Anthropic recently announced that its molecular biology lab, where Claude agents read and conjecture about biology problems, had made its first discovery. The system of 950 agents found a repeating pattern surrounding a known enzyme in 21 hours—a pattern Anthropic said had not been catalogued before.
Some biologists are unimpressed. A viral post by biologist Lucas Harrington, endorsed by Eli Lilly's chair and CEO, argued that finding a weird cluster of genes is often the easy part; the hard part is figuring out what the system actually does. Another biologist, Mario Rodríguez Mestre, said his team had already discovered the pattern, and though Anthropic denies learning from his chats, he says he is stopping use of Claude.
The broader issue is how AI companies present their systems. Rather than tools like microscopes, they claim AI is making discoveries itself. That framing can make genuine progress look like breakthrough-or-bust, and shifts goalposts—as with OpenAI's recent math claim, which skeptics questioned. Harrington suggests companies should set the bar high now so future real AI discoveries are appreciated.