Summary
An Anthropic-authored bioRxiv preprint reports that Claude Code agents surveyed 1.9 billion protein clusters and identified a family of phage reverse transcriptases associated with tandem DNA repeats. The researchers also report that the repeat arrays are highly expressed during Staphylococcus phage infection.
An Anthropic-authored bioRxiv preprint reports that autonomous Claude Code instances identified a family of reverse transcriptases associated with tandem repeat arrays while surveying 1.9 billion protein clusters. The researchers call the enzymes array-associated reverse transcriptases, or ARTs, and describe them as a previously unreported family found in jumbo phages.
Reverse transcriptases copy genetic information from RNA into DNA. The ART systems described in the preprint are linked to arrays made up of repeated units about 200 nucleotides long, as well as a dedicated partner gene. The scale of the search matters because metagenomic databases contain large collections of genetic sequences from microbes and their viruses, making it difficult to inspect every candidate manually.
What the researchers report
The authors say session transcripts show that Claude identified the ART family by examining DNA sequences and recognizing their unusual repeats. They attribute this pattern-recognition behaviour to specific internal signals in Mythos 5 that respond to repeated DNA.
The study also reports that ART arrays are highly expressed and appear as discrete units during Staphylococcus phage infection. The authors interpret this expression pattern as suggesting a reverse-transcription system directed by a repertoire of distinct RNAs. That proposed RNA-based organisation is an interpretation of the reported observations; the preprint abstract does not specify a biological role for the system.
The work is posted as a bioRxiv preprint. All authors are Anthropic employees and may hold company equity; the authors also disclose that Anthropic has filed a provisional patent application related to the work.