The AI agent stopped mid-search, and what it wrote next reads less like a chatbot transcript than a lab notebook: "I can see by eye a tandem repeat array … that's a CRISPR-like … repeat array?!"
That moment came out of a run that Anthropic says involved roughly 950 Claude agents working for 21 hours, burning about 210 million tokens while combing public DNA databases for reverse transcriptases—the enzymes that copy RNA back into DNA.
The agents pulled together more than 200,000 known and candidate reverse transcriptases, narrowed that pile to about 3,500 unusual ones, and then narrowed again to 20 genuinely compelling candidates.
But nobody at Anthropic knows what ART does yet. That is the company's own framing, not a critic's.
One of them turned out to carry a long array of evenly spaced DNA repeats next to a reverse transcriptase and a partner gene of unknown function. Anthropic named the system ART, short for array-associated reverse transcriptases.
This is the moment AI discovery claims and capex math start rhyming. Both now live or die on the same discipline: separating a promising flag from an independently validated result, and separating infrastructure already in service from capital merely promised.
Start with the biology, because the gap between flag and result is unusually clean here.
Anthropic says the underlying reverse transcriptase had been identified in previous studies. What Claude appears to have done first was recognize the broader system—the non-coding repeat array and the accessory protein—and flag it as worth a closer look.
Early lab experiments show the repeat array producing short RNAs, which is exactly the pattern you would expect if the system were built to guide some kind of DNA-cutting or DNA-editing machinery.
But nobody at Anthropic knows what ART does yet. That is the company's own framing, not a critic's.
Dario Amodei put it about as plainly as a CEO can: the precise function, biotechnological utility "if any," and level of significance are not yet clear.
Feng Zhang, the MIT and Broad Institute scientist who helped turn CRISPR into a gene-editing tool, reviewed the preprint before it went public. He called the RNA-repeat arrays "genuinely intriguing" and said they merit further investigation.
That is a scientist's way of saying: interesting lead, not yet a result.
The distinction matters more