It's fine, but the article makes some good points why - less cognitive load for GPT and less tokens. I think the transistor to logic gate analogy makes sense. You can build the thing perfectly with transistors, but just use the logic gate lol.