The emergence of large language models has revealed a broader problem that extends far beyond artificial intelligence itself: contemporary political discourse increasingly lacks stable conceptual foundations. Public debates often evolve into endless cycles of reinterpretation, rhetorical adjustment, and ideological reformulation, producing large volumes of commentary while generating little cumulative understanding. The same tendency can be observed in AI-generated political reasoning, where internally consistent narratives may drift away from observable reality.This paper proposes an alternative methodology based on the systematic use of prompts as explicit working assumptions. Rather than beginning with arguments, participants begin with prompts that function as provisional axioms. The objective is not to win a debate but to construct increasingly coherent prompt structures capable of organizing observation, interpretation, and judgment. New prompts emerge from previous ones, forming a dynamic framework for collective reasoning.As a test case, the paper examines the problem of contemporary Russia and the future of the Russian imperial system. Beginning from the prompt that the federalization and demilitarization of Russia should become a recognized objective of Western policy, the discussion develops a sequence of related prompts concerning imperial structures, social organization, human agency, fear, coercion, and the relationship between state preservation and human well-being. Particular attention is given to the possibility that the principal victims of imperial systems may be the populations living within them, and that conventional geopolitical discourse often obscures this perspective.The paper does not seek to provide definitive political conclusions. Instead, it explores whether prompt-based discourse can serve as a more productive and reality-oriented alternative to conventional political argumentation. The Russian question is therefore treated not only as a geopolitical problem but also as an experimental domain for investigating new forms of structured collective reasoning in the age of artificial intelligence.