As we reported last year, users saying “please” and “thank you” to AI tools is wasting millions in computing power. In an attempt to curb AI’s soaring costs, 404 Media has reported that developers at Nvidia, OpenAI, and GitHub are using a caveman plugin to make their AI tools speak like cavemen, which can cut Claude Code token consumption by ~75%. Users can pick their level of “grunt”: lite, full (which is the default setting), ultra, or Wenyan, which translates the output into classical Chinese characters for extreme compression.

(Maybe this is the magic, unidentified method that OpenAI engineers claim has cut inference costs in half?)

The S&P 500, Nasdaq 100, and Russell 2000 all gained yesterday, led by tech stocks. Stocks posted their best quarter in six years, and semiconductors had their best quarter ever.