①OpenAI released an article titled “Research Acceleration: OpenAI’s Internal Perspective”, publicly presenting RSI progress in internal data for the first time; ②OpenAI expects to achieve complete “automated AI researchers” by March 2028; ③Team member Kevin Liu noted on social media platforms: “RSI is likely to be the core driving force behind the leap in AI capabilities in the next few years.”

On September 7, 2026, OpenAI released an internal article titled “Research Acceleration: An OpenAI Perspective,” in which progress on Recursive Self-Improvement (RSI) was disclosed for the first time. The team expects to achieve a complete “automated AI researcher” by March 2028; this goal has been reached, and by September of this year, they will have developed “automated research interns” capable of performing specific research tasks. Team member Kevin Liu noted that RSI may be the core driving force behind the future leap in AI capabilities. Data shows that the number of internal OpenAI researchers using agents has increased significantly, with median users consuming over $600 per day in terms of token value, and high-frequency users over $7,000. As automation progresses, research tasks that are most difficult to replace will become new bottlenecks, further highlighting the importance of computing power.