TMTB: Anthropic CFO at UBS Key Quotes

Krishna Srinivasan (Anthropic CFO)

Mission & research focus (alignment + interpretability)

“Anthropic’s mission is to responsibly develop transformative AI that benefits humanity… we’re a research lab… doing actual research into these neural networks… Two areas we invest in are alignment and interpretability… interpretability I would think of as an MRI for the model to see inside it… it’s really important for trust and safety… but it also turns out if you can see inside the model you can build better models… it has a reinforcing effect on our research and the speed with which we can develop.”

Enterprise/B2B orientation

“We really think… the returns to frontier intelligence are much higher in B2B than they are in B2C… companies… are looking to hire the best employees who can learn the fastest… that’s also what companies are looking for in their AI models… frontier intelligence we think has less of a return [in consumer].”

Platform strategy: API-led, now a developer platform plus selective first-party apps

“We’re really focused on allowing companies to build transformative use cases and products on top of our platform… think of it like the early days of AWS, but it’s differentiated because we have an intelligence platform… We’re developing this Claude developer platform… tools and services that sit on top of the model layer… code execution, prompt caching, the ability to spin up virtual machines, the ability to call Claude Code… We also build on top of that platform… selectively build first-party products… all towards this vision of a virtual collaborator.”

Model layer vs apps layer (blurry boundary)

“Unlike some of the static platforms before where you’re debating infrastructure vs application, the lines are very blurred… we’re providing an intelligence platform… and we’re building applications on top of it, with a unique advantage because we know where the models are going.”

Claude Code: trajectory and why it worked