Anthropic’s AI Lab Tests What Counts as Scientific Discovery

Last Wednesday, Anthropic announced that earlier in the year it had launched a molecular biology lab where Claude agents read and conjecture about difficult biology problems while human scientists run experiments on their reports. [1] Anthropic said the AI-powered lab had made its first discovery. [1] According to the company, 950 agents worked for 21 hours and identified not a new DNA sequence, but a repeating pattern surrounding a known enzyme that had not previously been catalogued. [1]

Anthropic described the pattern as “reminiscent” of what led to CRISPR, a gene-editing technology that has transformed science and medicine. [1] The claim angered some biologists, who argued that finding an unusual cluster of genes and repeats is often easier than determining what the biological system actually does. [1] A viral post endorsed by Eli Lilly’s chair and CEO characterized the agents’ contribution as laboratory grunt work rather than a discovery. [1]

The dispute turns on where scientific discovery begins: at pattern detection, or at experimentally establishing what a biological system does. Anthropic’s 950-agent run shows how agents can search sequence data and surface an uncatalogued pattern, while the criticism emphasizes that novelty in a dataset does not establish biological function. The CRISPR comparison raises expectations, but the source provides no evidence that the reported pattern has comparable scientific or medical utility. What the sources don't address: whether follow-up experiments will reveal the pattern’s function or validate Anthropic’s use of the word “discovery.”