging, transitions, and chatty language that does not really matter"2

2

. By implementing simplified language output, Brussee reports achieving a 65 percent reduction in token usage2

2

. Instead of receiving friendly, conversational responses designed to mimic human interaction, users get direct, no-nonsense instructions that accomplish the same task with far fewer tokens.

Rising Pressure Forces Industry to Optimize AI Spending

The adoption of cost-saving measures for AI reflects mounting financial pressure across the industry. In recent weeks, major AI firms have shifted from flat-rate subscription models to per-use subscription models, acknowledging the prohibitively expensive upkeep of these services2

2

. Consulting giant Accenture identified that much of the "soaring token spend" stems from mundane tasks like converting PDFs to presentations1

1

. The financial strain is real: OpenAI shuttered Sora, their video generator that was bleeding a million dollars a day, complicating what could have been a lucrative deal with Disney2

2

. None of the major firms have found a path to natural profitability, making tools that curb rising AI costs essential for survival.

What This Means for AI Development

The caveman approach marks a stark reversal from years of efforts to anthropomorphize AI and showcase human-like intelligence. Companies invested heavily in making large language models sound conversational and empathetic, only to discover that this verbose output significantly inflates operational expenses. The fact that a senior OpenAI employee actively contributed to a tool that reduces their own models' verbosity signals how seriously the industry takes the need to reduce AI model token usage1

1

. As AI adoption spreads across enterprises, expect more companies to prioritize efficiency over personality. The trend suggests that future AI development may focus on delivering value with minimal computational overhead rather than maximizing human-like interaction. Watch for similar optimization tools to emerge as the industry grapples with the reality that maintaining conversational AI at scale remains financially unsustainable without significant innovation in how these systems operate.