Biologists push back on Anthropic's Claude discovery claim — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Anthropic announced last Wednesday that a lab powered by 950 Claude agents running for 21 hours identified an uncatalogued repeating pattern around a known enzyme, calling it the system's first discovery and describing it as "reminiscent" of what led to CRISPR.
- Biologist Lucas Harrington posted a viral critique saying "finding a weird cluster of genes and repeats is often the easy part" and that real discovery lies in figuring out what the system actually does — a post endorsed by Eli Lilly's chair and CEO — and urged AI firms to "set the bar high now" for what counts as an AI discovery.
- University of Copenhagen biologist Mario Rodríguez Mestre told the New York Times his team had already identified the same pattern and wondered whether Anthropic learned it from his Claude conversations; Anthropic denies this and Mestre says he's stopping all Claude use.
- OpenAI earlier this month claimed its team agents cracked a million-dollar math problem, but a follow-up questioned whether that problem was the one mathematicians care most about, and a mathematician accused the models of using his work without credit.
- Anthropic's agents reportedly helped narrow roughly 200,000 candidates down to a few worth exploring — work the article calls legitimate scientific labor even if it stops short of a discovery by traditional standards.
Why it matters: With Anthropic and OpenAI racing to claim scientific firsts, the dispute turns on whether calling routine pattern-finding a "discovery" will erode scientific endorsement before a genuinely novel AI-driven biology result lands. The cost is already material: Eli Lilly's CEO amplified a biologist's takedown, and Mestre — who says he found the same pattern first — is cutting off Claude entirely.
Ask SkimNews
