A big insight from this podcast is that intelligence will be commoditized and open source. It is an existential risk to Nvidia for intelligence to be closed source as it would make them captive to a few model companies who could chose to optimize on a different hardware stack. Nvidia needs the most used models to be optimized for their stack. It explains why Nvidia is putting so much effort into research and there is a clear trajectory for their open source models to saturate the biggest capability benchmarks that matter. SWE, general knowledge work, autonomous driving, physical AI. There are still a lot of domains where the frontier of capabilities can continuously expand and for closed source models to be dominant. However, the vast majority of tokens produced in the future will be produced by open source models.