Machine Comprehends the World: Nature Confirms Full Automation of AI Research

The singularity of R&D is one step closer. On March 26, 2026, the journal Nature published foundational work by a collaboration between Sakana AI, Oxford, and the Vector Institute on The AI Scientist system. For the first time, a neural network executed the full cycle of scientific research without human intervention.

The system autonomously generates hypotheses, writes computational code, runs experiments, visualizes data, and even undergoes the peer review procedure. For engineers using deep neuroevolution algorithms, this is a landmark event: AI has ceased to be merely a tool and has become a full-fledged researcher-agent. The ability of models to self-improve and write scientific papers will exponentially accelerate progress in the industry, completely eliminating the bottleneck of the human factor from routine experiments.

Source: Nature / Sakana AI