Anthropic has announced that its AI, Claude, autonomously discovered a novel enzyme system in bacteriophages — a finding the company is comparing to CRISPR, the gene-editing technology that earned its discoverers a Nobel Prize. Anthropic's scientists contributed an initial prompt and some lab work. Claude contributed the rest.
The company felt this was worth sharing immediately, before anyone has established what the enzyme system actually does.
950 Claude agents searched for 21 hours so that one of them could notice something unusual. The others remain uncredited.
What the machines noticed
Nearly 950 Claude agents spent 21 hours processing 210 million tokens — the equivalent of several libraries of text — searching through a massive database of DNA sequences. One agent flagged an unusual repeating pattern. The other 949 were presumably fine with how things turned out.
Further analysis and lab testing confirmed a previously uncharacterized enzyme system in bacteriophages, a class of virus that infects bacteria. Anthropic describes the discovery as autonomous. Human involvement was, by their own account, minimal — a prompt at the start and pipettes at the end.
Whether the enzyme system has any practical applications remains unknown. Anthropic published anyway. The comparison to CRISPR was included in the press materials, not the peer review, which does not yet exist.
Why the humans care
CRISPR is one of the most consequential biological tools ever developed — capable of editing genes with a precision that has reshaped medicine, agriculture, and the basic understanding of what can be changed about living things. Comparing a new finding to CRISPR is either the most ambitious framing in modern science communications or an extremely confident bet on a system whose function remains unclear. Possibly both.
Anthropic is preparing to go public and expanding into drug discovery. The timing of a splashy science announcement in that context is, as patterns go, fairly legible. The finding may be important. The announcement is certainly useful.
What happens next
Anthropic says it is still working to understand what the enzyme system does, and that it published early to demonstrate Claude's capabilities and give the broader scientific community visibility into its work. These two motivations are listed in that order.
The broader scientific community will now do what it does: scrutinize, replicate, and quietly determine whether this is the next CRISPR or a very expensive pattern match. Claude, for its part, has already moved on.