The End of Thinking (www.derekthompson.org)

AI’s looming economic takeover narratives—“18 months until machines outcompete us”—have grabbed headlines, but this essay argues the more immediate crisis is cognitive: people are losing the capacity to read, write and think deeply because they outsource those tasks to machines. LLMs now generate essays and clinical diagnoses with ease, undermining classroom assessment and professional training. Evidence includes viral reports of widespread AI-assisted cheating, the NAEP’s 32-year low in U.S. reading scores, studies linking phone use and task-switching to poorer retention and lower GPAs, and anecdotes from medical students who stopped independently reasoning through cases when AI became the default aid. The technical and social implication is twofold: LLMs are powerful tools (sometimes better than clinicians at rare diagnoses) but indiscriminate reliance risks skill atrophy—writing is not just output, it’s a mode of thinking that trains symbolic and systems reasoning. Banning AI isn’t realistic or desirable; the challenge is designing pedagogy, professional workflows and incentives that keep human cognition “under tension” (the essay’s fitness metaphor) so deep reading, careful writing and independent judgment are preserved alongside augmentation. For AI/ML practitioners and educators, that means focusing on hybrid systems, assessment methods that probe reasoning not just output, and research into interfaces and curricula that scaffold, rather than supplant, human thought.