帖子

The market may be dramatically mispricing where the long-term value in AI actually sits. If Super-Intelligence can design an even better model, then intelligence itself becomes increasingly commoditized. The real bottleneck isn't the model—it's the compute. A superintelligent system running on your hardware can help you build another superintelligent system. In that world, the scarce asset isn't intelligence. It's the infrastructure that powers and trains it. The atoms matter more than the bits. That's why I increasingly think the biggest winners in AI may not be the frontier labs themselves, but the companies controlling the compute stack, energy, chips, and data centers. Why pay a massive premium for a model that's 1% better when an open-source alternative is 99% as capable at a fraction of the cost? If compute owners can run any model they want and serve tokens directly to users, the moat around proprietary intelligence looks far weaker than many investors assume. Unless frontier labs can permanently keep Super-Intelligence closed, restrict access, or prevent open-source competition, it's hard to see how intelligence alone captures most of the value. The intelligence may be abundant. The compute may be priceless. I can also make it more provocative, hedge-fund style, or Crypto Twitter style if that's the audience you're targeting.