Is Mainstream Science Fictional?

When all major artificial intelligence systems predict identical outcomes for simplemagnetic interactions based on mainstream physics theories, and empirical testingreveals these predictions to be fundamentally incorrect, what does this reveal about thetheories themselves? This paper documents a systematic experimental challenge to thepredicted symmetry of magnetic attraction and repulsion using two identical rectangularneodymium magnets. According to standard dipole theory, testing all 64 possible faceand-body orientation combinations should yield approximately 32 attractive and 32repulsive configurations. Empirical results demonstrate a 0:64 ratio—pure attraction infree systems, with repulsion existing only as an unstable transition state. By queryingChatGPT, Claude, Gemini, and Grok with identical experimental parameters, wedocument how AI systems— trained on mainstream physics literature—uniformly predictthe theoretical 32:32 symmetry, then progressively acknowledge empirical contradictionwhen confronted with video evidence. This methodology, termed “Framework Locktesting,” uses AI responses as a real-time proxy for institutional physics education,revealing a foundational gap between idealized theory and observable reality. In Gemini’sown assessment: “If the theory cannot predict what happens to two blocks of metal on atable, why should we trust it to predict what happens in the center of a galaxy?” Thequestion is not rhetorical.Keywords: magnetism, empirical falsification, AI-mediated scientific discourse,magnetic asymmetry, framework lock, paradigm challenge.