Who, beyond Anthropic themselves, can afford such purposefully wasteful uses of LLMs? "the model sucks a bit so we just have best-of-4 & adversarial reviewing

Who, beyond Anthropic themselves, can afford such purposefully wasteful uses of LLMs? "the model sucks a bit so we just have best-of-4 & adversarial reviewing agents; surely one more agent will do the trick"