OpenAI Says It Reached Its Automated Research Intern Goal

OpenAI says it has reached its goal of creating an automated research intern, based on measurements detailed in a company blog post. It defines the system as one that completes well defined research tasks under human direction, including work that could take a skilled researcher several days.

By mid August, OpenAI's research organization was using 3.1 agent workdays for every workday of human labor, based on an eight hour workday. The median researcher used coding agents daily and consumed more than $600 of inference per day at API prices, while the 90th percentile user consumed more than $7,000 of tokens per day. Experiments per active researcher also reached their highest level since tracking began in January 2025.

The company frames the disclosure as a snapshot of its progress toward recursive self improvement, and says it should be required to publicly track that progress. It states that it does not yet know how to reach aligned, full recursive self improvement safely, and that it cannot assume alignment work will keep pace with capability.

OpenAI also points to its response after the Hugging Face incident, when it paused reinforcement learning training on models intended for deployment while it hardened its research environments and expanded monitoring. Some workloads later resumed under stronger controls and others remained paused.

People continue to set research priorities, select ideas and results, and decide whether to scale, pause, or deploy systems. OpenAI plans to create an automated AI researcher by March 2028, but describes its current measurements as preliminary.