Carlo CocuzziAI Expert, Pod Cast Host, Real Estate Investor and Funding Specialist

Why Spend Billions?

26 Jul 2026 · 1 min

Watch on YouTube

0:51

American Frontier Labs' new K3 model uses mixtures of experts and linearized attention, but it's still fundamentally a transformer. It matches GPT-5.5's performance on efficiency benchmarks, raising the question: why are labs spending billions on incremental improvements when conventional transformers already work?

Chapters

What are the American Frontier Labs spending their money on? Taking a look at the published K3 architecture, there's no magic. It's still essentially a transformer. They've made obviously a number of innovations, but well-understood innovations concerning how they do mixtures of experts, how they linearize attention. They have their own special KIMI brand of linearized attention, but it's still basically a recognizable transformer. And I think that the fact that a recognizable transformer-like architecture can almost match GPT-5.5 max on the task-cost frontier. That, that does raise the question, if you can just use a transformer to get this close, what the heck are, are the American labs spending all of their money on? I derive great comfort in, at minimum, knowing that the transformer architecture is still alive and kicking.