A few days ago, OpenAI released GPT-6 Astra, which claims to achieve AGI, sparking widespread discussions on recursive self-improvement architectures and drawing significant attention to AI's involvement in next-generation AI research and development. Recently, OpenAI disclosed detailed data on the accelerated progress of AI models in internal scientific research and announced that it has achieved the goal of an 'automated AI research intern,' a system capable of completing tasks that would take skilled researchers days to accomplish under human supervision. According to the data, as of mid-August 2026, for every human workday in OpenAI's research department, 3.1 AI Agent workdays were operating concurrently, and the experimental speed of each active researcher increased to 1.6 times the 2025 baseline. However, OpenAI also noted that the operating hours of AI Agents do not equate to equivalent productivity, as complex tasks still require significant human intervention, with over half of successful cases requiring at least one manual intervention. OpenAI plans to establish an 'automated AI researcher' system by March 2028, enabling AI to independently solve complex scientific research problems.