
Don't Worry About the Vase Podcast AI #135: OpenAI Shows Us The Money
Sep 25, 2025
This week, the conversation features insightful contributions from two regulars who explore the recent $100B NVIDIA investment by OpenAI and the implications of new compute-intensive offerings on the horizon. They dive into the everyday utility of LLMs like GPT-5, and discuss failures in complex contexts. The duo also evaluates the effects of AI on coding tasks and delves into the potential pitfalls of automation in venture capital. Lastly, they touch on political responses to AI regulation and the challenges of defining AGI in legal terms.
AI Snips
Chapters
Books
Transcript
Episode notes
Economics Papers Miss Broader AI Risks
- Economic papers largely ignore existential risk while modeling AI's productivity and distributional impacts.
- Growth could scale with compute but governance and control questions remain central.
Build Markets With Escalation And Verifiability
- Design AI markets and agents to know when to escalate to humans and price compute versus quality.
- Build resistance to manipulation and mechanisms for verification and trust.
AI Capex Will Boost Near‑Term GDP
- AI capex alone could add meaningful GDP growth in 2026–27, even without productivity gains.
- The scale of planned investment dwarfs earlier economist forecasts.






