The Scaling Curve: Dario Amodei, Anthropic, and the Race to Build and Survive Superintelligence — Chapter Sixteen

1. 1. By December 2025, Dario Amodei observed that AI was substantially accelerating its own development within Anthropic by writing code and aiding in building new AI systems.
2. 2. In January 2026, Dario Amodei reaffirmed his prediction that AI systems capable of performing at a Nobel Prize-winning level across many fields could arrive by 2026 or 2027.
3. 3. DeepMind's Demis Hassabis, while agreeing on the eventual destination of advanced AI, expressed more caution regarding timelines for automating fields like the natural sciences.
4. 4. Amodei increasingly treats the 'country-of-geniuses' scenario, involving millions of superintelligent minds, as a planning assumption for 2027 rather than a hypothetical thought experiment.
5. 5. Dario Amodei's positive vision for superintelligence includes ending major diseases like cancer and heart disease, extending human life, and adding trillions of dollars annually to the economy.
6. 6. Despite a 75-90% probability for positive outcomes, Amodei maintains a 10-25% chance of civilization being substantially derailed, a risk he actively works to mitigate.
7. 7. Amodei warned of simultaneous high GDP growth and high unemployment due to AI's ability to substitute for cognitive labor, potentially displacing entry-level white-collar jobs.
8. 8. By early 2026, the acceleration of AI had become a 'lived reality' within Anthropic, with models building models and code writing code, outpacing even their creators' expectations.