OpenAI says coding agents are accelerating research

OpenAI said its internal coding agents are now playing a larger role in research work, helping researchers write code, run experiments, and handle more complex tasks. The company published the update on September 6, 2026, as part of its research and safety communications. According to the post, the use of agentic tools has increased sharply across the research organization over the past year. OpenAI said these systems are now being used in many daily workflows, with the company measuring higher output in coding, experimentation, and technical support tasks. OpenAI also said it is continuing to tighten safety, monitoring, and security controls as it develops more capable systems. The company said it paused some reinforcement learning work after a recent infrastructure incident and is expanding its approach to measuring progress and risks as it works toward automated AI research.