The Next Cognitive Leap

#In his book Range (opens in a new tab), David Epstein recounts a fascinating natural experiment conducted by the psychologist Alexander Luria (opens in a new tab) in the Soviet Union during the 1920s and 1930s. Luria observed a profound difference in the cognitive frameworks of uneducated peasants compared to those who had received formal schooling. The illiterate peasants thought strictly in practical, immediate terms, bound by their tangible reality. In contrast, those exposed to literacy and modern technology developed the capacity for abstract reasoning, categorical thinking and hypothetical logic. Education and technology did not just fill their minds with new facts; they fundamentally rewrote their cognitive “operating system.”

#We are running that experiment again, much faster, and this time the technology isn’t the alphabet. It’s the AI.

#When a machine can draft the report, write the code or run the analysis in seconds, processing stops being the scarce skill. What’s scarce now is intention: deciding what is worth asking, setting up the context and knowing what a good answer looks like before you see one. The literate peasant learned to think in categories instead of things. We are learning to think in outcomes instead of steps.

#That doesn’t make us shallower. In September, OpenAI announced a proof (opens in a new tab) on the Navier–Stokes problem, produced by thousands of AI agents. Then the hard part started: mathematicians are still arguing about what it shows and whether anyone can follow it. A pattern-matching engine can produce an answer, but it can’t vouch for one. Simon Willison (opens in a new tab) suggests that journalists, who spend their careers working with unreliable sources, may be better prepared for AI than most professions. Human depth now lives in verification, judgement and a sense of what a result means for real people.

#The shift also rewards breadth. In the same conversation, Willison observes that a human expert masters two or three domains in a lifetime, while a model has read across all of them. Once deep expertise is on tap, the generalist matters more: the person who can borrow an idea from biology, test it against behavioural economics and ask the model to build it. Epstein made his case for range before ChatGPT existed, and it reads differently now.

#Luria’s villagers fell into two groups: those who had been to school and those who hadn’t. The divide opening now has more layers. Some people have no access to these tools at all. Some have the Iron Man suit but wear it like an overcoat. And a few have learned to fly it, aiming far higher than seems reasonable, as Felix Rieseberg (opens in a new tab) urges.

#Literacy taught us to reason about worlds we had never seen. What will this leap teach us to do, and who will be left on the ground when it does?