Your Prompts Are Wrong. Both OpenAI and Anthropic Just Said So. Simultaneously.

Both companies released official prompting guides in the same week, and both reach the same conclusion: your old prompts don't work anymore. The post hit 2,301 likes.

Here's what's fascinating. They arrive at the same destination from opposite directions.

Claude Opus 4.7 stopped guessing your intent. It does exactly what you type, literally. If your prompt is vague, you get vague output. If your prompt is specific, you get specific output. The model won't fill in gaps or add structure you didn't ask for. Anthropic's guide is essentially: be explicit about everything, because the model won't assume what you want.

GPT-5.5 went the other direction. It defaults to efficient, direct, task-oriented output. It strips away verbosity and filler unless you explicitly ask for it. OpenAI's guide is: if you want detail, say so, because the model will give you the minimum viable answer by default.

Both shifts break the "conversational" prompting style that worked in 2024. Back then, you could write "help me build a React component that..." and get a reasonable result. Now, with Claude you need to specify exactly what you want (structure, length, format, constraints) because it won't infer. With GPT, you need to specify when you want depth because it'll default to brevity.

I've been feeling this shift in my daily Claude Code usage. Prompts that worked three months ago produce different results now. Not worse, necessarily, but different. The model is more obedient and less creative. It does what I say instead of what I might have meant. That's better for coding (where precision matters) and worse for exploration (where you want the model to surprise you).

What to do about it: Read both guides today. Then pick your five most-used prompts and rewrite them. For Claude: add explicit structure, format, and constraint instructions. For GPT: add explicit depth and detail instructions where you want them. The models improved. Your prompts need to catch up.