Anthropic said its newly launched life sciences research group used Claude to identify a previously uncharacterized enzyme system in bacteriophage DNA, calling it array-associated reverse transcriptases, or ART, according to Anthropic and TechCrunch.
About 950 Claude agents spent 21 hours and consumed 210 million tokens scanning a large public DNA sequence database to find the system, which pairs a reverse transcriptase enzyme with a partner gene and an array of repeating DNA sequences, Anthropic said. The repeat pattern resembles CRISPR arrays, and early experiments show the array produces distinct short RNAs, though its function remains unknown, according to Anthropic.
Anthropic CEO Dario Amodei said the discovery was made "mostly, though not entirely, by Claude," TechCrunch reported. He also said Stanford researchers had previously found a system "in some ways similar" to the one Claude identified, meaning the finding may not be as novel as first described, according to TechCrunch. Feng Zhang, the MIT researcher known for his CRISPR work, called it "an exciting example of how AI agents can contribute to biological discovery" that merits further investigation, according to Anthropic.
The lab, opened this spring in the Bay Area, runs at biosafety levels 1 and 2 and does not handle human pathogens. All physical lab work, including verifying what Claude flagged, was done by human scientists, and Amodei said fully autonomous AI lab operation is not something Anthropic is doing today, according to TechCrunch.
The discovery is a research lead, not a finished result, given Amodei's own caveat about the earlier Stanford work. What is worth watching is the workflow: 950 agents searching a public database for 21 hours to surface a pattern is a preview of how much routine scientific pattern hunting can be handed to agents, with humans stepping in only to verify what they find.