The Rise of the Lone Scientist Under Agentic AI

Why it matters

A new sole-authored study, produced with extensive assistance from Claude Opus 5.5, finds that single-author papers on the arXiv preprint server have risen steeply since late 2025, with the strongest growth in mathematics, condensed matter physics, and quantitative biology. The author, based in Imperial College London's physics department, attributes the shift to agentic AI enabling individual scientists to conduct work that previously required collaborators, rather than to a small group of hyperproductive authors, as arXiv itself has suggested. Notably, the effect is absent in experimental particle physics, where large teams remain essential, and growth is strongest in mathematics branches relying on abstract reasoning. The study also distinguishes this surge from the arrival of ChatGPT, which left no detectable trace on submissions outside computer science. Writing in Unite.AI, Anderson frames the findings within unresolved legal questions: under US law, material generated without sufficient human creative control is not copyrightable, and as human contribution shifts from execution to direction, the basis of individual intellectual credit becomes harder to establish, potentially reshaping academic authorship and careers.