Anthropic CEO Dario Amodei

Background and Founding

• Dario Amodei trained in neuroscience and was a postdoc in computational biology at Stanford; he moved into AI after seeing early deep neural network results from Google and the University of Toronto.
• He joined OpenAI in the early days, led research there, and with the future Anthropic co-founders was among the first to document exponential scaling laws.
• Around 2019-2020 he concluded AI would reshape the economy and that, because models are grown rather than programmed line by line, safety and responsible deployment would be essential.
• He and seven co-founders, including his sister Daniela, left OpenAI to found Anthropic about 4.5 years ago.

Running Anthropic with Daniela

• Dario and Daniela grew up in San Francisco and attended Lowell High School four years apart.
• Daniela handles day-to-day operations, org design and business structure; Dario handles strategy, research direction, safety and vision. He says he couldn't run the company without her.
• Salesforce is a very early investor, and Anthropic now occupies the former Slack building, with Dario's office formerly Stuart's office.

Enterprise Focus and Competition

• Anthropic has a plurality, perhaps majority, of enterprise customers and enterprise revenue, while OpenAI has gone consumer, a divergence Dario calls a natural experiment.
• Google is both investor/partner and competitor via Gemini; co-opetition is normal across chip, cloud, model and distribution layers. He has been friends with Demis Hassabis for about 15 years and was happy about the AlphaFold Nobel Prize, dedicating his essay Machines of Loving Grace to it.

Agentic AI and Code

• The biggest current jump is agentic abilities, doing tasks end to end, most visible in code because developers are early adopters.
• On many Anthropic teams, 90% of code is written by AI, with humans editing and supervising; Claude once found an obscure cluster bug engineers had missed.
• By comparative advantage this is a rebalancing, not replacement, and may require as many or more engineers; he worries about broad labor disruption in the next two to five years.
• Coding benchmarks on random GitHub PRs have gone from about 5% to roughly 77% over the last year and a half.
• The ecosystem surprise he least expected: tools like Lovable and Replit letting non-coders write code end to end.

Healthcare

• Models should be treated like a second opinion; Claude listed diagnostic possibilities doctors missed during Daniela's pregnancy, but blindly following it would not have gone well.
• Human-model collaboration models take longer to work out in high-stakes regulated fields like radiology than in code.

Vision and Compute

• Revenue is approaching a $7 billion run rate, up 10x per year for three years; he wants Anthropic to put AI at the center of every enterprise and be known for trustworthiness.
• Future training will need millions of chips and many gigawatts of power; he is confident Anthropic can secure compute but worries some data center deals seem fishy and possibly double counting.