Anthropic announced that its AI model Claude autonomously discovered a novel enzyme system similar to the biological machinery behind Crispr gene-editing technology. The discovery, produced by the company’s newly launched wet lab, involved nearly 1,000 Claude agents searching through a massive DNA sequence database over 21 hours and consuming 210 million tokens before identifying an unusual repeating pattern for human review.

Anthropic noted that scientist involvement was limited to the initial prompt and laboratory testing, which confirmed the finding in bacteriophages. While the company acknowledges it remains unclear whether the discovery will yield practical applications, it chose to publish early results to demonstrate Claude’s scientific capabilities as the startup prepares for an eventual initial public offering.

The announcement highlights a broader shift among leading AI lab developers seeking to demonstrate model value through scientific breakthroughs. Anthropic aims to leverage these wet lab results to attract scientific talent and expand operations into drug discovery, competing with peers like OpenAI, which has similarly targeted advanced mathematical domains.

Why it matters

  • Demonstrates real-world scientific utility for agentic workflows, validating multi-agent compute scaling in R&D.

  • Signals Anthropic’s commercial expansion into life sciences and drug discovery ahead of its planned public listing.

  • Highlights early-stage automated research capabilities, though practical commercial applications in biotech remain unproven.

Source: theverge.com