A new study by researchers from the University of Texas at Austin, Texas A&M University, and Purdue University has found that large language models (LLMs) suffer cognitive decline when trained on viral social media content. The research indicates that models exposed to 100% viral data experience a significant drop in reasoning accuracy and long-context comprehension, a phenomenon termed 'LLM brain rot.' This degradation includes thought skipping and increased factual errors, with effects persisting even after retraining on clean data. The study highlights concerns that engagement-driven content may be altering AI cognition similarly to how social media affects human attention.