The field of system performance prediction and security is shifting towards text-based modeling, leveraging large language models (LLMs) to analyze complex system data such as configuration files, system logs, and user reports. This approach has shown promise in addressing challenging tasks like predicting resource efficiency, diagnosing hardware faults, and detecting anomalies in system logs. Innovative techniques, including text-to-text regression and decoder-only architectures, have been proposed to overcome the limitations of traditional methods. These advancements enable more accurate and efficient modeling of real-world outcomes, paving the way for universal simulators and scalable solutions for specialized domains like automotive communication systems and IoT security. Notable papers include: Performance Prediction for Large Systems via Text-to-Text Regression, which achieves near-perfect rank correlation and 100x lower MSE than tabular approaches. Evaluating LLMs and Prompting Strategies for Automated Hardware Diagnosis from Textual User-Reports, which evaluates 27 open-source models and identifies the best balance between size and performance for efficient inference on end-user devices.