“The primary metric for evaluating the success of AI labs like Anthropic, OpenAI, and DeepMind is their efficiency in converting capital and compute (flops) into model intelligence.”