ICLR 2025 workshop working building AI to answer a single question: Given low-level theory and computationally-expensive simulation code, how can we model complex systems on a useful time scale?