When we wrote about Midjourney Medical, we argued that a company’s real asset is not its product but the expertise it has built to make it. Eleven days later, Anthropic demonstrated the same principle from the opposite direction.
Yesterday, 30 June 2026, Dario Amodei presented Claude Science in San Francisco, a computational workspace for researchers and pharmaceutical companies, built around the existing Claude models (Claude Opus 4.8) and connected to more than sixty scientific databases.
Once again, the world is changing before our eyes.
What exactly it is
Dario Amodei presented Claude Science with this statement:
“Until now, humans have tackled the complexity of biology with their own minds alone”.
The aim is to do for the life sciences what Claude Code has already done for programming: become the reference operating layer for an entire industry.
In practice, Claude Science brings together in a single environment what a researcher today manages across dozens of separate tools: genomic databases, scientific literature (PubMed, bioRxiv, medRxiv), pharmacological compound databases (ChEMBL, Open Targets) and clinical trial management platforms (Medidata, ClinicalTrials.gov). The model does not change: only the context in which it operates changes, designed from the outset for the scientific workflow.
The partners already announced include Novo Nordisk, Allen Institute, AstraZeneca and Sanofi. For university researchers, Anthropic has opened a programme offering $30,000 in credits for fifty selected projects, open until 15 July 2026.
Midjourney and Anthropic: same direction, two methods
The timing of the two announcements says something that goes beyond the individual companies. Midjourney is entering medicine with physical hardware, building on its expertise in computing and design. Anthropic is entering science with vertical software, building on its expertise in language models and data integration.
In both cases, the health and life sciences sector becomes the most ambitious testing ground for whether expertise built elsewhere holds up when the stakes rise. In Anthropic’s case, the answer depends on how far Claude Science actually manages to shorten pharmaceutical research timelines, where every month saved is worth hundreds of millions of dollars (and perhaps lives too).
When Claude Code came out, within a few months it became the go-to tool for thousands of developers. Claude Science will do the same: or at least, it will try to, in a far more regulated environment, with far more demanding standards of evidence, and with far more direct consequences for people’s lives.
New Connections (FAQ)
Is Claude Science also available to those who do not work at a large pharmaceutical company?
Yes, in two distinct ways. For university researchers and doctoral students, Anthropic has opened a credits programme worth up to $30,000 for fifty selected projects, with applications open until 15 July 2026. For everyone else, Claude Science is available in beta today to all subscribers on the Pro, Max, Team and Enterprise plans, at no additional cost beyond their existing subscription.
What is the difference between Claude Science and Claude for Healthcare, announced in January?
They are different products for different audiences. Claude for Healthcare is aimed at hospitals, insurers and patients: it handles medical records, prior authorisations, diagnosis codes and personal health data, and is built on HIPAA-compliant infrastructure for processing sensitive clinical data. Claude Science is aimed at researchers and the pharmaceutical industry: it connects genomic databases, scientific literature and clinical trial platforms, and serves to accelerate drug discovery and computational research. The two products operate at different stages of the cycle: one in patient care, the other in the research that precedes that care.
If Claude Science is not a new model but uses the same existing models, where is the real innovation?
In the same way that made Claude Code relevant to programmers: not a more powerful model, but an environment designed for the specific workflow of the people who use it. Today a researcher loses hours moving from one database to another, extracting data in incompatible formats and reformatting output for different tools. Claude Science removes those intermediate steps, not because the model knows more, but because the environment in which it operates is built for that specific context. The power is not in the model: it is in the integration.