OpenAI devs go caveman to cut AI token costs by 65%

developers are deploying a 'caveman' plugin to combat skyrocketing ai costs by stripping pleasantries and prose from model outputs while preserving code and technical details. the tool reduces token usage by 65-75%, leading employees at openai and corporate giants like legrand to adopt it to stay within strict new budget quotas. a senior openai engineer even contributed code to support the project, signaling internal validation of the method. this trend highlights a growing tension between the expensive, verbose nature of frontier models and the enterprise need for cost-efficient, deterministic tooling.