As someone working in bioinformatics, to me it makes sense. Arguably their lead is not in “development” but agentic tool use more broadly. Opus does not have the most up-to-date knowledge despite its training cut-off in my testing, but excels at complex multi-tool calling. The life sciences have a ton of structured data, knowledgebases and some of the most complex use cases for running specialized software, fetching data from dozens of sourcing and contrasting and remixing it. So it fits agentic applications very naturally. Notably, Anthropic is not going into consumer health like OpenAI but rather pitching itself to all the big pharma companies (already partnering with some of them, including Novo).