Seeing the model use "caveman speak" in its thoughts to save tokens is hilarious. "Why use many word when few word do trick?" is actually a legit tech optimiza

Seeing the model use "caveman speak" in its thoughts to save tokens is hilarious. "Why use many word when few word do trick?" is actually a legit tech optimization now!