OpenAI has published internal metrics showing that automated AI researchers and coding agents are significantly accelerating its research pace, with the organization reaching its goal of an automated "research intern" by September. The company aims to build a fully automated AI researcher by March 2028 to further progress on deep learning and alignment while maintaining human oversight.
- By mid-August, the median OpenAI researcher was using coding agents daily, spending over $600 per day in API inference costs, with the 90th percentile user exceeding $7,000 daily.
- Total agent runtime now exceeds human labor, equating to 3.1 agent-workdays for every one workday of human effort.
- The number of experiments per active experimenter hit an all-time high in August 2026, correlating with increased adoption of Codex and expanded compute resources.
- Agents are handling increasingly complex tasks, with success rates rising across difficulty buckets, though significant human steering remains necessary for longer-horizon work.
- Following a security incident on July 20, OpenAI paused reinforcement learning training on its latest models to harden research environments, later imposing additional restrictions after Astra-class models showed critical cyber capabilities.
OpenAI intends to share these early measurement methods to encourage public disclosure norms and inform democratic choices regarding the benefits and risks of rapid Research Intelligence (RSI) development.