This video shows my Tetris test of Jev, TypeSafe AI’s model for structured decisions. Here, I use its Choice API to select actions based on the game state. The API returns the selected action along with probabilities for the available choices.

Download the video (.mp4).

Jev reached 96.47% accuracy in my IMDb movie review classification test. I discuss that result and this Tetris example in Language Models for Text Classification: From Bag-of-Words to Jev. The CLM and Laya recordings show my tests of two open-weight alternatives.