OpenAI says its researchers now use $2,400 a day of AI agents each

Nov 19, 2026
In a new report on research acceleration, OpenAI says agent workdays per researcher rose elevenfold and calls for labs to report signs of self-improving AI.

OpenAI has published a new report on how coding agents are speeding up its own research, saying AI labs should be required to track and publicly report indicators of recursive self-improvement, in which AI accelerates the development of future AI systems.

At API list prices, the median OpenAI researcher now uses around $2,400 a day of agentic AI tools and the 90th percentile researcher about $23,000, up 62% and 105% from September. Counting every eight hours of agent use as one workday, agent workdays per researcher have risen elevenfold, from 0.5 to 5.7 a day, with pretraining and posttraining teams averaging about eight. Lines of code changed per contributor have also risen sharply in 2026, and the report adds an internal version of the Epoch Capabilities Index to track model capability gains more directly.

OpenAI says the uplift is not uniform across research work and that agents still have significant flaws. It warns that past trends may not predict future progress, pointing to recent rapid breakthroughs in mathematics, and says it is monitoring for similar jumps in deep learning. The reporting "should not be read as an intention to pursue full RSI," the company writes, adding that a key task as AI improves itself will be to keep humanity part of the process, according to its report.

We hope you enjoyed this article

Consider subscribing to one of our newsletters like AI Programming Weekly or Daily AI Brief.

Also, consider following us on social media:

Free newsletter

AI Programming Weekly

Weekly news about AI tools for software engineers, AI enabled IDE's and much more.

Whitepaper

Tensordyne Napier: What If One Rack Could Do the Work of Nine?

Tensordyne

This Tensordyne whitepaper presents Napier, an inference-focused AI processor and rack-scale system based on the company’s TDN Math logarithmic number system. It examines infrastructure requirements for large mixture-of-experts and agentic models, compares major inference architecture approaches, and details the TDN AIP processor, TDN72 pod, TDN Link fabric, and Napier Ultra configuration. The paper reports simulation-based performance, cost, and accuracy-validation results, including Tensordyne’s projected comparison of one Napier rack with a nine-rack Nvidia Rubin plus Groq deployment; the chip is reported as taped out and in fabrication.

Read more
Free, six days a week

Daily AI Brief: the AI news that matters, in your inbox.