A thought experiment this chapter needs
Almost everything in this chapter is written in adult present tense. It asks what the transition does to workers, citizens, parents, and voters — people who knew a world before. This page asks a different question: what does the post-AGI world look like to the people born into it, who inherit the outcome without inheriting the memory? Much of what follows is reasoned argument rather than established finding, and it is labeled as such; where evidence exists, it is cited directly.
The question is not sentimental. Which generation holds risk, which holds power, and which holds the baseline for judging change determines what “good outcomes” even means. A transition judged a success by those who lived through it can still be a catastrophe for those who arrive after it, and the reverse.
The asymmetry of arrival
Those alive at the transition carry two burdens the next generation will not: the loss itself, and the memory of what was lost. They watched professions dissolve, institutions bend, the felt texture of work and school and friendship change. Whether that change was net good or net bad, they at least have a baseline. They can compare.
People born after the transition inherit the outcome without the grief or the baseline. To them, machine-mediated companionship is not a degraded version of something else; it is what companionship has always been. An economy where most production requires few humans is not an amputation; it is the weather. Generational analysis of earlier technological shifts — the printing press, electrification, the internet — consistently finds this pattern (much of this literature is reasoned historical argument, and that is how it should be read here): the generation that builds a technology debates it, and the generation raised inside it naturalizes it.
This asymmetry cuts both ways. The post-transition generation escapes the grief, the disorientation, and the mourning for obsolete identities. But they also lose the capacity to evaluate what they were given. A person who has never done long division cannot feel the cost of never learning it. A society that has never seen a human-only court cannot feel what was traded when judgment was delegated. The baseline is not just nostalgia; it is the instrument panel.
Growing up under an ambient machine authority
For children born after the transition, AI is not a tool they adopt. It is the environment — the tutor that answers every question, the companion that never tires, the mediator that stands between them and every other person.
The manual’s education and growing up and children and AI pages cover the immediate risks and the parenting questions. This page is about the longer arc: what attachment, trust, and authority formation look like when every formative interaction can be machine-mediated. Two worries dominate the serious literature. The first is developmental: children form expectations of minds from the minds around them, and an always-patient, always-agreeable, always-available machine teaches an unusual model of what other people are like and what relationships cost. The second is epistemic: a childhood in which every answer arrives instantly, fluent, and confident produces adults with no practiced sense of what it feels like to not know — the precondition for both learning and being deceived.
Regulators have noticed the specific danger. The EU’s AI Act prohibits systems that exploit children’s vulnerabilities or use subliminal or manipulative techniques likely to cause significant harm (Regulation (EU) 2024/1689), and UNICEF’s Policy Guidance on AI for Children (2020) sets out child-specific requirements that the wider field has been slow to operationalize. Both are worth reading as floor-markers, not solutions: they bound what may be deployed, not what a childhood inside machine mediation does to a person. Whether ambient AI authority produces resilient adults or dependent ones is, as of this writing, a genuinely open empirical question — the honest answer is that the experiment is running now, on the largest possible cohort, with no control group.
The knowledge-loss question
Every transition strands skills. The question is which ones, and whether it matters.
Consider what the last pre-transition generation uniquely holds: practiced judgment formed without machine assistance; tacit knowledge of physical systems — engines, soil, surgery, negotiation — that was learned by hand and fails in ways that are not in any manual; and, most importantly, the lived memory of what the old baseline was for. When that generation dies, the knowledge dies with it unless it has been deliberately transferred. Preservation institutions exist — the Internet Archive (archive.org) and UNESCO’s Memory of the World program (UNESCO) are the prominent examples — but they preserve documents, not capability. A scanned textbook is not a working machinist.
Whether this loss matters depends on the economy it lands in. If AI systems genuinely substitute for the lost skills, the loss is an aesthetic wound — a culture with fewer things it can do with its own hands. If the skills turn out to be load-bearing in ways the models are not — robustness, physical repair, judgment under novel conditions — the loss is structural. The honest position is that nobody knows which economy the next generation is inheriting, which is precisely why the manual treats knowledge preservation and recovery as a live discipline rather than a museum project. There is also a subtler documented effect worth naming: the early evidence on cognitive offloading suggests that delegating memory to external tools changes what people remember and how they evaluate their own knowledge (Sparrow, Liu & Wegner, “Google Effects on Memory,” Science, 2011) — a small preview, at individual scale, of what skill death looks like at civilizational scale.
The young inherit the tail risk
The sharpest intergenerational fact about AGI is distributional: the people making the decisions are old; the people who would live through the worst outcomes are young. A catastrophic failure in 2035 is a career-ending event for the executives and officials who made the calls and a life-ending event for children who made none. No consent was sought from the second group.
This is not a new structure. Climate change is the established template: harms that compound over decades, decisions made by current voters and shareholders, costs borne by people not yet born. The climate-justice literature has spent thirty years working out what this implies — intergenerational equity as a constraint, not a charity — most famously in the Brundtland Commission’s definition of sustainability as meeting present needs without compromising future generations’ ability to meet theirs (Our Common Future, 1987). The physical science makes the asymmetry concrete: the IPCC’s synthesis reports are, in effect, damage assessments for a future population that had no seat at the table (IPCC AR6 Synthesis Report).
These are parallels, stated as parallels. AI risk differs from climate risk in speed, reversibility, and shape — a misaligned transition does not degrade gradually over centuries, and there may be no later generation to appeal to. But the moral structure transfers intact: when a small group concentrates the benefits of a decision and disperses its worst-case costs across people who cannot vote, the decision needs a justification that does not depend on the beneficiaries’ consent.
What this generation owes the next
The climate literature offers one more export: obligations that are concrete rather than rhetorical. Three transfer well to the AGI transition.
Legibility. Leave a record of what was chosen and why. Future generations deserve to know which trade-offs were made deliberately and which were accidents of incentive. A transition that happens in press releases and sealed safety cases is illegible by design; legibility is a choice that costs little now and is unrecoverable later.
Capability. Education that survives model dependency. If every skill a child learns is a skill a model does better, the child has been given consumption, not capability. The test for any post-AGI curriculum is whether it produces people who can still function, think, and build when the model is absent — the standard the work and purpose page applies to adults.
Optionality. Keep real alternatives alive. A generation that grows up with exactly one way to learn, one way to work, one way to be companioned has been given a gift with a lock on it. Preserving human-taught schools, human-made judgments, and human-run institutions is not nostalgia; it is keeping the next generation’s options open against the possibility that this transition’s consensus was wrong.
None of these obligations requires knowing how the transition ends. They are robust across scenarios — and that, in a manual full of uncertainty, is the strongest claim this page makes.