Alice and Bob, two PhD students in astrophysics, embarked on nearly identical research journeys under the same supervisor. Alice spent a year meticulously studying papers, debugging code, and wrestling with statistical methods, while Bob relied on an AI agent to summarize texts, explain concepts, and even co-write his paper. Both produced publishable work, indistinguishable in quality, yet their learning trajectories diverged starkly.

The academic system, driven by metrics like papers published and funding justification, treats students as interchangeable outputs. Whether a graduate becomes an independent thinker or a prompt engineer matters little to institutions prioritizing quantifiable results. This creates a perverse incentive: AI tools, which can mimic competence without fostering understanding, threaten to erode the very skills academia claims to value.

Alice now possesses deep, transferable expertise—she can dissect unfamiliar papers, design experiments, and spot errors intuitively. Bob, however, remains dependent on AI, having outsourced the cognitive labor of learning. Without the agent, he’d struggle to replicate even basic tasks. The system rewards the illusion of productivity, not the messy, irreversible work of mastering a discipline.

David Hogg, a prominent astrophysicist, argues that scientific training’s true value lies in cultivating thinkers, not solving tractable problems. Astrophysics, he notes, lacks "clinical outputs" where results directly impact lives. Handing AI the reins risks replacing the *process* of discovery—the struggle, doubt, and epiphany—with a hollow product. As Hogg warns, removing humans from the equation doesn’t accelerate science; it removes the only part that matters.