OpenAI says coding agents can handle scientific maintenance, rewrites, and new systems

OpenAI · x · 2026-07-29

OpenAI says coding agents are already helping scientists spend more time on research by taking over routine maintenance, targeted optimization, full redesigns, and even new system builds.

The accompanying case-study graphic spans eight projects and shows the spectrum of work agents can handle: build modernization, byte-level and heuristic optimization, TensorFlow-to-PyTorch migration, Rust rewrites, workflow redesign, and GPU-native reimplementation.

The key caveat is that humans still own the scientific process:

OpenAI frames the shift as a productivity gain, but not a replacement for scientific judgment or stewardship.

Related event: OpenAI Says Coding Agents Are Taking Over Repetitive Research Tasks(2 posts)→

Original post →

More from coding & agent

coding & agent channel →