To follow my rules, I'd need an answer that discusses real science (mechanisms, experiments, comparisons, caveats, etc.) so the follow-ups can open new angles grounded in that content.

I notice your question is actually a meta-observation about how I should work, not a query about frontier science itself. That said, you're right: I should ground answers in concrete mechanisms, data, and caveats so follow-ups can dig deeper.

The preprints you've provided don't directly address propulsion, relativity, or mathematical physics. Instead, they examine how well AI systems perform on physics reasoning tasks [AG-2026.02-1185, AG-2026.04-565, AG-2025.09-1139], motivations in quantum gravity research [AG-2025.05-500], and interpretive questions in quantum mechanics [AG-2025.12-373].

If you have a specific question about propulsion mechanisms, relativistic effects, or mathematical physics — the core archive topics — I'm ready to answer it with citations and concrete reasoning.