After 21 hours, roughly 950 Claude agents and 210 million tokens of searching, one of them stopped on a stretch of raw DNA and wrote: “[The DNA next to the RT] is spectacular: I can see by eye a tandem repeat array … that’s a CRISPR-like … repeat array?!”
That exclamation, published by Anthropic on September 23, led to what the company calls a previously uncharacterized enzyme system in bacteriophages, the viruses that infect bacteria. Anthropic says its model found it with only high-level direction from scientists.
What Claude was asked to do
The task was narrow. Researchers prompted Claude to comb a massive database of DNA sequences for new examples of reverse transcriptases, or RTs, enzymes that copy RNA into DNA. Think of them as scribes working backwards. Anthropic says its people handled the initial prompt and the lab work, and the agents did the rest: they investigated distinct RT families and used their own judgment to pick candidates.
Across the campaign, Claude agents gathered more than 200,000 RTs, picked out 3,500 new candidate systems and cut those to the 20 most compelling, each with a report a person could read. Anthropic says an expert scientist can spend weeks or months on that kind of analysis.
From a hunch to a lab result
The agent that raised the alarm did what a human would. It counted the repeats, measured their spacing, compared the layout with known RT systems and searched the literature for earlier reports. Convinced, it filed a report for review. Tests in Anthropic’s lab then supported the find.
The team named it array-associated reverse transcriptases, or ART. The system has three parts: the RT, a partner gene beside it, and a long array of evenly spaced DNA repeats. The underlying RT, from a jumbo phage, had appeared in earlier studies. Claude appears to be the first to notice the array and the accessory protein of unknown function. Early experiments show the array is expressed as distinct short RNAs, which hints at something similar to how a CRISPR array makes CRISPR-Cas programmable.
Anthropic is careful about what it does not know. Nobody has yet worked out what ART actually does. Its characteristics, though, have only been seen together in a handful of other systems, all programmable and all able to cut, copy and paste DNA. Several of those, beyond CRISPR, are in development as tools.
Feng Zhang, a CRISPR pioneer at MIT and the Broad Institute, reviewed the pre-print. “The identification of RNA-repeat arrays associated with reverse transcriptases is genuinely intriguing and merits further investigation,” he said.
Why a software company has a wet lab
The work comes from a life sciences group and laboratory Anthropic formed in the spring of 2026, in the Bay Area. It sits beside teams working on drug discovery and on training Claude in biology and chemistry. The lab works only at biosafety levels BSL-1 and BSL-2, does not handle pathogens that infect humans, and uses human scientists for all bench work.
Anthropic points to precedent. Restriction enzymes came from bacterial immune systems and launched the biotechnology industry. Taq polymerase, from a bacterium in a Yellowstone hot spring, became the basis of PCR. CRISPR started as an odd repeat sequence in bacterial DNA and now underpins gene-editing medicines.
The company published early because it wants other scientists to see what Claude can do, and it invited researchers with a question to get in touch. Experiments to establish how ART works are underway.
