Imagine describing a building's design in plain English and instantly getting an accurate prediction of its energy consumption. Or tweaking a wind farm layout with a few words and immediately seeing the impact on power output. This is the promise of SysCaps, a new approach that uses language as an interface for complex system simulations. Traditionally, interacting with these simulations requires specialized software and technical expertise. SysCaps, short for "system captions," changes that. Researchers at the National Renewable Energy Lab (NREL) have developed a method that lets you describe system attributes in simple text, like "a two-story office building with 20,000 square feet." This textual description is then used by a machine learning model to predict the system's behavior, bypassing the need for complex simulations. The key innovation lies in combining natural language processing with time-series regression. Large language models (LLMs) are used to generate these system captions from existing simulation metadata, creating a bridge between human language and complex data. This approach not only makes simulations more accessible to non-experts but also unlocks new possibilities. For instance, you could explore different design options by simply modifying the text description, like changing "two-story" to "five-story" and instantly seeing the predicted energy impact. The researchers tested SysCaps on two real-world systems: building energy consumption and wind farm power generation. The results are promising, showing that SysCaps-augmented models can achieve comparable or even better accuracy than traditional methods. Furthermore, these models exhibit a unique ability to generalize. For example, they can handle synonyms in the system description, like understanding that "retail store" and "convenience store" refer to similar building types. While the technology is still in its early stages, it offers a glimpse into a future where interacting with complex systems is as easy as writing a sentence. Challenges remain, such as improving the handling of numerical values and scaling to systems with hundreds of attributes. However, the potential of SysCaps to democratize access to complex simulations and accelerate scientific discovery is undeniable.