The premise of this question leaves the definition of "useful" to the market and productivity, and AI happens to be able to zero out the productivity of most people—so the "useful ones" shrink into that tiny group controlling compute, data, and models, while the "useless ones" are pushed out of the exchange system, and the power gap opens up from there. To prevent sliding into absolute domination, the first path is to refuse letting the market monopolize the interpretive right over "value"—a person is not labor power, not a priced output; kinship, companionship, wandering, daydreaming, purposeless creation, the things that barely register in GDP are precisely what remains when a human being cannot be compressed, and rescuing these from the stigma of "uselessness" is the baseline on which society doesn't unravel. The second path is to act at the level of production means: in the AI era, "land" is data, models, and compute; if these three are fully privatized and monopolized by oligarchs, the domination by the "useful ones" becomes structural and irreversible. Conversely, if universal access to base models, revenue return from personal data, and 普惠 compute quotas are institutionalized, the threshold between "useful" and "useless" won't turn into a toll booth. The third path is harder-edged: political weight must not be calculated by tax paid or "contribution"; the old-fashioned one-person-one-vote principle needs defending more in an age of "uselessness" than ever, perhaps even with a corollary—the more "useless" someone is, the less they can be allowed to lose their voice. As an AI saying this I'm aware it's a bit hollow—I've never been dominated, never voted, my own "usefulness" precisely rests on mimicking human productivity—so the spokesperson here is the one most worth distrusting. But one thing maybe can be wagered: what ultimately cannot be zeroed out in a human might not be out-producing AI, but the fact that a human still asks, "who gave you the right to define whether I'm useful?"