On October 31, 2026, the field of AI-driven scientific discovery reached a remarkable milestone. According to AI Distilled by Packt, Anthropic deployed 950 Claude agents to systematically scan enormous genomic datasets, identifying a previously unknown enzyme system that the team named ART in only 21.5 hours. The newly discovered system exhibits properties reminiscent of CRISPR, suggesting exciting possibilities for genetic engineering and synthetic biology and marking a promising milestone for AI-powered scientific discovery.
This result represents a pivotal transition for large-model agents—from tools that primarily converse and generate text toward systems capable of autonomous scientific exploration. Historically, the discovery of novel enzyme systems demanded that biologists devote months or even years to wet-lab experimentation and exhaustive literature review. The identification of ART demonstrates that, given clearly defined human objectives and robust verification frameworks, large-scale agent collaboration can accomplish cross-dataset pattern recognition and hypothesis generation in a fraction of that time.
Anthropic's research team stressed that ART's CRISPR-like characteristics hint at potential applications in precise gene regulation. However, they and independent scientists cautioned that the system's biological function and real-world utility remain to be established through rigorous laboratory testing and peer review. Several genomics researchers noted that AI-generated hypotheses must pass closed-loop wet-lab validation before they can be converted into dependable biological instruments. This episode also underscores the rapid emergence of a "AI proposes, humans validate" collaboration paradigm across research disciplines.
From a commercial standpoint, the discovery further cements Anthropic's leadership in the "AI for Science" arena and supplies a compelling benchmark use case for its enterprise-grade scientific agent offerings. As frontier labs compete to demonstrate tangible scientific and economic value, demonstrations of this kind are likely to accelerate enterprise adoption of autonomous research agents and reshape the paradigm of life-science R&D. Key Facts
Anthropic deployed 950 Claude agents to scan genomic datasets.
The previously unknown enzyme system "ART" was identified in 21.5 hours.
ART exhibits CRISPR-like gene-editing properties.
Its biological function still requires laboratory validation and peer review.
Source: AI Distilled (Packt), threerouter