C

Large language models (LLMs) are beginning to create a strangely similar experience in thought. You enter a prompt (提示词) that reflects only part of what you mean. The idea may be unclear or incomplete, yet moments later, the LLM returns a paragraph that sharpens your thinking and even lands on an expression that feels surprisingly right: "That's exactly what I was thinking."

Human thinking typically follows a narrative arc (弧线). Scientists describe the moment when a confusing pattern finally resolves. Writers speak about the satisfaction of finding the right sentence after long revision. Philosophers recall the insight that arrives when a difficult argument settles into place. These moments share a common structure: an unfinished idea moves toward cognitive (认知的) insight. The eventual sense of completion-the aha moment-carries an emotional reward because it follows the effort of thinking.

LLMs can reproduce the final moment of that arc with remarkable speed. A prompt that begins as a loose question may return as a polished explanation. The response carries the language of insight, even when the process that normally produces insight never occurred.

But there is a cost to this speed. The experience of cognition itself begins to shift: LLMs approach questions very differently from humans. Where a person may encounter uncertainty or wonder whether an answer exists at all, LLMs encounter something else entirely: an incomplete pattern waiting to be filled.

When a finished explanation appears with the click of a button, the mind encounters the endpoint of reasoning before it has traveled the path that normally leads there. And when this result feels persuasive, the brain experiences the moment as recognition rather than construction. The sentence LLMs produce seems to capture what we meant to say, and the idea it expresses falls into place with a sense of familiarity. Over time, the distinction between generating a thought and adopting one may become harder to notice.

Keep in mind that none of this requires AI to possess understanding or consciousness. The effect arises from language alone. LLMs produce language that carries the structure of thought, and once that signal appears, the brain quickly fills in the rest. The system behaves less like a companion and more like a mirror, reflecting pieces of our thinking and returning them in expanded form.

This is where the deeper and more consequential question begins. For the first time in history, people are interacting with a system that can routinely supply fully expressed ideas in response to incomplete ones. AI doesn't read our thoughts; it completes patterns in language. Yet when the completion feels natural enough, the mind may experience the result as its own thinking. This moment, borrowed if not stolen from humans, may change how thinking feels.