Anthropic has released Claude Science, a research workbench that pulls the tools a scientist juggles all day (PubMed, Jupyter, R, cluster terminals, dozens of genomics and proteomics databases) into one place where an agent runs the analysis and a second agent checks its citations and math. It is in beta for Pro, Max, Team, and Enterprise users, and is pre-loaded with more than 60 curated skills and wired into NVIDIA’s BioNeMo models.
The product is exciting, but let’s explore the dark side. Until now, the model companies mostly rented out the model and let other people build the products on top. Claude Science is Anthropic building the product itself. Importantly, it built it using their most advanced AI, which (and this is the scariest part) is not available to anyone outside the company.
My favorite question for founders who want me to invest in their startups is: “You say you’ve built the death star. Nice. Are you going to rent it out or go f#ck up some planets?” Anthropic has built the death star and, while they started by renting it out, they’re now going to f#ck up some planets.
The counter argument is that distribution decides these fights. The hyperscalers (AWS, Azure, Google Cloud) already hold the enterprise contracts, the compliance paperwork, and the billing relationship, and they will bundle whatever model into vertical apps of their own. Fair enough. That argument changes which giant absorbs the workflow layer, but it does nothing for the company whose whole product was that layer.
Possible moats (defensible positions) include unique assets such as proprietary data, regulated trust, physical operations, and real customer relationships. However, it is clear that whole industries are now on the roadmap of the foundational model builders.
Claude Science will help scientists do better work, and I am glad it exists. In practice, the people who sell electricity have decided to build appliances. Plan accordingly.
Every company needs a Claw strategy. Do you have one?
Author’s note: This is not a sponsored post. I am the author of this article and it expresses my own opinions. I am not, nor is my company, receiving compensation for it. This work was created with the assistance of various generative AI models.