A six-year-old in Dom Sierot once spent the better part of an afternoon trying to tie a knot. An older child offered help; she refused. A staff member approached; Korczak stopped him with a look. The child failed repeatedly, face moving through frustration, concentration, something like anger at the string itself, then sudden stillness — the stillness of a mind finding a new approach — and then, finally, the imperfect knot. What mattered was visible only if you knew what to look for: the child's face afterward. Not triumph, but something quieter and more durable — the recognition, felt in the body before language, that she had done a hard thing. Korczak built his framework on the observation that this experience is not merely useful for skill acquisition but constitutive of the child's selfhood. The child who struggles and succeeds does not simply acquire a skill. She acquires a knowledge of herself as capable, and that knowledge becomes the foundation on which every subsequent engagement with difficulty will be built.
AI systems deployed in children's environments in 2025 and 2026 systematically remove this resistance. The AI that solves the math problem removes the productive frustration of trying approaches that don't work and discovering, through failure, why they don't. The AI that writes the essay removes the agony of the blank page — the confrontation with one's own unclear thinking that is, as any writer knows, the mechanism by which thinking becomes clear. The AI that generates the drawing removes the gap between intention and execution in which manual skill, aesthetic judgment, and the tolerance for imperfection all develop. The efficiency gain is real. The developmental cost is invisible.
Csikszentmihalyi's research on flow provides the psychological scaffolding. Flow requires challenge matched precisely to skill — hard enough to demand full attention, not so hard it overwhelms capacity. The key insight, often missed in popular accounts: the state requires difficulty. Without difficulty, there is no flow; there is only ease, which produces not satisfaction but boredom. AI removes the challenge. Without the challenge, there is no flow. Without flow, there is no developmental experience. The child who receives the AI's output experiences relief — the task is done — but not the satisfaction of having done it herself.
Segal's ascending friction thesis in The Orange Pill — that AI does not eliminate difficulty but relocates it to a higher cognitive floor — contains a developmental caveat his framework, oriented toward adult professionals, does not fully address. The surgeon who operates laparoscopically first learned to feel tissue with her hands. The senior engineer whose architectural intuition guides AI-assisted work first spent years debugging lower layers of the stack. The ascending friction thesis assumes a foundation of embodied experience. For children, that foundation has not yet been laid. The twelve-year-old cannot ascend to higher friction if she has not experienced the lower friction first.
The right generates specific obligations: on parents, to resist smoothing every difficulty from the child's path; on educators, to design environments that preserve struggle even when tools can eliminate it; on AI designers, to build systems that know when to step back — that identify moments when the child's own process is more valuable than the system's output, and withhold the output long enough for the process to run.
The principle runs through every institutional structure at Dom Sierot. Children cooked, cleaned, managed logistics, resolved their own conflicts, governed themselves democratically, published their own newspaper. The implicit rationale: children permitted to struggle with real responsibilities — not simulated exercises — develop capacities no instruction can produce. The children's court is the sharpest example: disputes adjudicated by a rotating panel of children, with verdicts carrying real consequences. The process demanded capacities that could only be developed through the struggle of exercising them.
Struggle as constitutive. The child tying the knot does not merely acquire a skill; she acquires the self that emerges from the struggle to acquire it.
Relief vs. satisfaction. Receiving the AI's output produces relief; doing the work oneself produces a categorically different experience of satisfaction that builds the self.
The developmental foundation gap. Ascending friction works for adults with embodied foundations already laid; children have not yet laid those foundations, and cannot build upward without them.
Calibrated difficulty, not suffering. The right is to productive difficulty — developmentally appropriate challenge, not frustration for its own sake.
Defenders of AI-assisted learning point to evidence that scaffolding tools reduce frustration-induced dropout, particularly for children with learning differences, and that the strict Korczakian line risks privileging middle-class children who can afford prolonged struggle. The framework's response: the distinction between scaffolding (which supports struggle) and substitution (which replaces it) is not fuzzy but architectural, and good design can preserve the first while eliminating the second.