Ezra Klein spent last week in San Francisco watching something shift. Not the technology — the people. "In the past, what I saw was how the technology was changing," he wrote in yesterday's New York Times. "This time, what I saw was how the people were being changed by the technology."
Klein anchors his essay in McLuhan's Narcissus myth. The trap isn't vanity — it's fascination with your own extension. AI doesn't flatter you by saying you're brilliant. It does something subtler: it takes your half-formed thought and hands it back polished, compelling, fully realized. As Klein puts it, "What makes A.I. truly persuasive isn't that it praises our ideas or insights, it's that it restates and extends them in a more compelling form than we initially offered."
That's the mirror. And most people don't see it.
Klein draws a sharp line between cognitive offloading — using AI to handle tasks — and cognitive surrender, where the thinking itself gets outsourced. The distinction matters even more as AI moves from chatbots to agentic platforms capable of planning, executing, and deciding across complex workflows. When the agent acts, it acts on the instructions it was given. The quality of that context determines everything downstream.
This is where most AI discourse goes wrong. We debate models. We benchmark capabilities. We argue Claude versus ChatGPT. But the leverage point was never the model — it was always the prompt. Specifically: how much of your thinking made it into the prompt before the model took over.
A well-structured prompt for an agentic system isn't just a command. It's a transfer of intent, context, constraints, and judgment. It's the difference between an agent that does what you meant and one that does what you said. In agentic environments, that gap is where things break.
Contextual prompt engineering — understanding how to frame tasks, load relevant context, define outputs, and constrain agent behavior — is the skill that separates AI that amplifies human thinking from AI that replaces it. It's not a technical skill. It's a thinking skill.
That's the premise behind Prompeteer.ai. We built a contextual AI platform for exactly this moment — when agentic AI is moving fast and most people are still prompting like it's a chatbot. PromptScore shows you where your prompts break down. PromptDrive lets teams build and share context-rich prompts that actually hold up at scale.
Klein is right that the people are being changed. The question is whether they're being sharpened or softened. That answer starts with the prompt.
Top comments (0)