This article addresses a key issue in contemporary pedagogy: what and how schools should teach in an era of widespread generative artificial intelligence. The authors proceed from the premise that the new mission of schools is not to convey information per se, but to develop children’s agency—their ability to think, learn, make choices, act, and take responsibility. It is shown that AI is changing not only the tools but also the very logic of school education: it makes standard intellectual operations (retelling, model essays, formulaic solutions) less valuable and thereby increases the value of independent understanding, effort, and judgment. Based on an analysis of international studies (OECD/PISA, UNESCO, EEF, K-12 AI tutoring) and Russian data (VTsIOM, HSE University), the paper argues that the previous school model is inadequate and justifies the need to shift from a “school of answers” to a “school of judgment and action.” A practical seven-element model of themodern school is proposed; the boundaries of pedagogically appropriate AI application are described; the risks of replacing human effort with machine assistance are identified; and recommendations are formulated across three implementation horizons.

   Attention is paid to age-based differentiation in the use of AI in elementary, middle, and high school, as well as to the teacher’s new role as an architect of the educational environment, a source of professional judgment, and a mentor.

   In conclusion, a manifesto for a new school is presented, in which technology reinforces the right to education without replacing human connection, pedagogical judgment, and the process of growing up itself.