Anthropic’s AI agents identify novel enzyme system in 21-hour search

Key points

• Anthropic’s newly launched life sciences lab used Claude to search massive DNA databases, leading to the identification of a previously uncharacterized enzyme system.
• The discovery, named array-associated reverse transcriptases (ART), involves a reverse transcriptase, a partner gene, and a DNA repeat array resembling CRISPR structures.
• The process required nearly 950 Claude agents, 21 hours of computation, and 210 million tokens, with human scientists providing only the initial prompt and laboratory verification.
• Anthropic is sharing these early findings to attract scientific talent and validate Claude’s capabilities in biology, particularly as the company prepares for an initial public offering.
• The practical applications of the ART system remain unknown, though its structural similarities to CRISPR suggest potential for future programmable biological tools.

Background

This announcement follows Anthropic’s recent launch of a dedicated wet lab in the Bay Area, a move intended to expand the company’s reach into drug discovery and fundamental biology. The timing coincides with intense scrutiny of Anthropic’s upcoming IPO, where investors are evaluating its revenue growth and competitive position against rivals like OpenAI. While OpenAI has focused on solving advanced mathematical problems, Anthropic is pivoting toward biological research to demonstrate the tangible value of its more capable models.

How outlets are covering it

Anthropic emphasizes the autonomous nature of the discovery, highlighting that human involvement was limited to the initial prompt and lab work, and compares the finding to the transformative impact of CRISPR. The Verge, however, notes that the practical applications of the enzyme system remain unclear and questions whether the announcement is premature. The Verge also contrasts Anthropic’s approach with OpenAI’s focus on mathematics, suggesting that AI companies are increasingly using scientific research to prove the value of their models. Feng Zhang, a CRISPR pioneer, praised the discovery as an exciting example of AI contributing to biological discovery, while acknowledging the need for further investigation.

Why it matters

The discovery underscores the growing role of AI in scientific research, potentially accelerating the identification of new biological tools and therapies. For Anthropic, it serves as a strategic move to attract scientific talent and validate its models ahead of a high-stakes IPO. The broader implications include a shift in how AI companies demonstrate their capabilities, moving from abstract problem-solving to tangible scientific contributions that could reshape fields like genomics and medicine.

What to watch

Anthropic is continuing to investigate the function of the ART system and plans to share more detailed findings in a technical report. The company is also seeking collaboration with other scientists to extend its AI-driven hypothesis generation approach to other areas of genomics and biology. Meanwhile, the scientific community will be watching to see if the ART system can be developed into a practical tool, similar to CRISPR, and how Anthropic’s lab expands its research into drug discovery and other domains.