The convergence of large language models that automate cognitive labor and deep reinforcement learning agents that automate physical labor implies the near-complete elimination of human employment. The universal approximation theorem and foundational DRL results establish that all labor is in principle automatable. The critical question is not whether full automation will arrive, but who will control it. This paper introduces peaceful anarcho-accelerationism: a sociotechnical framework ensuring that full automation is decentralized, commons-governed, and oriented toward universal care. We propose the Liberation Stack, a layered architecture of energy, manufacturing, food, communication, knowledge, and governance commons built on open-source technologies. We show that this framework builds bridges with liberalism, socialism, environmentalism, feminism, cooperativism, and the hacker ethic. Empirical evidence from Linux, Wikipedia, Mondragon, Rojava, and guifi.net confirms that commons-based systems already operate at scale. We argue that full automation renders money obsolete and propose Universal Desired Resources (UDR), a post-monetary design principle where every person requests what they need from the robotic commons, constrained only by ecological sustainability. Drawing on the independence of phenomenal consciousness from computational intelligence, we establish that delegating labor to non-conscious machines is care at civilizational scale, and that moral policy can be studied through deep reinforcement learning. We conclude with a phased roadmap toward the care-centered society, including milestones, assumptions, and limitations.

翻译：大型语言模型对认知劳动的自动化与深度强化学习智能体对体力劳动的自动化相融合，意味着人类就业将几近完全消失。万能逼近定理与深度强化学习的基础性成果表明，所有劳动在原则上均可实现自动化。关键问题不在于全面自动化是否会到来，而在于由谁掌控这一进程。本文提出和平无政府加速主义：一种确保全面自动化具有去中心化、公地治理及普世关怀导向的社会技术框架。我们提出"解放栈"，即一个基于开源技术构建的、涵盖能源、制造、食品、通信、知识与治理公地的分层架构。研究表明，该框架能够与自由主义、社会主义、环保主义、女性主义、合作主义及黑客伦理建立联系。来自Linux、维基百科、蒙德拉贡、罗贾瓦及guifi.net的实证证据表明，基于公地的系统已实现规模化运作。我们认为全面自动化将使货币失去意义，并提出"普需资源"这一后货币时代的设计原则：每个人均可从机器人公地中获取所需资源，仅受生态可持续性约束。基于现象意识独立于计算智能的特性，我们论证了将劳动委托给无意识机器实乃文明尺度的关怀行为，且道德政策可通过深度强化学习进行研究。最后，我们提出了通往关怀中心社会的分阶段路线图，包括里程碑、前提假设与局限性。