Imagine a power outage after a major storm. Instead of waiting days for the lights to come back on, AI could jump in to restore power in mere minutes. That's the promise of a groundbreaking new approach explored by researchers, using the power of Large Language Models (LLMs) like the ones behind ChatGPT to revolutionize how we handle power grid disasters. Traditionally, getting the grid back online after a disruption is a complex puzzle. Operators have to manually figure out the best way to reroute power, balancing supply and demand while avoiding further damage. This is where AI could offer a dramatic improvement. This new research introduces the “Physics-Informed Decision Transformer” (PIDT), an AI system that learns the physics of power grids and uses an LLM to make lightning-fast decisions about restoring power flow. Think of it as an expert operator with instant access to all the information and the ability to make the right choices at incredible speed. In simulations, PIDT showed its potential. Compared to existing AI methods, it was able to restore power more effectively in a model of a 123-bus power distribution network – a simplified representation of a real-world system. While still in its early stages, this research suggests a future where AI helps bring power back online rapidly after natural disasters or cyberattacks, minimizing disruption and improving grid resilience. This is just the beginning, and researchers are already working on enhancing this model to handle larger, more realistic grid scenarios and increase reliability. The potential is huge – faster recovery, reduced economic impact, and increased safety, all thanks to the power of AI working hand-in-hand with human expertise.