The announcement is real and the numbers in the post match the primary source closely. Anthropic states that after 21 hours spent searching this data by roughly 950 agents using 210 million tokens, one of the agents spotted a repeating pattern of DNA sequences occurring next to the gene for an odd-looking reverse transcriptase, and that after further analysis and testing in its lab it recognized that this pattern marked a previously uncharacterized enzyme system found in bacteriophages, which it calls array-associated reverse transcriptases (ART).

Anthropic says Claude agents gathered over 200,000 RTs, picked out 3,500 new candidate systems, and narrowed those to the 20 most compelling candidates that they analyzed to produce human-readable reports.

The critical qualifier sits in Anthropic's own text. Anthropic states that while the underlying RT, found in a jumbo phage, had been identified in previous studies, Claude appears to be the first to notice the system's defining features, an associated array of non-coding DNA sequences and an additional accessory protein of unknown function. An independent expert put the same point bluntly: "The enzyme itself was already known," Perrin explained, adding that what is new is the recognition that it may form part of this larger system.

On what the lab work established, the ceiling is low. Initial laboratory experiments found that the ART repeat array is expressed as multiple distinct short RNAs, which is of interest because CRISPR arrays also generate RNAs, but researchers caution that ART has not been shown to function like CRISPR and its biological role remains unknown.

Per the preprint, the team has not yet shown that the enzyme is active, or that it works on these RNAs. The same outlet reports that the campaign ran for 21.5 hours without human input, across 949 agent sessions and 215.6 million tokens, using Claude Mythos 5 against a database of 1.9 billion protein clusters.

On independent standing: the result was announced in a blog post and a preprint that has not been peer reviewed, and it is the first result from a new molecular biology group and lab inside the company. The named outside expert quoted by Anthropic endorsed the direction, not the conclusion: Feng Zhang, a CRISPR pioneer at MIT and the Broad Institute, reviewed the pre-print and called the identification of RNA-repeat arrays associated with reverse transcriptases "genuinely intriguing" and said it merits further investigation. A reproducibility caveat reported from the preprint cuts against the broader "AI can do this fast" reading: one digest records the honest caveat that ten reruns of the same campaign failed to rediscover it , and the preprint's reported follow-up campaigns did not rediscover the pattern in ten repeat runs, according to The Next Web's review.