Can AI help accelerate its own development?
Artificial intelligence is increasingly being used to assist the researchers who build AI systems.
A new study by nonprofit research organization METR suggests this process may already be affecting the pace of AI development.
According to reporting published November 5, 2026, researchers estimated that Anthropic’s AI models improved on certain benchmark tasks approximately 1.9 times faster after October 2025 than during the preceding period.
The study examined confidential research information, interviews, and model-performance data over an 11-week assessment.
Importantly, the researchers described the estimate as uncertain. Their analysis suggested a possible range of approximately 1.5 to 2.5 times the historical pace.
How AI supports technical research
AI tools can assist researchers with programming, analyzing experimental results, identifying problems in software, and carrying out repetitive technical work.
These capabilities may help research teams test ideas more quickly.
In AI development, this creates a potentially important feedback process: increasingly capable systems can help researchers develop subsequent systems.
However, using AI as a research assistant is not equivalent to having fully autonomous AI systems design and train their successors.
METR’s findings do not establish that AI can independently perform the most consequential stages of frontier model development.
What the study actually shows
The reported acceleration concerns model performance on particular benchmark tasks.
It should not be interpreted as proof that all aspects of AI research are improving at the same rate.
The researchers also cautioned against treating informal productivity estimates as reliable measurements.
Differences in benchmarks, research methods, computing resources, and development priorities can complicate attempts to isolate the effect of AI assistance.
The study therefore offers evidence of a possible acceleration mechanism rather than a definitive measurement of AI’s overall contribution.
Implications for technical careers
If AI systems increasingly assist with programming and research, the skills required in technical careers may evolve.
Researchers and engineers may spend more time designing experiments, evaluating outputs, understanding system limitations, and making higher-level technical decisions.
Skills such as critical thinking, scientific reasoning, software architecture, and evaluation could become increasingly valuable.
For women pursuing careers in AI, machine learning, computer science, and research engineering, these developments reinforce the importance of understanding both how to use AI tools and how to assess their reliability.
Why safety research matters
Faster AI development also creates challenges for safety and governance.
If capabilities advance more quickly, organizations need evaluation methods capable of identifying risks before systems are widely deployed.
Researchers must understand not only whether models can complete more tasks, but also whether they remain predictable, controllable, and appropriately aligned with human objectives.
This is one reason independent evaluation organizations play an important role in the AI ecosystem.
What comes next?
METR’s findings contribute to a growing debate about the relationship between AI-assisted research and the speed of technological progress.
Further studies will be needed to establish how much of the observed acceleration can be attributed to AI tools and whether similar patterns appear across other research organizations.
The broader lesson is that AI’s influence on scientific work may extend beyond individual productivity gains to the pace at which new technology is developed.
Sources and further reading
Reuters — METR study, November 5, 2026
METR — Independent AI evaluation research
Looking for Your Next Opportunity in STEM?
Fems in STEM Career Connections highlights women-led innovation, career opportunities, and emerging technology developments.
Subscribe to Fems in STEM Career Connections for curated updates.
This article is independent editorial coverage based on publicly available research reporting. Findings and estimates remain subject to methodological limitations.
Discover more from Fems in STEM
Subscribe to get the latest posts sent to your email.
