Simon Willison's experiment highlights a persistent gap in the generative AI hype cycle: the chasm between a model's technical capability and its practical accessibility. While the MLX port is a clever piece of engineering that leverages Apple's hardware ecosystem, the resource demands—both in storage and time—render it a novelty for all but the most dedicated researchers and hobbyists. This serves as a useful reality check; the industry's focus should arguably be on efficiency and integration before declaring every new model a breakthrough. The real story here isn't the video output, but the reminder that raw model capability means little without a viable path to actual use.