University of Arizona

Solution

At the University of Arizona, computational astrophysicist Chi-kwan Chan uses Codex to accelerate the code and algorithm work behind black hole simulations for the Event Horizon Telescope effort. The scientific workflow starts with telescope observations and the need to model how plasma behaves in extreme environments, but the underlying equations are so complex that even supercomputers struggle and new approximations can take years to invent. Chan uses Codex to generate coordinate transformations, numerical algorithms, and supporting scientific code through an agent skill he wrote, then implements and verifies those outputs before comparing simulation results with observational data. Codex also reviews code and helps catch mistakes in statistics and numerical methods, making the simulation loop both faster and more reliable. In effect, Codex sits between astrophysics questions and software production, letting Chan spend more of his time on research instead of manual coding.

Data flow

Observational goals and simulation problems from the Event Horizon Telescope workflow are turned into coding tasks for Codex, which proposes algorithms and code that Chan validates and runs before comparing the resulting simulations against telescope observations.