My favourite comment (by Vijay):

Palantir talking about AI sovereignty is like Hitler talking about multiculturalism and human rights. F@ck off

I have doubts about the Palantir model. This company doesn't even build its own AI. And the strategy they're pushing ultimately seems to be an attempt to standardize Palantir's topology in defense and healthcare.

In other words, they want to create the top-level spec and make all the subcontractors conform to it. As you know, if you're a hardware programmer, you know how varied the standards are across different devices. So I know that system integration would be incredibly difficult.

I understand they want to become like NVIDIA's CUDA ecosystem, but NVIDIA is a hardware manufacturer at its core, so that kind of strategy makes sense for them. Can a middleware layer actually pull that off?

If they were an AI foundation company, and their AI was overwhelmingly superior, then this strategy could work in the end. But they're not even a foundation AI company. I'm not sure this is even possible.

The so-called technical moat is ultimately an integration layer called an ontology. How do they prove that it's superior? I fundamentally don't understand it. The idea of a middleware layer that can absorb all complexity, as a programmer, I naturally arrive at the conclusion that it would become a single point of failure.