OpenAI has opened the curtains on its internal research operations, revealing that coding agents are now a central part of how the organisation works. The company describes these agents as reshaping the research environment, suggesting a shift from manual coding to automated assistance for experimental work.

The organisation shared early data on how these agents are being deployed, focusing on usage patterns, experiment velocity, and the complexity of tasks they can handle. By automating parts of the coding process, OpenAI claims to be able to move faster through research cycles, though the specific metrics and benchmarks used in the data are not detailed in the public post.