The speaker believes the market reacted "exactly the wrong way" to Kimi K3, seeing its "wildly bullish" response as an overreaction for "all of compute," "AI," and "trade." Conversely, they argue the market "really underreacted to Grok 4.5 and Muse 1.1 from Meta."
The core argument centers on Grok 4.5, which is described as being on, and even "outside the Pareto frontier" on some metrics, meaning it offers unparalleled "intelligence per dollar." While Grok 4.5 is "not as intelligent as the Anthropic or OpenAI models," it is "more cost-effective" – providing "more intelligence per dollar."
The speaker clarifies that Grok 4.5 is "almost as intelligent as the frontier models, the best of the best," stating it's "within a few percent," yet available at "a fraction the cost." They emphasize that Grok 4.5 is "not retarded" but "extremely intelligent," merely "just a fraction of the absolute best brain out there."
This cost-effectiveness, coupled with near-frontier intelligence, is identified as "where the majority of the opportunity is." The speaker uses the analogy that "you don't need Einstein for every single task." Instead, if one can get "almost Einstein" performance for "a fraction of the cost," that's "where most usage will occur." They acknowledge that for specific, demanding tasks, one might still "pay for the Einstein, the conductor, per Gavin's analogy," but for "the rest of the time," the opportunity lies in getting "near Einstein level performance on a McDonald's salary."