Plain colored shapes in prompts shift AI models' coding choices, DeepMind and Stanford find

1. 11:14 AM
   Rohan Paul@rohanpaul_aiX

   Google DeepMind and Stanford find simple shapes in prompts steer AI coding choices

   AIA Google DeepMind and Stanford paper reports that adding a plain colored shape, such as a red heart, to a coding prompt changed how models solved a problem. Across 288 shapes, the study tracked how often Claude Opus 4.7, Gemini 3.1 Pro, and GPT-5.4 chose recursion. A red heart moved Claude Opus 4.7 and GPT-5.4 from 0% to 100% recursive solutions on a binary tree problem, and 71.5% of Gemini 3.1 Pro's thinking traces that were steered called the image irrelevant.

   

The trend compares only continuously observed participants, so its scope may be smaller than the current heat total. Hover or tap for hourly heat; use arrow keys with a keyboard.