The AI scientist arrives: a new epoch in autonomous discovery
In 2026, three landmark Nature papers collectively announced that artificial intelligence has crossed a threshold, moving from a laboratory instrument to an autonomous research agent. The Empirical Research Assistant (ERA) writes expert-level scientific software across multiple domains. The AI Scientist executes the entire research lifecycle from hypothesis to peer-reviewed manuscript. MIRA navigates clinical workflows with diagnostic accuracy exceeding that of physicians. Together, these systems signal a profound paradigm shift, albeit one laden with technical fragility, epistemic ambiguity, and unresolved ethical tensions. This article reviews the breakthroughs, dissects the persistent limitations (including hallucinations, silent errors, and reproducibility failures), and argues that the real revolution lies not in replacement of humans but in reconfiguring how human curiosity and machine scalability co-evolve.