Scientific Computing in the Age of Agentic AI
OpenAI, Tuesday, July 28th, 2026
A field report shows scientists using AI coding agents to modernize legacy scientific computing codebases.
OpenAI published a field report on how scientists are using AI coding agents to modernize scientific computing.
Much of the field runs on decades-old Fortran and C code that few current researchers can safely modify, which has become a genuine constraint on research velocity.
The report covers work in genomics and other domains where agents translated, tested, and accelerated legacy code. It addresses what the agents handled well and where researcher oversight remained essential. The findings are drawn from observed practice rather than controlled evaluation.