OpenAI, Anthropic and Google all price output tokens several times higher than input tokens — the same pattern across every provider. Here's the actual reason, and what it means for how you should write prompts.