Nature | Towards end-to-end automation of AI research
Nature | Towards end-to-end automation of AI research
Chris Lu, Cong Lu, Robert Tjarko Lange, Yutaro Yamada, Shengran Hu, Jakob Foerster, David Ha & Jeff Clune
DOI: https://doi.org/10.1038/s41586-026-10265-5
Background
The automation of science is a long-standing ambition in artificial intelligence research. Although substantial progress has been made in automating individual components of the scientific process, a system that autonomously navigates the entire research life cycle—from conception to publication—has remained out of reach.
Methods
Here we present a pipeline for automating the entire scientific process end to end, termed The AI Scientist. The system creates research ideas, writes code, runs experiments, plots and analyses data, writes the full scientific manuscript, and performs its own peer review. It leverages modern foundation models within a complex agentic system, and is evaluated in two settings: a focused mode using human-provided code templates and a template-free, open-ended mode leveraging agentic search for broader scientific exploration.
Results
Both evaluation settings produce diverse ideas and automatically test, report, and evaluate them. The system’s ideas, execution, and presentation are of sufficient quality that a manuscript generated by it passed the first round of peer review for a workshop of a top-tier machine learning conference (acceptance rate 70%).
Fig. 1 The AI Scientist workflow.
Fig. 2 Selected sections from a paper generated by The AI Scientist that was accepted via peer review at a top-tier machine learning conference workshop.
Fig. 3 The phases and compute scaling of the AI Scientist.
Conclusion
This achievement demonstrates the growing capacity of AI for making scientific contributions and suggests a potential paradigm shift in how research is conducted. However, it may also introduce risks such as burdening peer review systems and adding noise to the scientific literature; if developed responsibly, such autonomous systems could substantially accelerate scientific discovery.
Reference
Lu, C., Lu, C., Lange, R.T. et al. Towards end-to-end automation of AI research. Nature 651, 914–919 (2026). https://doi.org/10.1038/s41586-026-10265-5
