Digital twins – virtual representations of real-world systems – hold immense potential for predicting behavior, understanding complex processes, and making better decisions across various fields. However, traditional digital twins often struggle to generalize to new situations, especially when data is scarce. They also need to adapt as the real-world systems they mimic evolve. This is where the concept of *hybrid digital twins* comes into play. These twins combine the best of both worlds: the explainability of physics-based models with the flexibility of AI. By incorporating known scientific equations, these hybrid models can make accurate predictions even with limited data. But building them has always been a complex, expert-driven process. Researchers have now developed a new technique called HDTwinGen, an AI-powered system that *automatically* designs and optimizes these hybrid digital twins. HDTwinGen uses large language models (LLMs), the technology behind tools like ChatGPT, to propose and refine the structure of the twin. Think of it as an AI scientist that experiments with different combinations of equations and machine learning components to find the perfect balance for the system being modeled. The results are impressive. In tests on complex systems, from simulating cancer treatment to predicting the spread of COVID-19, HDTwinGen created twins that outperformed existing models. These AI-generated twins were better at generalizing to new scenarios, learning from small datasets, and adapting to changes in the real-world system. This breakthrough could significantly accelerate the development and deployment of digital twins. Imagine AI designing personalized medical treatment plans based on a patient's digital twin or optimizing energy grids by simulating different scenarios. While this technology holds enormous promise, it's important to be mindful of its limitations. The effectiveness of HDTwinGen depends on the quality of the initial information provided by human experts and the underlying knowledge of the LLMs. Further research will focus on addressing these limitations, such as incorporating human feedback and expanding to a wider range of systems. As AI continues to advance, the future of digital twins looks brighter than ever.