What Gets Lost in the Crossing

There are people alive today who grew up without television. A smaller number grew up without radio. Almost no one alive in prosperous countries grew up without electricity. Each generation has been the last to remember life before something that subsequent generations take as background condition. Our generation — roughly speaking, anyone born before 1995 — is the last to remember life before always-on connectivity, before search engines that answer any question in seconds, and before AI that can hold conversation, generate text, write code, and make decisions at scale. We remember what it felt like to not know something and have to live with not knowing for a while. We remember looking things up in physical reference books. We remember being unreachable. What we teach the generations behind us, and what we fail to teach them, is being decided right now without much systematic thought about it.

The Skills That Get Handed Down

Every generation transmits skills to the next, both deliberately and incidentally. The skills transmitted deliberately are those that show up in curriculum — reading, writing, arithmetic, historical knowledge, scientific method. The skills transmitted incidentally are those absorbed by living in a particular kind of world — how to navigate ambiguity, how to manage boredom, how to sustain attention through difficulty, how to disagree with someone you have to keep seeing. The AI transition poses a specific pedagogical challenge: some of the skills that matter most in an AI-abundant world are the same skills that AI makes less immediately necessary. Critical evaluation of information matters more when information is abundant and easily fabricated. But the immediate experience of having instant answers available reduces the practice of suspending judgment until sufficient evidence accumulates. The skill is more necessary but less practiced.

What Research Suggests About Transfer

The cognitive science of skill transfer is relevant here. Transfer refers to the application of skills learned in one context to different contexts — the ability to take what you learned solving one kind of problem and use it when facing a different kind. Research from Carnegie Mellon University's Human-Computer Interaction Institute examining how AI tool use affects cognitive skill development in students has found a pattern that is becoming familiar: short-term performance improves when AI assistance is available, but transfer to unassisted contexts suffers. Students who learn to write with AI writing tools write better-looking assignments than those without them. They write worse when the tool is removed. The tool improved the product without improving the skill. This matters because the skills worth transmitting are not the skills for performing in the presence of specific tools. They are the skills for thinking, evaluating, and creating under varying conditions — including conditions where the tool is not available, not reliable, or not helpful for the specific problem.

The Tangent: What Previous Last Generations Chose to Keep

The generation that was last to live without widespread electricity made choices about what to transmit that subsequent generations benefited from. Physical navigation using map and landmark. Practical mechanical skills for maintaining equipment. Food preservation techniques. Weather reading from natural signs. Some of these were transmitted successfully; others were lost. The loss of some practical skills was not clearly harmful — electricity was reliable enough that the backup skills were rarely needed. But the 20th century also saw catastrophic failures of electrical infrastructure — during wars, natural disasters, and grid failures — where communities with residual traditional skills fared better than those without them. The question for AI is similar. Which capabilities matter enough to preserve through deliberate teaching, even when AI makes the immediate exercise of those capabilities unnecessary? Navigation? Mental arithmetic? Writing from scratch? Research without algorithmic mediation? Long-form attention? The list is contestable, but the question is not trivial.

What Schools Are Actually Designed For

Schools were designed for the world that existed when they were built. Industrial-era schools were built to produce workers who could follow instructions reliably and perform repetitive tasks with accuracy. Those schools persist, in modified form, in an economy that no longer needs those workers in those numbers. Research from OECD's Education at a Glance assessment program examining curriculum alignment across member countries has found persistent lag between labor market requirements and educational content, typically running fifteen to twenty years behind. Schools are teaching for the world of fifteen years ago. They are not teaching for the world of fifteen years hence. The AI transition is compressing this problem. The skills that will matter in a world saturated with AI are different from the skills that matter in a world where AI is an emerging tool, and both are different from the skills that mattered before AI. Schools are not designed to respond at this speed.

The Responsibility We Are Leaving to the Next Generation

There is something that should be said plainly: the generation that is last to remember life before AI is making choices right now about what kind of AI world will exist. The values embedded in these systems, the governance structures built around them, the norms established in their early years — these will be inherited by people who have no memory of the alternative. That is a significant responsibility. It is not being exercised with sufficient seriousness across most of the institutions that are shaping it. The curriculum question is one part of this, and not the smallest one.