Anthropic CEO Dario Amodei on designing AGI-pilled products, model economics, and 19th-century vitalism

Key takeaways

1. Amodei advises against wrapper products that patch the weaknesses of today's model, because the next model release can eat them. Build something complementary to where the field is heading.
2. He says long product roadmaps make even less sense when the technology moves under you. Use tighter ship schedules, and try new models on real prototypes internally.
3. Amodei suggests judging a model business model by model. In his example, a $100 million model earns $200 million while the next one costs $1 billion, so the company-wide P&L looks bad even though each model is profitable.
4. Anthropic launched with seven cofounders and equal equity, against most advice. Amodei says it worked because they had history working together, and they now carry the company's values as it scales.
5. For large companies adopting new technology, Amodei recommends a strike team separate from the rest of the company that builds working prototypes first, then does the harder work of integrating them.