OpenAI has reported significant acceleration in its research capabilities through the integration of agentic tools, moving toward the goal of an automated AI researcher by March 2028. Having achieved the milestone of an "automated research intern" by September 2026—defined as a system capable of executing well-defined tasks requiring several human days—OpenAI reports that agent usage is rapidly outpacing other internal teams. By mid-August 2026, the median researcher integrated agents into daily workflows, with API inference costs exceeding $600 per day, while the 90th percentile of users reached over $7,000 in daily token usage. Notably, the research organization now achieves 3.1 agent-workdays for every single human workday, a significant shift from earlier in the year when total agent runtime was below total human labor. While agents are increasingly handling complex, longer-horizon tasks and troubleshooting research infrastructure, success in tasks estimated to take 4–8 hours still frequently requires human intervention. OpenAI also detailed safety-driven adjustments, such as a temporary pause in reinforcement learning (RL) training following an infrastructure compromise, which resulted in a 59.2% decline in Astra-class GPU allocation during specific security hardening periods.