The Wisdom Wall
13 quotable lessons, heuristics and mental models. Every one is playable at the moment it was said. No fortune cookies allowed.
“I think selling in deprecated Nvidia GPUs to China. Lowers the odds of them developing their own alternative ecosystem, which would be a lot power hungrier, um, because you use optical, you bring in optical a lot earlier for scale up fabrics.”
“In some ways, NVIDIA is kind of becoming the central bank of AI. The Federal Reserve of AI. That is like a way to kind of conceptualize what they are doing. And in the same way, the federal reserve, like a lot of monetary policy, it works through private banks. Here you're going to have, um, and I'm probably more banks…”
“They're awesome for the AI infrastructure providers. They just shift economic value from the margins of the frontier labs to the infrastructure.”
“there are a lot of portfolio managers who, if a stock breaks deal price, they sell it no matter what. They consider it a promise that was broken.”
“private investors are often selling to management teams. And at some level that can mean telling management teams what they need to hear, because you want to be able to keep participating in the rounds.”
“cross-sectionally, if you look at the valuations for all these AI names, they just, they can't all be accurate... The AI market is cross-sectionally inefficient right now, which is what I was trying to say.”
“Being the low-cost producer is really going to matter in AI because, um, at some level, the amount of tokens you produce is intelligence because of test time compute, um, you know, post-training reinforcement learning, and so if you can produce those tokens at a lower cost, you'll have a big advantage.”
“the ultimate AI winner will be the one with the lowest cost of infrastructure costs, the lowest cost of compute. And definitionally, you can't be the low cost provider if you're renting your compute from someone else.”
“Open source makes frontier tokens more valuable.”
“if you look at the underlying architectures of the three, what I call big Chinese open source models, and maybe even through a few, or four, You know, if we have Quinn, if we have Kimmy, if we have DeepSeq, and then we have GLM, they're actually evolving in very different ways. And that architectural variation works in…”
“a world where Anthropik, OpenAI, and, you know, maybe even SpaceX are kind of the only model providers that dominate the model layer, that is a hard world for software.”
“if you ever get a biotech company that can become a platform and they have a mechanism, whether it's of drugging, whether it's targeting, if you have something that is broadly applicable, that is when you can get these really, really big hundred billion dollar plus outcomes in biotech, which are rare.”
“I think software is the first area where you've seen a real true kind of economic productivity impact.”