OpenAI reports how coding agents are changing research work
OpenAI reports growing use of coding agents across its research organization, alongside more code contributions and experiments. Its analysis also finds that complex tasks still need substantial human steering. The company cautions that these preliminary activity measures do not directly establish the overall pace of scientific progress, and that increased compute is another factor.
Context & source check
For teams adopting agents, track completed, reviewed outcomes alongside usage and speed. More automated activity is useful only when it clears a real bottleneck without weakening quality or oversight.
Company research report. Company source. Source checked Sep 7, 2026. Not independently tested.
