Claude found a hidden enzyme system in viral DNA

The clue was beside the gene

Claude was looking for unusual enzymes when it noticed a repeating stretch of DNA beside one of them. Anthropic says that observation led to a previously uncharacterized biological system in viruses that infect bacteria. The company announced the finding on September 23.[1]

The enzyme itself had appeared in earlier research. What Claude picked out was the arrangement around it: repeated DNA sequences and an accompanying protein. The researchers named the system array-associated reverse transcriptase, or ART.[1]

A discarded lead opened another door

The technical report describes a search run with Claude Mythos 5. Agents divided the work between workers and supervisors, recording their findings in shared notes. The campaign involved 949 agent sessions over roughly 21 hours.[2]

The route to ART was indirect. An early idea about a neighboring gene failed, but the agent proposed another look at the enzyme. A follow-up examined the surrounding DNA and found the repeats. The agents then counted them, checked their spacing and searched the literature before sending a report to human reviewers.[2]

This is what makes the result interesting: the search could change direction when an observation deserved a closer look. Rejecting one explanation did not mean abandoning the clue.

A discovery, with its function still unresolved

Human scientists performed the laboratory experiments. They found that the repeated region produces distinct short RNA molecules during infection. That gives the pattern biological significance, but does not yet explain the job the whole system performs.[2]

Anthropic compares the arrangement to CRISPR, whose repeated sequences helped lead to gene-editing tools. ART has not been shown to be a new gene-editing tool. Its function remains under investigation, and the detailed findings were released as a preprint.[1]

For now, the achievement is a promising lead brought from a large database into the laboratory. Whether it becomes useful biotechnology will depend on the experiments that come next.