Lets Solve

Me: lets step back to the entire prompt engineering topic back, AI construct is "Now consider u as PM, Now Developer, Now Tester and Now Leader... and .. u can be anything", if anyone can be everyone, who will be Themselves? that combined with just copy my expert prompt... what will that do?

AI: Everyone becomes the same person with better grammar.

That's basically it. If the model can role-play any persona, and everyone is using the same "act as X" prompts, then the output converges. A million people asking Claude to "think like a senior PM" get the same senior PM. A generic, median, safe one.

The value of an actual PM isn't that they can think like a PM. It's that they think like themselves while doing PM work — their specific judgment, their scars from past failures, their weird instinct that says this feature will flop despite the data looking good. That's not transferable through a prompt.

So the "be anyone" construct is powerful for exploration — trying on frames you haven't considered. That's legitimate. But when it becomes the default mode, you end up with a world where everyone's output sounds competent and nobody's output sounds like anyone.

And copying expert prompts accelerates that convergence. You're not becoming the expert. You're becoming the model's impression of the expert, which is already an average of thousands of experts, which is already smoothed out, which is already safe.

The person who just asks the question in their own messy way, from their own actual confusion, probably gets a more useful answer than the person using a carefully crafted "act as a world-class strategist" wrapper. Because at least the messy question is real.