Claude discovered a new CRISPR-like enzyme system after scanning DNA data for 21 hours. Anthropic stated that Claude autonomously identified a previously undocumented enzyme system in bacteriophage DNA—viruses that infect bacteria. The new system has been named ART, short for array-associated reverse transcriptases, and exhibits several features reminiscent of CRISPR. During the research, approximately 950 parallel Claude agents worked for 21 hours, analyzing over 200,000 reverse transcriptase enzymes using about 210 million tokens. The AI narrowed the dataset down to 3,500 unusual systems, then selected the 20 most promising candidates before discovering ART. Anthropic noted that a similar effort by a human expert could take weeks to months. ART consists of a reverse transcriptase, a companion gene, and a structured DNA repeat array similar to a CRISPR array. Initial experiments suggest this array produces multiple distinct short RNAs, raising the possibility that ART may also have a “programmable” mechanism akin to some known gene-editing systems. Feng Zhang, one of the pioneering CRISPR scientists at MIT and the Broad Institute, called the discovery “genuinely intriguing” and highlighted it as a notable example of how AI agents can contribute to biological discovery. However, the significance of this finding remains unclear. Anthropic itself acknowledges it does not yet know ART’s precise function, whether it has biotechnological applications, or how significant the system truly is. Laboratory experiments are ongoing. What is particularly noteworthy is not just ART itself, but the discovery process: AI autonomously searched data, identified anomalous patterns, cross-referenced literature, and formed hypotheses before human researchers conducted experimental validation. If this model proves effective at scale, AI agents could become a major accelerator for biological research and drug discovery.