OpenAI reports increased use of AI agents in its research work
- OpenAI
- recursive self-improvement
- agentic AI
- research acceleration

Median researcher used about $2,400 a day in agentic AI tools by October 2026
Agent workdays rose elevenfold, from 0.5 to 5.7 per researcher per day
Safety and Alignment recorded the largest absolute rise in researcher contributions during 2026
OpenAI published an update on research acceleration based on internal measurements of how coding agents are being used in its research organization. The report describes increased use of agents and changes to researchers’ workflows, while noting that the findings provide only a partial picture of research acceleration.
OpenAI reports that, as of October 2026, the median researcher used about $2,400 per day in agentic AI tools, valued at API list prices. Agent workdays—a measure that counts each eight hours of agent use as one workday—increased from 0.5 to 5.7 per researcher per day. The report says this measure generally captures interactive use and may undercount automated work.
The report also says that, as of September 2026, the API-equivalent value of agentic AI use in OpenAI’s research organization was about twice researchers’ non-equity compensation. That comparison excludes equity compensation. OpenAI says Safety & Alignment had the largest absolute increase in documented researcher workweek contributions in 2026.
OpenAI says the findings do not show uniform gains across research tasks: estimated agent success rates varied by task category, and the report notes that agents still have significant flaws. The company plans to continue monitoring its OpenAI ECI capability metric and other indicators to assess whether capabilities are accelerating because of recursive self-improvement.
Key terms explained
- OpenAI
OpenAI is a research and deployment organization focused on creating artificial general intelligence. The company operates with the goal of ensuring that its advanced AI technology provides benefits to all people.
- agentic AI
Agentic AI refers to autonomous systems capable of acting independently to reach specific objectives. These systems can perform complex tasks with minimal human oversight by making contextual decisions.




